test(handlers): add coverage for BroadcastHandler (workspace_broadcast.go)

Add 13 tests covering every path in the POST /workspaces/:id/broadcast handler and its unexported broadcastTruncate helper: broadcastTruncate: - short string → unmodified - exactly max length → unmodified - exceeds max length → truncated + ellipsis - unicode runes → rune-boundary truncation Broadcast: - invalid workspace ID → 400 - missing message → 400 - workspace not found → 404 - broadcast_disabled=true → 403 with hint - recipient query fails → 500 - no recipients → 200 delivered=0 - one recipient → 200 delivered=1 - recipient insert fails → 200 delivered=0 (best-effort continue) - sender log fails → 200 delivered=1 (best-effort, non-blocking) Uses sqlmock for the DB layer and newTestBroadcaster() for the *events.Broadcaster injection. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-16 22:11:47 +00:00
974 changed files with 86953 additions and 61587 deletions
@@ -50,8 +50,11 @@ MOLECULE_ENV=development                       # Environment label (development/
 # Container/runtime detection
 # MOLECULE_IN_DOCKER=                    # Set when running the platform inside Docker (accepts 1/0, true/false). Triggers A2A proxy to rewrite 127.0.0.1:<port> agent URLs to Docker bridge hostnames. Auto-detected via /.dockerenv; only set if detection fails or to force off.

+# Observability (Awareness)
+# AWARENESS_URL=                         # If set, injected into workspace containers along with a deterministic AWARENESS_NAMESPACE derived from workspace ID. Enables the cross-session memory MCP server.
+
 # GitHub
-# GITHUB_REPO=owner/repo                 # Target repo for agent initial_prompt clone (e.g. Molecule-AI/molecule-core). Read inside workspace containers.
+# GITHUB_REPO=owner/repo                 # Target repo for agent initial_prompt clone (e.g. Molecule-AI/molecule-monorepo). Read inside workspace containers.
 # GITHUB_TOKEN=                          # Personal access token / installation token used by agents that clone private repos. Register as a global secret via POST /admin/secrets for propagation to workspace env. Token is used in-URL during clone and then scrubbed from .git/config via `git remote set-url`.

 # Webhooks
@@ -1 +0,0 @@
-refire:1778784369
@@ -18,24 +18,15 @@
 # per §SOP-6 security model). No-op when merged=false.
 #
 # Required env (set by the workflow):
-#   GITEA_TOKEN, GITEA_HOST, REPO, PR_NUMBER
-#   plus one of REQUIRED_CHECKS_JSON (preferred) or REQUIRED_CHECKS (legacy)
+#   GITEA_TOKEN, GITEA_HOST, REPO, PR_NUMBER, REQUIRED_CHECKS
 #
-# REQUIRED_CHECKS_JSON is a JSON object keyed by branch name. Each value
-# is an array of status-check context names that branch protection
-# requires for that branch. The script looks up the PR's base branch and
-# evaluates only the checks declared for that branch.
-#
-#   {"main": ["CI / all-required (pull_request)", ...],
-#    "staging": ["CI / all-required (pull_request)", ...]}
-#
-# REQUIRED_CHECKS (legacy) is a newline-separated list used when the
-# JSON variable is not set. Declared in the workflow YAML rather than
-# fetched from /branch_protections (which needs admin scope — sop-tier-bot
-# has read-only). Trade dynamism for simplicity: when the required-check
-# set changes, update both branch protection AND this env. Keeping them
-# in sync is less complexity than granting the audit bot admin perms on
-# every repo.
+# REQUIRED_CHECKS is a newline-separated list of status-check context
+# names that branch protection requires. Declared in the workflow YAML
+# rather than fetched from /branch_protections (which needs admin
+# scope — sop-tier-bot has read-only). Trade dynamism for simplicity:
+# when the required-check set changes, update both branch protection
+# AND this env. Keeping them in sync is less complexity than granting
+# the audit bot admin perms on every repo.

 set -euo pipefail

@@ -43,10 +34,7 @@ set -euo pipefail
 : "${GITEA_HOST:?required}"
 : "${REPO:?required}"
 : "${PR_NUMBER:?required}"
-if [ -z "${REQUIRED_CHECKS_JSON:-}" ] && [ -z "${REQUIRED_CHECKS:-}" ]; then
-  echo "::error::Either REQUIRED_CHECKS_JSON or REQUIRED_CHECKS must be set"
-  exit 1
-fi
+: "${REQUIRED_CHECKS:?required (newline-separated context names)}"

 OWNER="${REPO%%/*}"
 NAME="${REPO##*/}"
@@ -77,14 +65,10 @@ if [ -z "$MERGE_SHA" ]; then
  exit 0
 fi

-# 2. Required status checks — branch-aware JSON dict takes precedence.
-if [ -n "${REQUIRED_CHECKS_JSON:-}" ]; then
-  REQUIRED=$(echo "$REQUIRED_CHECKS_JSON" | jq -r --arg branch "$BASE_BRANCH" '.[$branch] // [] | .[]')
-else
-  REQUIRED="$REQUIRED_CHECKS"
-fi
+# 2. Required status checks declared in the workflow env.
+REQUIRED="$REQUIRED_CHECKS"
 if [ -z "${REQUIRED//[[:space:]]/}" ]; then
-  echo "::notice::REQUIRED_CHECKS empty for branch '$BASE_BRANCH' — force-merge not applicable."
+  echo "::notice::REQUIRED_CHECKS empty — force-merge not applicable."
  exit 0
 fi

@@ -203,17 +203,12 @@ def ci_jobs_all(ci_doc: dict) -> set[str]:

 def ci_job_names(ci_doc: dict) -> set[str]:
    """Set of job keys in ci.yml MINUS the sentinel itself MINUS jobs
-    whose `if:` gates on `github.event_name` or `github.ref` (those are
-    event-scoped and can legitimately be `skipped` for a given trigger;
-    if we required them under the sentinel `needs:`, every PR-only job
+    whose `if:` gates on `github.event_name` (those are event-scoped
+    and can legitimately be `skipped` for a given trigger; if we
+    required them under the sentinel `needs:`, every PR-only job
    would be `skipped` on push and the sentinel would interpret
    `skipped != success` as failure). RFC §4 spec.

-    `github.ref` is the companion gate for jobs that run only on direct
-    pushes to specific branches (e.g. `github.ref == 'refs/heads/main'`).
-    These never execute in a PR context, so flagging them as missing
-    from `all-required.needs:` is a false positive (mc#958 / mc#959).
-
    Used for F1 (jobs missing from sentinel needs). NOT used for F1b
    (typos in needs) — see `ci_jobs_all` for that."""
    jobs = ci_doc.get("jobs")
@@ -226,9 +221,7 @@ def ci_job_names(ci_doc: dict) -> set[str]:
            continue
        if isinstance(v, dict):
            gate = v.get("if")
-            if isinstance(gate, str) and (
-                "github.event_name" in gate or "github.ref" in gate
-            ):
+            if isinstance(gate, str) and "github.event_name" in gate:
                continue
        names.add(k)
    return names
@@ -274,8 +267,7 @@ def required_checks_env(audit_doc: dict) -> set[str]:
                    found.append(v)
    if not found:
        sys.stderr.write(
-            f"::error::REQUIRED_CHECKS env not found in any step of "
-            f"{AUDIT_WORKFLOW_PATH}\n"
+            f"::error::REQUIRED_CHECKS env not found in any step of {AUDIT_WORKFLOW_PATH}\n"
        )
        sys.exit(3)
    if len(found) > 1:
@@ -385,15 +377,10 @@ def detect_drift(branch: str) -> tuple[list[str], dict]:
    contexts = set(protection.get("status_check_contexts") or [])

    # ----- F1: job exists in CI but not under sentinel.needs -----
-    # Post-#1766 contract: the sentinel may deliberately have no `needs:`
-    # and instead poll path-relevant statuses dynamically. In that case
-    # F1 is a false positive — skip it. F1b (typos in existing needs)
-    # is naturally skipped when needs is empty.
    missing_from_needs = sorted(jobs - needs)
-    if missing_from_needs and needs:
+    if missing_from_needs:
        findings.append(
-            "F1 — jobs in ci.yml NOT under sentinel `needs:` "
-            "(sentinel doesn't gate them):\n"
+            "F1 — jobs in ci.yml NOT under sentinel `needs:` (sentinel doesn't gate them):\n"
            + "\n".join(f"  - {n}" for n in missing_from_needs)
        )

@@ -403,8 +390,7 @@ def detect_drift(branch: str) -> tuple[list[str], dict]:
    stale_needs = sorted(needs - jobs_all)
    if stale_needs:
        findings.append(
-            "F1b — sentinel `needs:` lists jobs NOT present in ci.yml "
-            "(typo or removed job):\n"
+            "F1b — sentinel `needs:` lists jobs NOT present in ci.yml (typo or removed job):\n"
            + "\n".join(f"  - {n}" for n in stale_needs)
        )

@@ -412,9 +398,7 @@ def detect_drift(branch: str) -> tuple[list[str], dict]:
    # Compute the contexts the CI YAML actually produces. The sentinel
    # is in (B) intentionally (`ci / all-required (pull_request)`); we
    # whitelist it explicitly.
-    emitted_contexts = {
-        expected_context(j) for j in jobs
-    } | {expected_context(SENTINEL_JOB)}
+    emitted_contexts = {expected_context(j) for j in jobs} | {expected_context(SENTINEL_JOB)}
    # Contexts NOT produced by ci.yml may still come from other
    # workflows in the repo (Secret scan etc). We can't enumerate
    # every workflow's emissions cheaply; instead, flag only contexts
@@ -427,9 +411,8 @@ def detect_drift(branch: str) -> tuple[list[str], dict]:
    )
    if stale_protection:
        findings.append(
-            "F2 — protection `status_check_contexts` entries with `ci / ` "
-            "prefix that NO job in ci.yml emits "
-            "(stale name → silent advisory gate):\n"
+            "F2 — protection `status_check_contexts` entries with `ci / ` prefix that NO "
+            "job in ci.yml emits (stale name → silent advisory gate):\n"
            + "\n".join(f"  - {c}" for c in stale_protection)
        )

@@ -504,8 +487,7 @@ def render_body(branch: str, findings: list[str], debug: dict) -> str:
        f"# Drift detected on `{REPO}/{branch}`",
        "",
        "Auto-filed by `.gitea/workflows/ci-required-drift.yml` "
-        "(RFC [internal#219]"
-        "(https://git.moleculesai.app/molecule-ai/internal/issues/219) §4 + §6).",
+        "(RFC [internal#219](https://git.moleculesai.app/molecule-ai/internal/issues/219) §4 + §6).",
        "",
        "## Findings",
        "",
@@ -516,11 +498,8 @@ def render_body(branch: str, findings: list[str], debug: dict) -> str:
            "",
            "## Resolution",
            "",
-            "- **F1 / F1b**: if the sentinel job has a `needs:` block, add "
-            "the missing job to it in `.gitea/workflows/ci.yml`, or remove "
-            "the stale entry. If the sentinel deliberately has no `needs:` "
-            "(path-aware polling sentinel per post-#1766 contract), this "
-            "finding is expected and F1 is skipped.",
+            "- **F1 / F1b**: add the missing job to `all-required.needs:` "
+            "in `.gitea/workflows/ci.yml`, or remove the stale entry.",
            "- **F2**: rename the protection context to match an emitter, "
            "or remove it from `status_check_contexts` "
            "(PATCH `/api/v1/repos/{owner}/{repo}/branch_protections/{branch}`).",
@@ -561,12 +540,12 @@ def file_or_update(

    if dry_run:
        print(f"::notice::[dry-run] would file/update drift issue for {branch}")
-        print("::group::[dry-run] title")
+        print(f"::group::[dry-run] title")
        print(title)
-        print("::endgroup::")
-        print("::group::[dry-run] body")
+        print(f"::endgroup::")
+        print(f"::group::[dry-run] body")
        print(body)
-        print("::endgroup::")
+        print(f"::endgroup::")
        return

    existing = find_open_issue(title)
@@ -1,176 +0,0 @@
-#!/usr/bin/env python3
-"""Shared path-filter helper for Gitea Actions workflows.
-
-Computes changed files against the PR base SHA or push-before SHA and writes
-boolean outputs to GITHUB_OUTPUT. If the diff base is missing or untrusted, the
-helper fails open by setting every output in the selected profile to true.
-"""
-
-from __future__ import annotations
-
-import argparse
-import os
-import re
-import subprocess
-import sys
-from pathlib import Path
-
-PROFILES: dict[str, dict[str, str]] = {
-    "ci": {
-        "platform": r"^workspace-server/",
-        "canvas": r"^canvas/",
-        "python": r"^workspace/",
-        "scripts": r"^tests/e2e/|^scripts/|^infra/scripts/",
-    },
-    "handlers-postgres": {
-        "handlers": (
-            r"^workspace-server/internal/handlers/"
-            r"|^workspace-server/internal/wsauth/"
-            r"|^workspace-server/migrations/"
-            r"|^\.gitea/workflows/handlers-postgres-integration\.yml$"
-        ),
-    },
-    "e2e-api": {
-        "api": r"^workspace-server/|^tests/e2e/|^\.gitea/workflows/e2e-api\.yml$",
-    },
-}
-
-
-def classify(profile: str, paths: list[str]) -> dict[str, bool]:
-    patterns = PROFILES[profile]
-    return {
-        name: any(re.search(pattern, path) for path in paths)
-        for name, pattern in patterns.items()
-    }
-
-
-def all_true(profile: str) -> dict[str, bool]:
-    return {name: True for name in PROFILES[profile]}
-
-
-def resolve_base(event_name: str, pr_base_sha: str, push_before: str) -> str:
-    if event_name == "pull_request" and pr_base_sha:
-        return pr_base_sha
-    return push_before
-
-
-def is_zero_sha(value: str) -> bool:
-    return not value or bool(re.fullmatch(r"0+", value))
-
-
-def run_git(args: list[str], *, timeout: int = 30) -> subprocess.CompletedProcess[str]:
-    return subprocess.run(
-        ["git", *args],
-        check=False,
-        text=True,
-        stdout=subprocess.PIPE,
-        stderr=subprocess.PIPE,
-        timeout=timeout,
-    )
-
-
-def base_exists(base: str) -> bool:
-    return run_git(["cat-file", "-e", base]).returncode == 0
-
-
-def fetch_base(base: str, base_ref: str) -> None:
-    # Gitea may reject fetching an arbitrary unadvertised SHA from a shallow
-    # PR checkout. Fetch the advertised base branch first, then fall back to
-    # the SHA for hosts that allow it.
-    if base_ref:
-        run_git(["fetch", "--depth=1", "origin", base_ref])
-    if not base_exists(base):
-        run_git(["fetch", "--depth=1", "origin", base])
-
-
-def deepen_base_ref(base_ref: str) -> None:
-    if base_ref:
-        run_git(["fetch", "--deepen=200", "origin", base_ref], timeout=60)
-
-
-def merge_base(base: str) -> str | None:
-    proc = run_git(["merge-base", base, "HEAD"])
-    if proc.returncode != 0:
-        return None
-    value = proc.stdout.strip()
-    return value or None
-
-
-def changed_paths(base: str, *, use_merge_base: bool) -> list[str] | None:
-    compare_base = base
-    if use_merge_base:
-        compare_base = merge_base(base) or ""
-        if not compare_base:
-            return None
-
-    proc = run_git(["diff", "--name-only", compare_base, "HEAD"])
-    if proc.returncode != 0:
-        return None
-    return [line for line in proc.stdout.splitlines() if line]
-
-
-def write_outputs(values: dict[str, bool], output_path: str | None) -> None:
-    lines = [f"{name}={'true' if value else 'false'}" for name, value in values.items()]
-    if output_path:
-        with Path(output_path).open("a", encoding="utf-8") as fh:
-            for line in lines:
-                fh.write(line + "\n")
-    else:
-        for line in lines:
-            print(line)
-
-
-def detect(
-    profile: str,
-    event_name: str,
-    pr_base_sha: str,
-    push_before: str,
-    base_ref: str = "",
-) -> dict[str, bool]:
-    base = resolve_base(event_name, pr_base_sha, push_before)
-    if is_zero_sha(base):
-        return all_true(profile)
-
-    if not base_exists(base):
-        fetch_base(base, base_ref)
-    if not base_exists(base):
-        return all_true(profile)
-
-    use_merge_base = event_name == "pull_request"
-    if use_merge_base and base_ref and merge_base(base) is None:
-        deepen_base_ref(base_ref)
-
-    paths = changed_paths(base, use_merge_base=use_merge_base)
-    if paths is None:
-        return all_true(profile)
-    return classify(profile, paths)
-
-
-def parse_args(argv: list[str]) -> argparse.Namespace:
-    parser = argparse.ArgumentParser(description=__doc__)
-    parser.add_argument("--profile", required=True, choices=sorted(PROFILES))
-    parser.add_argument("--event-name", default=os.environ.get("GITHUB_EVENT_NAME", ""))
-    parser.add_argument("--pr-base-sha", default="")
-    parser.add_argument("--base-ref", default="")
-    parser.add_argument(
-        "--push-before",
-        default=os.environ.get("GITHUB_EVENT_BEFORE", ""),
-    )
-    return parser.parse_args(argv)
-
-
-def main(argv: list[str]) -> int:
-    args = parse_args(argv)
-    values = detect(
-        args.profile,
-        args.event_name,
-        args.pr_base_sha,
-        args.push_before,
-        args.base_ref,
-    )
-    write_outputs(values, os.environ.get("GITHUB_OUTPUT"))
-    return 0
-
-
-if __name__ == "__main__":
-    sys.exit(main(sys.argv[1:]))
@@ -65,11 +65,6 @@ class ApiError(RuntimeError):
    pass


-class MergePermissionError(ApiError):
-    """Merge failed with a permanent permission error (403/404/405).
-    The queue should skip this PR and move to the next one."""
-
-
@dataclasses.dataclass(frozen=True)
 class MergeDecision:
    ready: bool
@@ -153,40 +148,15 @@ def latest_statuses_by_context(statuses: list[dict]) -> dict[str, dict]:
    return latest


-def _is_tier_low_pending_ok(
-    latest_statuses: dict[str, dict],
-    context: str,
-    pr_labels: set[str],
-) -> bool:
-    """Return True if tier:low PR can tolerate sop-checklist pending state.
-
-    Per sop-checklist-config.yaml tier_failure_mode, tier:low uses soft-fail:
-    sop-checklist posts state=pending when acks are satisfied (missing
-    manager/ceo acks are informational only). The queue should accept
-    pending instead of waiting for success.
-    """
-    if "tier:low" not in pr_labels:
-        return False
-    if "sop-checklist" not in context:
-        return False
-    status = latest_statuses.get(context) or {}
-    return status_state(status) == "pending"
-
-
 def required_contexts_green(
    latest_statuses: dict[str, dict],
    contexts: list[str],
-    pr_labels: set[str] | None = None,
 ) -> tuple[bool, list[str]]:
    missing_or_bad: list[str] = []
    for context in contexts:
        status = latest_statuses.get(context)
        state = status_state(status or {})
        if state != "success":
-            if pr_labels and _is_tier_low_pending_ok(
-                latest_statuses, context, pr_labels
-            ):
-                continue  # tier:low soft-fail: accept pending sop-checklist
            missing_or_bad.append(f"{context}={state or 'missing'}")
    return not missing_or_bad, missing_or_bad

@@ -239,7 +209,6 @@ def evaluate_merge_readiness(
    pr_status: dict,
    required_contexts: list[str],
    pr_has_current_base: bool,
-    pr_labels: set[str] | None = None,
 ) -> MergeDecision:
    # Check push-required contexts explicitly instead of combined state.
    # Combined state can be "failure" due to non-blocking jobs
@@ -259,7 +228,7 @@ def evaluate_merge_readiness(
    # The required_contexts list is the authoritative gate — it includes only
    # the checks that actually block merges.
    latest = latest_statuses_by_context(pr_status.get("statuses") or [])
-    ok, missing_or_bad = required_contexts_green(latest, required_contexts, pr_labels)
+    ok, missing_or_bad = required_contexts_green(latest, required_contexts)
    if not ok:
        return MergeDecision(False, "wait", "required contexts not green: " + ", ".join(missing_or_bad))
    return MergeDecision(True, "merge", "ready")
@@ -284,32 +253,27 @@ def get_combined_status(sha: str) -> dict:
    _, combined = api("GET", f"/repos/{OWNER}/{NAME}/commits/{sha}/status")
    if not isinstance(combined, dict):
        raise ApiError(f"status for {sha} response not object")
-    combined_statuses: list[dict] = combined.get("statuses") or []
+    # Fetch full statuses list; 200 covers >99% of real-world runs.
+    # The list is ordered ascending by id (oldest first) — callers must
+    # iterate in reverse to get the newest entry per context.
+    # Best-effort: large repos (main with 550+ statuses) may time out.
+    # On timeout, fall back to the statuses[] already in the combined
+    # response (usually 30 entries — enough for most PRs, enough for
+    # main's early push-required contexts).
    try:
-        _, all_statuses_raw = api(
+        _, all_statuses = api(
            "GET",
            f"/repos/{OWNER}/{NAME}/commits/{sha}/statuses",
            query={"limit": "50"},
        )
-        if isinstance(all_statuses_raw, list):
-            all_statuses: list[dict] = list(all_statuses_raw)
-        else:
-            all_statuses = []
+        if isinstance(all_statuses, list):
+            combined["statuses"] = all_statuses
    except (ApiError, urllib.error.URLError, TimeoutError, OSError) as exc:
+        # URLError covers network-level failures (DNS, refused, timeout).
+        # TimeoutError and OSError cover socket-level timeouts.
        sys.stderr.write(f"::warning::could not fetch full statuses list for {sha[:8]}: {exc}\n")
-        all_statuses = []
-    # Build latest per context: process combined (ascending→reverse=newest
-    # first), then fill gaps from all_statuses (already newest-first).
-    latest: dict[str, dict] = {}
-    for status in reversed(sorted(combined_statuses, key=lambda s: s.get("id") or 0)):
-        ctx = status.get("context")
-        if isinstance(ctx, str) and ctx not in latest:
-            latest[ctx] = status
-    for status in all_statuses:
-        ctx = status.get("context")
-        if isinstance(ctx, str) and ctx not in latest:
-            latest[ctx] = status
-    combined["statuses"] = list(latest.values())
+        # Fall back to the statuses[] already in the combined response.
+        pass
    return combined


@@ -374,16 +338,7 @@ def merge_pull(pr_number: int, *, dry_run: bool) -> None:
    print(f"::notice::merging PR #{pr_number}")
    if dry_run:
        return
-    try:
-        api("POST", f"/repos/{OWNER}/{NAME}/pulls/{pr_number}/merge", body=payload, expect_json=False)
-    except ApiError as exc:
-        # Re-raise permission-like errors so process_once can skip this PR.
-        # 403 = no push access, 404 = repo/pr not found, 405 = not allowed.
-        msg = str(exc)
-        for code in ("403", "404", "405"):
-            if code in msg:
-                raise MergePermissionError(msg) from exc
-        raise  # re-raise other ApiErrors unchanged
+    api("POST", f"/repos/{OWNER}/{NAME}/pulls/{pr_number}/merge", body=payload, expect_json=False)


 def process_once(*, dry_run: bool = False) -> int:
@@ -425,13 +380,11 @@ def process_once(*, dry_run: bool = False) -> int:
    commits = get_pull_commits(pr_number)
    current_base = pr_has_current_base(pr, commits, main_sha)
    pr_status = get_combined_status(head_sha)
-    pr_labels = label_names(pr)
    decision = evaluate_merge_readiness(
        main_status=main_status,
        pr_status=pr_status,
        required_contexts=contexts,
        pr_has_current_base=current_base,
-        pr_labels=pr_labels,
    )

    print(f"::notice::PR #{pr_number} decision={decision.action}: {decision.reason}")
@@ -454,25 +407,7 @@ def process_once(*, dry_run: bool = False) -> int:
                "deferring to next tick"
            )
            return 0
-        try:
-            merge_pull(pr_number, dry_run=dry_run)
-        except MergePermissionError as exc:
-            # Permanent merge failure (HTTP 403/404/405). Post a comment so
-            # maintainers know why, then return 0 so this tick is done.
-            # The PR stays in the queue; future ticks can retry after the
-            # permission issue is resolved.
-            sys.stderr.write(f"::error::merge permission error for PR #{pr_number}: {exc}\n")
-            post_comment(
-                pr_number,
-                (
-                    "merge-queue: merge failed with HTTP 405 'User not allowed to merge PR'. "
-                    "No available token has Can-merge permission on this repo. "
-                    "Fix: grant Can-merge to a token, or add a maintain/admin collaborator. "
-                    "Skipping to next queued PR on next tick."
-                ),
-                dry_run=dry_run,
-            )
-            return 0
+        merge_pull(pr_number, dry_run=dry_run)
        return 0
    return 0

@@ -482,21 +417,7 @@ def main() -> int:
    parser.add_argument("--dry-run", action="store_true")
    args = parser.parse_args()
    _require_runtime_env()
-    try:
-        return process_once(dry_run=args.dry_run)
-    except ApiError as exc:
-        # API errors (401/403/404/500) are transient for a queue tick —
-        # log and exit 0 so the workflow is not marked failed and the next
-        # tick can retry. Returning non-zero would permanently fail the
-        # workflow run, blocking future ticks.
-        sys.stderr.write(f"::error::queue API error: {exc}\n")
-        return 0
-    except urllib.error.URLError as exc:
-        sys.stderr.write(f"::error::queue network error: {exc}\n")
-        return 0
-    except TimeoutError as exc:
-        sys.stderr.write(f"::error::queue timeout: {exc}\n")
-        return 0
+    return process_once(dry_run=args.dry_run)


 if __name__ == "__main__":
@@ -13,9 +13,11 @@ from __future__ import annotations
 import argparse
 import glob
 import re
+import sys
 from pathlib import Path
 from typing import NamedTuple

+
 SELF = ".gitea/workflows/lint-curl-status-capture.yml"


@@ -283,7 +283,7 @@ def _ensure_labels(repo: str, names: list[str]) -> list[int]:
    if status != "ok" or not isinstance(labels, list):
        return []
    out: list[int] = []
-    by_name = {label["name"]: label["id"] for label in labels if isinstance(label, dict)}
+    by_name = {l["name"]: l["id"] for l in labels if isinstance(l, dict)}
    for n in names:
        if n in by_name:
            out.append(by_name[n])
@@ -82,7 +82,7 @@ import sys
 import urllib.error
 import urllib.parse
 import urllib.request
-from datetime import datetime, timezone
+from datetime import datetime, timedelta, timezone
 from pathlib import Path
 from typing import Any

@@ -641,15 +641,6 @@ def main(argv: list[str] | None = None) -> int:

    base_workflows = workflows_at_sha(BASE_SHA)
    head_workflows = workflows_at_sha(HEAD_SHA)
-    # Ignore workflow files that are identical on both sides — old branches
-    # that haven't rebased onto main carry stale copies of workflows that
-    # were updated later. Comparing those stale copies against the current
-    # base produces false-positive "flips".
-    base_workflows = {
-        p: t for p, t in base_workflows.items()
-        if p in head_workflows and head_workflows[p] != t
-    }
-    head_workflows = {p: t for p, t in head_workflows.items() if p in base_workflows}
    flips = detect_flips(base_workflows, head_workflows)

    if not flips:
@@ -61,7 +61,6 @@ import os
 import shutil
 import subprocess
 import sys
-import time
 import urllib.error
 import urllib.parse
 import urllib.request
@@ -90,28 +89,6 @@ API = f"https://{GITEA_HOST}/api/v1" if GITEA_HOST else ""
 # match by exact title without parsing.
 TITLE_PREFIX = "[main-red]"

-# Contexts that are scheduled or non-required — their pending/failure
-# state should not block stale-issue closeout (mc#1789).
-SCHEDULED_CONTEXT_PATTERNS = (
-    "Staging SaaS smoke",
-    "Continuous synthetic E2E",
-    "main-red-watchdog",
-    "ci-arm64-advisory",
-)
-
-# Settling window (seconds) between initial red detection and the
-# pre-file recheck. The recheck filters out the two largest false-
-# positive classes seen in mc#1597..1630 (task #394, 2026-05-21):
-#   1. HEAD moved on (a new commit landed mid-tick) — the prior red SHA
-#      is no longer authoritative; let the next cron tick re-evaluate.
-#   2. Combined status recovered on the SAME SHA (transient
-#      cancel-cascade rolled forward to success on retry).
-# 90s is well below the hourly cron cadence; a real failure that
-# persists past it is the one we want surfaced.
-# Override with WATCHDOG_RECHECK_DELAY_SECS for tests / local probes
-# (the test suite stubs time.sleep to a no-op).
-RECHECK_DELAY_SECS = int(_env("WATCHDOG_RECHECK_DELAY_SECS", default="90"))
-

 def _require_runtime_env() -> None:
    """Enforce env contract — called from `main()` only.
@@ -195,49 +172,6 @@ def api(
        return status, {"_raw": raw.decode("utf-8", errors="replace")}


-# --------------------------------------------------------------------------
-# action_run.status resolver — extensibility hook for task #394.
-# --------------------------------------------------------------------------
-def _resolve_action_run_status(target_url: str) -> int | None:
-    """Resolve the underlying Gitea `action_run.status` integer for the
-    run referenced by `target_url`, returning None if the resolver
-    cannot reach an authoritative source from the runner.
-
-    Canonical Gitea 1.22.6 enum (per `models/actions/status.go` +
-    `reference_gitea_action_status_enum_corrected_2026_05_19`):
-        1=Success, 2=Failure, 3=Cancelled, 4=Skipped,
-        5=Waiting,  6=Running, 7=Blocked
-    Only `status == 2` is a real defect; status=3 is cancel-cascade and
-    status=1 is an emission artifact (Gitea wrote a 'failure' commit_status
-    row for a run that actually succeeded — observed empirically on
-    `publish-canvas-image` jobs at SHAs in mc#1597..1630).
-
-    CURRENT STATE (2026-05-20, verified): Gitea 1.22.6 exposes NO REST
-    endpoint for `action_run.status`. Probed:
-        /api/v1/repos/{o}/{r}/actions/runs/{id}   → HTTP 404
-        /api/v1/repos/{o}/{r}/actions/jobs/{id}   → HTTP 404
-        /api/v1/repos/{o}/{r}/actions/tasks/{id}  → HTTP 404
-        /swagger.v1.json paths containing 'actions' → secrets+variables+runners only
-    The SPA backend (`/{repo}/actions/runs/{id}/jobs/{idx}` POST) requires
-    a session CSRF token, unreachable from a runner. The only authoritative
-    source today is direct DB access (`mol_action_status` on op-host,
-    `docker exec molecule-postgres-1 psql ...`), which the runner cannot
-    reach.
-
-    Therefore: this hook returns None on every call. Callers MUST fall
-    back to the description-string filter (existing) plus the HEAD
-    recheck (this PR). When a future Gitea release (>=1.23 expected) or
-    an op-host proxy exposes the endpoint, replace the body of this
-    function with an `api(...)` call — the caller contract is stable.
-
-    See also:
-        - `reference_chronic_red_sweep_cancelled_vs_failed_filter`
-        - `feedback_gitea_status_enum_use_helper_not_raw_int`
-    """
-    _ = target_url  # noqa: F841 — intentional placeholder
-    return None
-
-
 # --------------------------------------------------------------------------
 # Gitea reads
 # --------------------------------------------------------------------------
@@ -274,11 +208,6 @@ def get_combined_status(sha: str) -> dict:
    return body


-def _entry_state(s: dict) -> str:
-    """Per-entry status key in Gitea 1.22.6 is `status`; fall back to `state`."""
-    return s.get("status") or s.get("state") or ""
-
-
 def is_red(status: dict) -> tuple[bool, list[dict]]:
    """Return (is_red, failed_statuses).

@@ -289,31 +218,6 @@ def is_red(status: dict) -> tuple[bool, list[dict]]:

    `failed_statuses` is the list of per-context entries whose own
    `state` is in the red set; useful for the issue body.
-
-    Cancel-cascade filter (mc#1564, 2026-05-19):
-      Gitea maps BOTH `action_run.status=2 (Failure)` AND
-      `action_run.status=3 (Cancelled)` to commit-status string
-      `"failure"`. On a busy main with
-      `concurrency: cancel-in-progress: true`, every merge burst
-      cancels prior in-flight runs (status=3) — those bubble to the
-      combined-status `failure` and inflate the watchdog's red%,
-      generating phantom `[main-red]` issues (mc#1562/#1552/#1540/...).
-      Canonical Gitea 1.22.6 enum per `models/actions/status.go` +
-      `reference_gitea_action_status_enum_corrected_2026_05_19`:
-          1=Success, 2=Failure, 3=Cancelled, 4=Skipped,
-          5=Waiting, 6=Running, 7=Blocked
-      We only want status=2 (real defects) to file. At the
-      commit-status layer we don't have the integer enum directly
-      (only the `failure` rollup string), so we use the description
-      string Gitea writes when a run is cancelled — empirically
-      `"Has been cancelled"` (verified 2026-05-19 via #1562 body).
-      Real failures show `"Failing after Ns"` and are unaffected.
-      This is option B from mc#1564 (description-string filter, no
-      extra API call). Description-string stability is a soft contract
-      with Gitea; if a future release renames it, the cancel-cascade
-      entries will simply leak back through (visible-not-silent), and
-      we'll either re-pin the string or upgrade to option A (resolve
-      the underlying action_run.status integer via target_url).
    """
    combined = status.get("state")
    statuses = status.get("statuses") or []
@@ -326,30 +230,14 @@ def is_red(status: dict) -> tuple[bool, list[dict]]:
    # "no per-context entries were in a red state" fallback even when
    # the combined-state correctly flagged red. See
    # `feedback_smoke_test_vendor_truth_not_shape_match`.
-    def _is_cancel_cascade(s: dict) -> bool:
-        """status=3 entry per Gitea 1.22.6 description-string contract.
-        Match exactly (after strip) — substring match would catch
-        legitimate test names like "Has been cancelled by the user
-        unexpectedly" in failure logs."""
-        desc = (s.get("description") or "").strip()
-        return desc == "Has been cancelled"
+    def _entry_state(s: dict) -> str:
+        return s.get("status") or s.get("state") or ""

    failed = [
        s for s in statuses
-        if isinstance(s, dict)
-        and _entry_state(s) in red_states
-        and not _is_cancel_cascade(s)
+        if isinstance(s, dict) and _entry_state(s) in red_states
    ]
-    # Combined state alone is no longer sufficient — combined=failure
-    # may be 100% cancel-cascade. Drive `red` off the FILTERED list:
-    # if every red-shaped per-entry was cancel-cascade, `failed` is
-    # empty and we report green. Combined-failure with no per-entry
-    # detail (empty `statuses[]`) still trips red — that's the
-    # "CI emitter set combined-status directly" edge case from
-    # render_body's fallback path; we keep filing on it so the
-    # operator sees the breadcrumb.
-    combined_red_no_detail = combined in red_states and not statuses
-    return (bool(failed) or combined_red_no_detail, failed)
+    return (combined in red_states or bool(failed), failed)


 # --------------------------------------------------------------------------
@@ -364,15 +252,6 @@ def title_for(sha: str) -> str:
    return f"{TITLE_PREFIX} {REPO}: {sha[:10]}"


-def _is_scheduled_context(context: str) -> bool:
-    """Return True if `context` is a known scheduled/non-required job.
-
-    These contexts run on a schedule and should not block stale-issue
-    closeout when main's required CI has recovered (mc#1789).
-    """
-    return any(pattern.lower() in context.lower() for pattern in SCHEDULED_CONTEXT_PATTERNS)
-
-
 def list_open_red_issues() -> list[dict]:
    """All open issues whose title starts with `[main-red] {repo}: `.

@@ -382,34 +261,23 @@ def list_open_red_issues() -> list[dict]:
    file-or-update path to POST a duplicate — exactly the regression
    class the helper-raises contract closes.

-    Pagination is exhausted (mc#1789). The old "by design ≤ 1" invariant
-    was false — backlog can exceed 50 open issues.
+    Gitea issue search returns at most 50/page; we only need open
+    `[main-red]` issues which are by design ≤ 1 at any time per repo,
+    so a single page is enough.
    """
-    prefix = f"{TITLE_PREFIX} {REPO}: "
-    all_issues: list[dict] = []
-    page = 1
-    limit = 50
-    while True:
-        _, results = api(
-            "GET",
-            f"/repos/{OWNER}/{NAME}/issues",
-            query={"state": "open", "type": "issues", "limit": str(limit), "page": str(page)},
+    _, results = api(
+        "GET",
+        f"/repos/{OWNER}/{NAME}/issues",
+        query={"state": "open", "type": "issues", "limit": "50"},
+    )
+    if not isinstance(results, list):
+        raise ApiError(
+            f"issue search returned non-list body (got {type(results).__name__})"
        )
-        if not isinstance(results, list):
-            raise ApiError(
-                f"issue search returned non-list body (got {type(results).__name__})"
-            )
-        matched = [
-            i for i in results
-            if isinstance(i, dict)
+    prefix = f"{TITLE_PREFIX} {REPO}: "
+    return [i for i in results if isinstance(i, dict)
            and isinstance(i.get("title"), str)
-            and i["title"].startswith(prefix)
-        ]
-        all_issues.extend(matched)
-        if len(results) < limit:
-            break
-        page += 1
-    return all_issues
+            and i["title"].startswith(prefix)]


 def find_open_issue_for_sha(sha: str) -> dict | None:
@@ -605,156 +473,10 @@ def file_or_update_red(
        sys.stderr.write(f"::warning::label '{RED_LABEL}' not found on repo\n")


-def close_stale_red_issues(
-    current_sha: str,
-    current_status: dict,
-    *,
-    dry_run: bool = False,
-) -> int:
-    """Close open [main-red] issues whose specific failing contexts have
-    all recovered on `current_sha`, even though `main` is still red for
-    other reasons (mc#1789).
-
-    When main stays red across consecutive SHAs for *different* causes,
-    `close_open_red_issues_for_other_shas` never fires (it only runs when
-    main is green). This function prevents stale issues from accumulating
-    indefinitely by comparing per-context recovery across SHAs.
-
-    An issue is considered stale when every context that was in a failed
-    state on the issue's SHA is now either `success` on the current HEAD
-    or absent (workflow removed / renamed). Issues whose original SHA had
-    a combined-red-with-no-detail (empty statuses list) are skipped — we
-    cannot verify recovery without per-context data.
-
-    Returns the number of issues closed.
-    """
-    open_red = list_open_red_issues()
-    if not open_red:
-        return 0
-
-    current_statuses = current_status.get("statuses") or []
-    closed = 0
-
-    for issue in open_red:
-        title = issue.get("title", "")
-        prefix = f"{TITLE_PREFIX} {REPO}: "
-        if not title.startswith(prefix):
-            continue
-        short_sha = title[len(prefix):]
-        if short_sha == current_sha[:10]:
-            continue
-
-        # Query status for the old SHA. Short SHA should resolve; if it
-        # doesn't (GC'd, force-pushed, ambiguous), skip conservatively.
-        try:
-            old_status = get_combined_status(short_sha)
-        except ApiError:
-            continue
-
-        old_red, old_failed = is_red(old_status)
-        if not old_red:
-            # Open issue for a now-green SHA — close it via the normal path.
-            num = issue.get("number")
-            if isinstance(num, int):
-                comment = (
-                    f"Commit `{short_sha}` is no longer red. Closing as the "
-                    f"failure context has recovered or expired."
-                )
-                if dry_run:
-                    print(
-                        f"::notice::[dry-run] would close issue #{num} "
-                        f"({title}) — old SHA is now green"
-                    )
-                    closed += 1
-                    continue
-                api(
-                    "POST",
-                    f"/repos/{OWNER}/{NAME}/issues/{num}/comments",
-                    body={"body": comment},
-                )
-                api(
-                    "PATCH",
-                    f"/repos/{OWNER}/{NAME}/issues/{num}",
-                    body={"state": "closed"},
-                )
-                print(
-                    f"::notice::Closed stale main-red issue #{num} "
-                    f"(old SHA {short_sha} is now green)"
-                )
-                closed += 1
-            continue
-
-        if not old_failed:
-            # Combined red with no per-context detail — can't verify recovery.
-            continue
-
-        # Verify every failed context from the old SHA has recovered.
-        all_recovered = True
-        recovered_ctxs: list[str] = []
-        still_failing_ctxs: list[str] = []
-        for s in old_failed:
-            ctx = s.get("context", "")
-            if not ctx:
-                continue
-            current_match = None
-            for cs in current_statuses:
-                if isinstance(cs, dict) and cs.get("context") == ctx:
-                    current_match = cs
-                    break
-            if current_match is None:
-                recovered_ctxs.append(ctx)
-            elif _entry_state(current_match) == "success":
-                recovered_ctxs.append(ctx)
-            else:
-                all_recovered = False
-                still_failing_ctxs.append(ctx)
-
-        if not all_recovered:
-            continue
-
-        num = issue.get("number")
-        if not isinstance(num, int):
-            continue
-
-        comment = (
-            f"The failing contexts from this SHA (`{short_sha}`) have "
-            f"recovered on current HEAD `{current_sha[:10]}`: "
-            f"{', '.join(recovered_ctxs)}. "
-            f"Main is still red for other reasons; see the current "
-            f"`[main-red]` issue for `{current_sha[:10]}`."
-        )
-        if dry_run:
-            print(
-                f"::notice::[dry-run] would close stale issue #{num} "
-                f"({title}) — contexts recovered"
-            )
-            closed += 1
-            continue
-
-        api(
-            "POST",
-            f"/repos/{OWNER}/{NAME}/issues/{num}/comments",
-            body={"body": comment},
-        )
-        api(
-            "PATCH",
-            f"/repos/{OWNER}/{NAME}/issues/{num}",
-            body={"state": "closed"},
-        )
-        print(
-            f"::notice::Closed stale main-red issue #{num} "
-            f"(contexts recovered at {current_sha[:10]})"
-        )
-        closed += 1
-
-    return closed
-
-
 def close_open_red_issues_for_other_shas(
    current_sha: str,
    *,
    dry_run: bool = False,
-    close_same_sha: bool = False,
 ) -> int:
    """When main is green at current_sha, close any open `[main-red]`
    issues whose title references a different SHA. Returns the number
@@ -763,25 +485,15 @@ def close_open_red_issues_for_other_shas(
    Lineage note: we only close issues whose title prefix matches; if
    a human renamed the issue or added a suffix this won't touch it.
    That's intentional — manual editorial state takes precedence.
-
-    Args:
-        close_same_sha: set True when the caller already knows main is
-            green at current_sha (e.g. recovery block) and wants to close
-            the open issue for THIS SHA too. Defaults False so the
-            green-path callers never accidentally close an issue they just
-            filed on the same tick.
    """
    target_title = title_for(current_sha)
    open_red = list_open_red_issues()
    closed = 0
    for issue in open_red:
        if issue.get("title") == target_title:
-            if not close_same_sha:
-                # Same SHA — caller should not have invoked this if main is
-                # green. Skip defensively (guards against green-path callers
-                # that accidentally pass the SHA they just filed for).
-                continue
-            # close_same_sha=True: close even this SHA's issue (recovery path)
+            # Same SHA — caller should not have invoked this if main is
+            # green. Skip defensively.
+            continue
        num = issue.get("number")
        if not isinstance(num, int):
            continue
@@ -858,130 +570,29 @@ def run_once(*, dry_run: bool = False) -> int:
    }

    if red:
-        # HEAD recheck (task #394 — guards mc#1597..1630 false-positive
-        # cluster). After the initial detection, wait RECHECK_DELAY_SECS
-        # (default 90s; tests stub time.sleep) and re-evaluate:
-        #
-        #   1. Re-fetch HEAD SHA. If HEAD moved, a new commit landed
-        #      mid-tick — the prior red SHA is no longer authoritative
-        #      and the next cron run will re-evaluate against the new
-        #      HEAD. Skip-file.
-        #
-        #   2. If HEAD unchanged, re-fetch the combined status. If it
-        #      recovered (combined state no longer in {failure,error}
-        #      after the cancel-cascade filter), a transient retry
-        #      rolled the run forward. Skip-file.
-        #
-        # Both paths emit a Loki event distinguishable from the real
-        # `main_red_detected` so obs queries can track filter activity.
-        # The settling window is well below the hourly cron cadence —
-        # genuine failures persist past it and are surfaced normally.
-        time.sleep(RECHECK_DELAY_SECS)
-
-        recheck_sha = get_head_sha(WATCH_BRANCH)
-        if recheck_sha != sha:
-            emit_loki_event("main_red_skipped_head_drift", sha, [])
-            print(
-                f"::notice::skip-file (HEAD moved): initial red at "
-                f"{sha[:10]} but HEAD is now {recheck_sha[:10]} on "
-                f"{WATCH_BRANCH}; next cron tick will re-evaluate."
-            )
-            # HEAD drifted — close any stale main-red issue for the prior SHA
-            # before returning, so we don't leave stale open issues when main
-            # is no longer pointing at the red commit.
-            close_open_red_issues_for_other_shas(recheck_sha, dry_run=dry_run)
-            return 0
-
-        recheck_status = get_combined_status(sha)
-        recheck_red, recheck_failed = is_red(recheck_status)
-        if not recheck_red:
-            emit_loki_event("main_red_skipped_recovered", sha, [])
-            print(
-                f"::notice::skip-file (recovered after settling): "
-                f"combined state at {sha[:10]} flipped to "
-                f"{recheck_status.get('state')!r} on recheck; "
-                f"initial red was a transient cancel-cascade."
-            )
-            # CI recovered on the same SHA — close any stale main-red issue
-            # that was filed on a prior tick for this SHA.
-            close_open_red_issues_for_other_shas(sha, dry_run=dry_run, close_same_sha=True)
-            return 0
-
-        # Still red after settling — file/update. Use the recheck data
-        # as authoritative so the issue body reflects the latest state.
-        failed = recheck_failed
-        debug["recheck_combined_state"] = recheck_status.get("state")
-        debug["recheck_failed_contexts"] = [
-            s.get("context") for s in failed
-        ]
-
        failed_ctxs = [s.get("context") for s in failed if s.get("context")]
        emit_loki_event("main_red_detected", sha, failed_ctxs)
        print(f"::warning::main is RED at {sha[:10]} on {WATCH_BRANCH}: "
              f"{len(failed)} failed context(s)")
        file_or_update_red(sha, failed, debug, dry_run=dry_run)
-        stale_closed = close_stale_red_issues(sha, recheck_status, dry_run=dry_run)
-        if stale_closed:
-            emit_loki_event("main_red_stale_closed", sha, [])
-            print(
-                f"::notice::Closed {stale_closed} stale main-red issue(s) "
-                f"whose contexts recovered at {sha[:10]}"
-            )
    else:
-        # Green or pending-with-no-real-failures. Close stale issues
-        # from earlier SHAs when required CI has recovered.
-        #
-        # mc#1789: main often sits at combined `pending` because
-        # scheduled/non-required contexts (Staging SaaS smoke,
-        # Continuous synthetic E2E, main-red-watchdog itself,
-        # ci-arm64-advisory) are still running. We close stale issues
-        # as long as no *non-scheduled* context has failed and no
-        # *non-scheduled* context is still pending — i.e. required CI
-        # is effectively green.
-        #
-        # The success-only gate is preserved for the canonical green
-        # path; the extended check below only fires when combined is
-        # `pending` but all required work is done.
-        combined_state = status.get("state")
-        if combined_state == "success":
-            should_close = True
-            close_reason = "GREEN"
-        else:
-            statuses = status.get("statuses") or []
-            non_scheduled_pending = [
-                s for s in statuses
-                if isinstance(s, dict)
-                and (_entry_state(s) == "pending")
-                and not _is_scheduled_context(s.get("context", ""))
-            ]
-            non_scheduled_failed = [
-                s for s in statuses
-                if isinstance(s, dict)
-                and (_entry_state(s) in {"failure", "error"})
-                and not _is_scheduled_context(s.get("context", ""))
-            ]
-            # Cancel-cascade already filtered by is_red(); red=False
-            # here means no real failures. We additionally check that
-            # no non-scheduled context is still pending.
-            should_close = not non_scheduled_pending and not non_scheduled_failed
-            close_reason = "pending-but-required-green"
-
-        if should_close:
+        # Green (or pending — pending is treated as not-red so we don't
+        # spam during the post-merge CI window). Close any stale issues
+        # from earlier SHAs only when we're actually green; pending
+        # means CI hasn't finished and the prior issue might still be
+        # accurate.
+        if status.get("state") == "success":
            closed = close_open_red_issues_for_other_shas(sha, dry_run=dry_run)
            if closed:
                emit_loki_event(
                    "main_returned_to_green", sha,
                    [],
                )
-            print(
-                f"::notice::main is {close_reason} at {sha[:10]} on {WATCH_BRANCH} "
-                f"(closed {closed} stale issue(s))"
-            )
+            print(f"::notice::main is GREEN at {sha[:10]} on {WATCH_BRANCH} "
+                  f"(closed {closed} stale issue(s))")
        else:
-            print(
-                f"::notice::main has pending-or-failed required CI at {sha[:10]} "
-                f"on {WATCH_BRANCH} (combined state={combined_state!r}; no action)"
-            )
+            print(f"::notice::main is PENDING at {sha[:10]} on {WATCH_BRANCH} "
+                  f"(combined state={status.get('state')!r}; no action)")
    return 0


@@ -17,14 +17,18 @@ import urllib.error
 import urllib.request
 from urllib.parse import quote

+
 TRUE_VALUES = {"1", "true", "yes", "on", "disabled", "disable"}
 PROD_CP_URL = "https://api.moleculesai.app"
 DEFAULT_REQUIRED_CONTEXTS = [
+    "CI / Platform (Go) (push)",
+    "CI / Canvas (Next.js) (push)",
+    "CI / Shellcheck (E2E scripts) (push)",
+    "CI / Python Lint & Test (push)",
    "CI / all-required (push)",
    "Secret scan / Scan diff for credential-shaped strings (push)",
 ]
 TERMINAL_FAILURE_STATES = {"failure", "error", "cancelled", "canceled", "skipped"}
-REDEPLOY_PATH = "/cp/admin/tenants/redeploy-fleet"


 def truthy_flag(value: str | None) -> bool:
@@ -67,12 +71,6 @@ def build_plan(env: dict[str, str]) -> dict:
        "soak_seconds": _int_env(env, "PROD_AUTO_DEPLOY_SOAK_SECONDS", 60, minimum=0),
        "batch_size": _int_env(env, "PROD_AUTO_DEPLOY_BATCH_SIZE", 3),
        "dry_run": truthy_flag(env.get("PROD_AUTO_DEPLOY_DRY_RUN", "")),
-        # confirm:true ack required by CP /cp/admin/tenants/redeploy-fleet
-        # contract (cp#228 / task #308) for fleet-wide intent. Empty body
-        # / {confirm:false} / {only_slugs:[]} → 400. This caller is the
-        # production auto-deploy step that rolls every live tenant (canary
-        # + fan-out), no slug scoping, so confirm:true is correct.
-        "confirm": True,
    }
    if canary_slug:
        body["canary_slug"] = canary_slug
@@ -130,217 +128,6 @@ def required_contexts(env: dict[str, str]) -> list[str]:
    return [line.strip() for line in raw.replace(",", "\n").splitlines() if line.strip()]


-def chunks(items: list[str], size: int) -> list[list[str]]:
-    return [items[i : i + size] for i in range(0, len(items), size)]
-
-
-class RolloutFailed(RuntimeError):
-    def __init__(self, message: str, response: dict):
-        super().__init__(message)
-        self.response = response
-
-
-def slugs_from_redeploy_response(body: dict) -> list[str]:
-    slugs: list[str] = []
-    for row in body.get("results") or []:
-        slug = str(row.get("slug") or "").strip()
-        if slug:
-            slugs.append(slug)
-    return slugs
-
-
-def scoped_redeploy_body(base: dict, slugs: list[str]) -> dict:
-    body = dict(base)
-    body.pop("canary_slug", None)
-    body["only_slugs"] = slugs
-    body["soak_seconds"] = 0
-    body["batch_size"] = max(1, len(slugs))
-    return body
-
-
-def cp_api_json(method: str, url: str, token: str, body: dict | None = None) -> tuple[int, dict]:
-    data = None
-    headers = {
-        "Authorization": f"Bearer {token}",
-        "Accept": "application/json",
-    }
-    if body is not None:
-        data = json.dumps(body).encode("utf-8")
-        headers["Content-Type"] = "application/json"
-    req = urllib.request.Request(url, data=data, headers=headers, method=method)
-    try:
-        with urllib.request.urlopen(req, timeout=120) as resp:
-            return resp.status, json.loads(resp.read())
-    except urllib.error.HTTPError as exc:
-        raw = exc.read().decode("utf-8", errors="replace")
-        try:
-            parsed = json.loads(raw)
-        except json.JSONDecodeError:
-            parsed = {"error": raw[:500]}
-        return exc.code, parsed
-
-
-def plan_rollout_slugs(cp_url: str, token: str, body: dict, redeploy=None) -> list[str]:
-    if redeploy is None:
-        redeploy = redeploy_scoped
-    dry_run_body = dict(body)
-    dry_run_body["dry_run"] = True
-    status, resp = redeploy(cp_url, token, dry_run_body)
-    if status != 200:
-        raise RuntimeError(f"dry-run redeploy-fleet returned HTTP {status}: {resp.get('error', '')}")
-    if resp.get("ok") is not True:
-        raise RuntimeError(f"dry-run redeploy-fleet reported ok={resp.get('ok')}: {resp.get('error', '')}")
-    slugs = slugs_from_redeploy_response(resp)
-    if not slugs:
-        raise RuntimeError("dry-run redeploy-fleet returned no rollout candidates")
-    return slugs
-
-
-def redeploy_scoped(cp_url: str, token: str, body: dict) -> tuple[int, dict]:
-    return cp_api_json("POST", f"{cp_url}{REDEPLOY_PATH}", token, body)
-
-
-def _raise_for_redeploy_result(status: int, body: dict, slugs: list[str]) -> None:
-    if status != 200 or body.get("ok") is not True:
-        raise RuntimeError(
-            "redeploy scoped call failed for "
-            f"{','.join(slugs)}: HTTP {status}, ok={body.get('ok')}"
-        )
-
-
-def rollout_stragglers(enumerated: list[str], results: list[dict]) -> list[str]:
-    """Return every enumerated tenant NOT proven on the target build.
-
-    A straggler is any tenant the rollout was supposed to cover that the
-    CP could not verify is running the target image tag — whether it
-    errored, was skipped, or SSM-succeeded onto the wrong image
-    (internal#724). CP marks each per-tenant result row with
-    ``verified_on_target`` (the REDEPLOY_RUNNING_IMAGE docker-inspect
-    proof). A tenant enumerated for the rollout but absent from the
-    result set (no batch ever ran it) is also a straggler — that is the
-    exact agents-team silent-skip class.
-
-    Backward-compat: an OLDER CP that doesn't emit ``verified_on_target``
-    yet returns rows without the key. Treat a missing key as verified so
-    this surfacing degrades to the previous (ok-based) behavior against an
-    un-upgraded CP, rather than failing every deploy spuriously. Once the
-    CP fix is deployed the key is always present and real stragglers are
-    caught.
-    """
-
-    verified: set[str] = set()
-    for row in results:
-        if str(row.get("ssm_status") or "") == "DryRun":
-            continue
-        slug = str(row.get("slug") or "").strip()
-        if not slug:
-            continue
-        # Missing key (old CP) => assume verified; present key is authoritative.
-        if "verified_on_target" not in row or row.get("verified_on_target"):
-            verified.add(slug)
-    return sorted(s for s in dict.fromkeys(enumerated) if s not in verified)
-
-
-def assert_full_coverage(enumerated: list[str], aggregate: dict, dry_run: bool) -> None:
-    """Fail the rollout if any enumerated tenant is not on the target build.
-
-    This is the no-silent-skip gate (internal#724). A dry run proves
-    nothing landed, so coverage is not asserted for it.
-    """
-
-    if dry_run:
-        return
-    stragglers = rollout_stragglers(enumerated, aggregate.get("results") or [])
-    if stragglers:
-        msg = (
-            f"incomplete rollout: {len(stragglers)} tenant(s) not verified on target "
-            f"after redeploy-fleet: {', '.join(stragglers)} "
-            f"(enumerated {len(set(enumerated))})"
-        )
-        aggregate["ok"] = False
-        aggregate["error"] = msg
-        aggregate["stragglers"] = stragglers
-        raise RolloutFailed(msg, aggregate)
-
-
-def execute_scoped_rollout(
-    plan: dict,
-    token: str,
-    list_slugs=plan_rollout_slugs,
-    redeploy=redeploy_scoped,
-    sleep=time.sleep,
-) -> dict:
-    cp_url = plan["cp_url"]
-    base_body = plan["body"]
-    all_slugs = list_slugs(cp_url, token, base_body)
-    batch_size = int(base_body.get("batch_size") or 1)
-    canary_slug = str(base_body.get("canary_slug") or "").strip()
-    dry_run = bool(base_body.get("dry_run"))
-    aggregate = {"ok": True, "results": []}
-
-    if canary_slug:
-        if canary_slug not in all_slugs:
-            raise RuntimeError(f"configured canary slug {canary_slug!r} is not a running tenant")
-        body = scoped_redeploy_body(base_body, [canary_slug])
-        print(f"POST {cp_url}{REDEPLOY_PATH} only_slugs={','.join(body['only_slugs'])}")
-        status, resp = redeploy(cp_url, token, body)
-        aggregate["results"].extend(resp.get("results") or [])
-        try:
-            _raise_for_redeploy_result(status, resp, [canary_slug])
-        except RuntimeError as exc:
-            aggregate["ok"] = False
-            aggregate["error"] = str(exc)
-            raise RolloutFailed(str(exc), aggregate) from exc
-        soak_seconds = int(base_body.get("soak_seconds") or 0)
-        if soak_seconds > 0 and not dry_run:
-            print(f"Canary passed; soaking locally for {soak_seconds}s")
-            sleep(soak_seconds)
-
-    remaining = [slug for slug in all_slugs if slug != canary_slug]
-    for group in chunks(remaining, batch_size):
-        body = scoped_redeploy_body(base_body, group)
-        print(f"POST {cp_url}{REDEPLOY_PATH} only_slugs={','.join(group)}")
-        status, resp = redeploy(cp_url, token, body)
-        aggregate["results"].extend(resp.get("results") or [])
-        try:
-            _raise_for_redeploy_result(status, resp, group)
-        except RuntimeError as exc:
-            aggregate["ok"] = False
-            aggregate["error"] = str(exc)
-            raise RolloutFailed(str(exc), aggregate) from exc
-
-    # No-silent-skip coverage gate (internal#724): every enumerated tenant
-    # must be PROVEN on the target build. A per-tenant HTTP-200/ok response
-    # is not proof — a tenant that SSM-succeeded but stayed on the old tag,
-    # or one enumerated but never batched, is a straggler. Surfacing it as
-    # a RolloutFailed makes the deploy step exit non-zero instead of
-    # silently reporting success (the exact agents-team failure mode).
-    assert_full_coverage(all_slugs, aggregate, dry_run)
-
-    return aggregate
-
-
-def rollout_from_plan_file(plan_path: str, response_path: str, env: dict[str, str]) -> None:
-    token = env.get("CP_ADMIN_API_TOKEN", "").strip()
-    if not token:
-        raise ValueError("CP_ADMIN_API_TOKEN is required for production auto-deploy")
-    with open(plan_path, "r", encoding="utf-8") as fh:
-        plan = json.load(fh)
-    if not plan.get("enabled"):
-        raise RuntimeError("production auto-deploy plan is disabled")
-    try:
-        response = execute_scoped_rollout(plan, token)
-    except RolloutFailed as exc:
-        response = exc.response
-        with open(response_path, "w", encoding="utf-8") as fh:
-            json.dump(response, fh, sort_keys=True)
-            fh.write("\n")
-        raise
-    with open(response_path, "w", encoding="utf-8") as fh:
-        json.dump(response, fh, sort_keys=True)
-        fh.write("\n")
-
-
 def _api_json(url: str, token: str) -> dict:
    req = urllib.request.Request(url, headers={"Authorization": f"token {token}"})
    try:
@@ -442,9 +229,6 @@ def main() -> int:
    sub.add_parser("plan", help="print production deploy plan as JSON")
    sub.add_parser("assert-enabled", help="fail if production deploy is currently disabled")
    sub.add_parser("wait-ci", help="block until required CI context is green")
-    rollout_parser = sub.add_parser("rollout", help="execute canary-first scoped production rollout")
-    rollout_parser.add_argument("--plan", required=True, help="path to prod-auto-deploy plan JSON")
-    rollout_parser.add_argument("--response", required=True, help="path to write aggregate response JSON")
    args = parser.parse_args()

    try:
@@ -457,9 +241,6 @@ def main() -> int:
        if args.command == "wait-ci":
            wait_for_ci_context(dict(os.environ))
            return 0
-        if args.command == "rollout":
-            rollout_from_plan_file(args.plan, args.response, dict(os.environ))
-            return 0
    except Exception as exc:  # noqa: BLE001 - CLI should render operator-friendly errors.
        print(f"::error::{exc}", file=sys.stderr)
        return 1
@@ -1,5 +1,4 @@
 #!/usr/bin/env bash
-# shellcheck disable=SC2016,SC2329
 # review-check — evaluate whether a PR satisfies a single team-review gate.
 #
 # RFC#324 Step 1 of 5 — qa-review + security-review check workflows.
@@ -12,7 +11,6 @@
 #   ≥ 1 review on the PR where:
 #     • state == APPROVED
 #     • review.dismissed == false
-#     • review.official != false (excludes draft/mis-filed APPROVED reviews)
 #     • review.user.login != PR.user.login (non-author)
 #     • review.user.login ∈ team-members
 #
@@ -102,12 +100,11 @@ printf 'header = "Authorization: token %s"\n' "$GITEA_TOKEN" > "$CURL_AUTH_FILE"
 # (bash trap 'function' EXIT expands variables at trap-fire time, not def time).
 PR_JSON=$(mktemp)
 REVIEWS_JSON=$(mktemp)
-COMMENTS_JSON=$(mktemp)
 TEAM_PROBE_TMP=$(mktemp)
 NA_STATUSES_TMP=""  # declared here so cleanup() always has the var

 cleanup() {
-  rm -f "$CURL_AUTH_FILE" "$PR_JSON" "$REVIEWS_JSON" "$COMMENTS_JSON" "$TEAM_PROBE_TMP" "${NA_STATUSES_TMP-}"
+  rm -f "$CURL_AUTH_FILE" "$PR_JSON" "$REVIEWS_JSON" "$TEAM_PROBE_TMP" "${NA_STATUSES_TMP-}"
 }
 trap cleanup EXIT

@@ -130,7 +127,6 @@ fi
 PR_AUTHOR=$(jq -r '.user.login // ""' "$PR_JSON")
 PR_HEAD_SHA=$(jq -r '.head.sha // ""' "$PR_JSON")
 PR_BASE_REF=$(jq -r '.base.ref // ""' "$PR_JSON")
-PR_BASE_SHA=$(jq -r '.base.sha // ""' "$PR_JSON")
 PR_STATE=$(jq -r '.state // ""' "$PR_JSON")
 DEFAULT_BRANCH="${DEFAULT_BRANCH:-main}"
 debug "pr_author=${PR_AUTHOR} pr_head=${PR_HEAD_SHA:0:7} pr_base=${PR_BASE_REF} pr_state=${PR_STATE}"
@@ -139,10 +135,6 @@ if [ "$PR_STATE" != "open" ]; then
  echo "::notice::PR ${PR_NUMBER} is ${PR_STATE} — exiting 0 (closed PRs do not gate)"
  exit 0
 fi
-if [ "$PR_HEAD_SHA" = "$PR_BASE_SHA" ]; then
-  echo "::notice::PR ${PR_NUMBER} has no diff (head == base) — exiting 0 (empty PRs do not gate)"
-  exit 0
-fi
 if [ "$PR_BASE_REF" != "$DEFAULT_BRANCH" ]; then
  echo "::notice::PR ${PR_NUMBER} targets ${PR_BASE_REF:-<unknown>} not ${DEFAULT_BRANCH} — ${TEAM}-review gate not applicable"
  exit 0
@@ -202,7 +194,6 @@ fi
 JQ_FILTER='.[]
  | select(.state == "APPROVED")
  | select(.dismissed != true)
-  | select(.official != false)
  | select(.user.login != $author)'
 if [ "${REVIEW_CHECK_STRICT:-}" = "1" ]; then
  JQ_FILTER="${JQ_FILTER}
@@ -211,82 +202,11 @@ fi
 JQ_FILTER="${JQ_FILTER}
  | .user.login"

-REVIEW_CANDIDATES=$(jq -r --arg author "$PR_AUTHOR" --arg head "$PR_HEAD_SHA" "$JQ_FILTER" "$REVIEWS_JSON" | sort -u)
-debug "candidate non-author approvers: $(echo "$REVIEW_CANDIDATES" | tr '\n' ' ')"
+CANDIDATES=$(jq -r --arg author "$PR_AUTHOR" --arg head "$PR_HEAD_SHA" "$JQ_FILTER" "$REVIEWS_JSON" | sort -u)
+debug "candidate non-author approvers: $(echo "$CANDIDATES" | tr '\n' ' ')"

-if [ -z "$REVIEW_CANDIDATES" ]; then
-  # --- Guardrail (internal#503): explain the most common false
-  # "no candidates" red. Gitea's review event enum is EXACTLY
-  # APPROVED/REQUEST_CHANGES/COMMENT/PENDING. A wrong value ("APPROVE",
-  # lowercase, ...) is silently accepted (HTTP 200) and stored as
-  # state=PENDING. A correctly-started draft review has an EMPTY body;
-  # a NON-empty body + state==PENDING by a non-author == an intended
-  # verdict mis-filed by a wrong event string. Surface it actionably.
-  # This does NOT change the gate result (still fail-closed below) — it
-  # only converts a mystery red into a named, self-fixing error.
-  MISFILED_FILTER='.[]
-    | select(.state == "PENDING")
-    | select(.dismissed != true)
-    | select(.user.login != $author)
-    | select(((.body // "") | gsub("^\\s+|\\s+$";"") | length) > 0)
-    | "\(.id)\t\(.user.login)"'
-  MISFILED=$(jq -r --arg author "$PR_AUTHOR" "$MISFILED_FILTER" "$REVIEWS_JSON" 2>/dev/null || true)
-  if [ -n "$MISFILED" ]; then
-    echo "::error::${TEAM}-review: non-author review(s) were SUBMITTED but stored as PENDING — almost certainly the wrong Gitea review event string (internal#503)."
-    echo "::error::Gitea accepts ONLY the exact enum APPROVED / REQUEST_CHANGES / COMMENT. 'APPROVE' or lowercase is silently (HTTP 200) filed as PENDING and is invisible to this gate."
-    printf '%s\n' "$MISFILED" | while IFS="$(printf '\t')" read -r _rid _rl; do
-      [ -n "${_rid:-}" ] && echo "::error::  review id=${_rid} by '${_rl}': RE-SUBMIT via POST ${API}/repos/${OWNER}/${NAME}/pulls/${PR_NUMBER}/reviews with {\"event\":\"APPROVED\"} (correct enum) — do NOT edit the DB."
-    done
-  fi
-
-fi
-
-# --- Fallback/extension (internal#348): check issue comments for agent-approval ---
-# core-qa-agent and core-security-agent can approve via issue comments. Always
-# include comment candidates, even if the reviews API returned approvals for a
-# different team; team membership below is the authoritative filter.
-COMMENT_CANDIDATES=""
-AGENT_PATTERN=""
-case "$TEAM" in
-  qa)       AGENT_PATTERN="\\[core-qa-agent\\]" ;;
-  security) AGENT_PATTERN="\\[core-security-agent\\]" ;;
-esac
-HTTP_CODE=$(curl -sS -o "$COMMENTS_JSON" -w '%{http_code}' \
-  -K "$CURL_AUTH_FILE" "${API}/repos/${OWNER}/${NAME}/issues/${PR_NUMBER}/comments")
-debug "GET /issues/${PR_NUMBER}/comments → HTTP ${HTTP_CODE}"
-if [ "$HTTP_CODE" = "200" ]; then
-  # JQ expression: select non-author comments that match either the
-  # agent-prefix pattern (case-insensitive) OR a generic approval keyword.
-  JQ_APPROVALS='
-    .[] |
-    select(.user.login != $author) |
-    . as $cmt |
-    if ($agent_pattern | length) > 0 and ($cmt.body // "" | test($agent_pattern; "i")) then
-      $cmt.user.login
-    elif ($cmt.body // "" | test("\\b(APPROVED|LGTM|ACCEPTED)\\b"; "i")) then
-      $cmt.user.login
-    else
-      empty
-    end
-  '
-  COMMENT_CANDIDATES=$(jq -r \
-    --arg author "$PR_AUTHOR" \
-    --arg agent_pattern "$AGENT_PATTERN" \
-    "$JQ_APPROVALS" \
-    "$COMMENTS_JSON" 2>/dev/null | sort -u)
-  debug "comment-based approval candidates: $(echo "$COMMENT_CANDIDATES" | tr '\n' ' ')"
-
-  if [ -n "$COMMENT_CANDIDATES" ]; then
-    echo "::notice::${TEAM}-review: found $(echo "$COMMENT_CANDIDATES" | wc -w | xargs) comment-based approval candidate(s) — verifying team membership..."
-  fi
-else
-  debug "could not fetch issue comments (HTTP ${HTTP_CODE})"
-fi
-
-CANDIDATES=$(printf '%s\n%s\n' "$REVIEW_CANDIDATES" "$COMMENT_CANDIDATES" | sed '/^$/d' | sort -u)
-
-if [ -z "${CANDIDATES:-}" ]; then
-  echo "::error::${TEAM}-review awaiting non-author APPROVE from ${TEAM} team (no candidates from reviews API or issue comments)"
+if [ -z "$CANDIDATES" ]; then
+  echo "::error::${TEAM}-review awaiting non-author APPROVE from ${TEAM} team (no candidates yet)"
  exit 1
 fi

@@ -296,15 +216,7 @@ fi
 #   403     → token owner is not in this team (Gitea 1.22.6 'Must be a team
 #             member' constraint — see follow-up issue for token-provisioning)
 #   404     → not a member
-# Track whether every candidate returned 403 (token owner not in team).
-# When this happens the root cause is a token-provisioning issue, not a
-# reviewer-eligibility issue — surface it clearly so ops don't waste time
-# verifying team roster (Bug C / RFC#324 follow-up).
-_ALL_CANDIDATES_403="yes"
-_CANDIDATE_COUNT=0
-
 for U in $CANDIDATES; do
-  _CANDIDATE_COUNT=$((_CANDIDATE_COUNT + 1))
  CODE=$(curl -sS -o "$TEAM_PROBE_TMP" -w '%{http_code}' \
    -K "$CURL_AUTH_FILE" "${API}/teams/${TEAM_ID}/members/${U}")
  debug "probe ${U} in team ${TEAM} (id=${TEAM_ID}) → HTTP ${CODE}"
@@ -314,31 +226,22 @@ for U in $CANDIDATES; do
      exit 0
      ;;
    403)
-      # Token owner is not in the team being probed; Gitea 1.22.6 refuses
-      # to confirm membership in this case. Do NOT hard-fail the gate on a
-      # 403 — doing so would fail the entire gate if ANY candidate triggers
-      # a 403, even when other valid team-members exist. Instead skip this
-      # candidate and continue checking others. If all candidates produce
-      # 403 (token owner can't query any of them) the final exit fires.
-      echo "::warning::team-probe for ${U} in ${TEAM} returned 403 (token owner not in ${TEAM} team — skipping; cannot confirm membership)"
+      # Token owner is not in the team being probed; the API refuses to
+      # confirm membership. This is the RFC#324 follow-up token-scope gap.
+      # Fail closed — never grant approval on a 403; surface clearly.
+      echo "::error::team-probe for ${U} in ${TEAM} returned 403 (token owner not in ${TEAM} team — RFC#324 token-scope follow-up). Cannot confirm membership; failing closed."
      cat "$TEAM_PROBE_TMP" >&2
-      continue
+      exit 1
      ;;
    404)
-      _ALL_CANDIDATES_403="no"
      debug "${U} not a member of ${TEAM}"
      ;;
    *)
-      _ALL_CANDIDATES_403="no"
      echo "::warning::team-probe for ${U} in ${TEAM} returned unexpected HTTP ${CODE}"
      cat "$TEAM_PROBE_TMP" >&2
      ;;
  esac
 done

-if [ "$_ALL_CANDIDATES_403" = "yes" ] && [ "$_CANDIDATE_COUNT" -gt 0 ]; then
-  echo "::error::${TEAM}-review FAILED — every candidate returned 403 (token owner is not a member of the ${TEAM} team). This is a TOKEN PROVISIONING issue, not a reviewer-eligibility issue. Add the token owner to the '${TEAM}' Gitea team (id=${TEAM_ID}) or use a token whose owner is already in that team."
-else
-  echo "::error::${TEAM}-review awaiting non-author APPROVE from ${TEAM} team (candidates: $(echo "$CANDIDATES" | tr '\n' ',' | sed 's/,$//') — none are in team)"
-fi
+echo "::error::${TEAM}-review awaiting non-author APPROVE from ${TEAM} team (candidates: $(echo "$CANDIDATES" | tr '\n' ',' | sed 's/,$//') — none are in team)"
 exit 1
@@ -13,26 +13,20 @@ set -euo pipefail
 OWNER="${REPO%%/*}"
 NAME="${REPO##*/}"
 API="https://${GITEA_HOST}/api/v1"
-# Branch-protection requires the (pull_request_target) context variant.
-# The refire path must post the EXACT BP-required name so the gate flips.
-CONTEXT="${TEAM}-review / approved (pull_request_target)"
+CONTEXT="${TEAM}-review / approved (pull_request)"
 TARGET_URL="https://${GITEA_HOST}/${OWNER}/${NAME}/pulls/${PR_NUMBER}"

 authfile=$(mktemp)
-post_authfile=$(mktemp)
 prfile=$(mktemp)
 postfile=$(mktemp)
 # shellcheck disable=SC2329 # invoked by EXIT trap
 cleanup() {
-  rm -f "$authfile" "$post_authfile" "$prfile" "$postfile"
+  rm -f "$authfile" "$prfile" "$postfile"
 }
 trap cleanup EXIT

-chmod 600 "$authfile" "$post_authfile"
+chmod 600 "$authfile"
 printf 'header = "Authorization: token %s"\n' "$GITEA_TOKEN" > "$authfile"
-# STATUS_POST_TOKEN is narrow-scoped write:repository for explicit status POST.
-# Falls back to GITEA_TOKEN for backward compatibility (e.g. local test).
-printf 'header = "Authorization: token %s"\n' "${STATUS_POST_TOKEN:-$GITEA_TOKEN}" > "$post_authfile"

 code=$(curl -sS -o "$prfile" -w '%{http_code}' -K "$authfile" \
  "${API}/repos/${OWNER}/${NAME}/pulls/${PR_NUMBER}")
@@ -74,7 +68,7 @@ body=$(jq -nc \
  '{state:$state, context:$context, description:$description, target_url:$target_url}')

 code=$(curl -sS -o "$postfile" -w '%{http_code}' -X POST \
-  -K "$post_authfile" -H "Content-Type: application/json" \
+  -K "$authfile" -H "Content-Type: application/json" \
  -d "$body" \
  "${API}/repos/${OWNER}/${NAME}/statuses/${head_sha}")
 if [ "$code" != "200" ] && [ "$code" != "201" ]; then
@@ -6,8 +6,8 @@
 # RFC#351 Step 2 of 6 (implementation MVP).
 #
 # Invoked by .gitea/workflows/sop-checklist.yml on:
-#   - pull_request_target: [opened, edited, synchronize, reopened, labeled, unlabeled]
-#   - issue_comment:       [created]  # edited/deleted omitted (Gitea 1.22.6 job-parsing quirk)
+#   - pull_request_target: [opened, edited, synchronize, reopened]
+#   - issue_comment:       [created, edited, deleted]
 #
 # Flow:
 #   1. Load .gitea/sop-checklist-config.yaml (from BASE ref — trusted).
@@ -64,41 +64,11 @@ import argparse
 import json
 import os
 import re
-import resource
 import sys
 import urllib.error
 import urllib.parse
 import urllib.request
-from typing import Any, Callable, Iterator
-
-# ---------------------------------------------------------------------------
-# Address-space guardrail (RFC#369 / task #369 follow-up to mc#1242-class OOM).
-#
-# `get_issue_comments` paginates the full comment history of a PR. On
-# bot-relay-heavy PRs (e.g. mc#291, mc#1242) this can balloon past the
-# runner's cgroup memory limit and 137 the job. Cap virtual-address-space
-# at 2 GiB so the script OOMs as a `MemoryError` (catchable / surfaceable)
-# rather than a SIGKILL we can't post a status for.
-#
-# 2 GiB is generous — a 5000-comment PR with 1 KiB minimal-dicts (see
-# get_issue_comments below) fits in ~10 MiB, leaving plenty of headroom
-# for the Python runtime + urllib + json buffers.
-#
-# Skipped under pytest / dry-run where RLIMIT_AS would interfere with
-# test runner memory needs (set SOP_CHECKLIST_NO_RLIMIT=1 to opt out).
-if not os.environ.get("SOP_CHECKLIST_NO_RLIMIT"):
-    try:
-        resource.setrlimit(resource.RLIMIT_AS, (2 * 1024**3, 2 * 1024**3))
-    except (ValueError, OSError):
-        # macOS sometimes refuses RLIMIT_AS; not fatal — the Linux runner
-        # is the only place this matters for the OOM-prevention goal.
-        pass
-
-# Per-comment body cap (task #369). The directive parser walks the body
-# line-by-line looking for ^/sop-ack ^/sop-revoke ^/sop-n/a markers — only
-# the first few KiB matter for that. Cap each comment body so a single
-# pasted-log comment can't push us past the cgroup limit.
-_MAX_BODY_BYTES = int(os.environ.get("SOP_CHECKLIST_MAX_BODY_BYTES") or 8 * 1024)
+from typing import Any


 # ---------------------------------------------------------------------------
@@ -140,7 +110,7 @@ def normalize_slug(raw: str, numeric_aliases: dict[int, str] | None = None) -> s
 # for /sop-revoke (RFC#351 open question 4 — reason is captured but not
 # yet validated; future iteration may require a min-length).
 _DIRECTIVE_RE = re.compile(
-    r"^[ \t]*/(sop-ack|sop-revoke|sop-n/a)[ \t]+([A-Za-z0-9_\- ]+?)(?:[ \t]+(.*))?[ \t]*$",
+    r"^[ \t]*/(sop-ack|sop-revoke)[ \t]+([A-Za-z0-9_\- ]+?)(?:[ \t]+(.*))?[ \t]*$",
    re.MULTILINE,
 )

@@ -148,21 +118,19 @@ _DIRECTIVE_RE = re.compile(
 def parse_directives(
    comment_body: str,
    numeric_aliases: dict[int, str],
-) -> tuple[list[tuple[str, str, str]], list[tuple[str, str, str]]]:
-    """Extract /sop-ack, /sop-revoke, and /sop-n/a directives from a comment body.
+) -> tuple[list[tuple[str, str, str]], list]:
+    """Extract /sop-ack and /sop-revoke directives from a comment body.

-    Returns (directives, na_directives) where each is a list of
-    (kind, canonical_slug, note) tuples:
-      kind is "sop-ack", "sop-revoke", or "sop-n/a"
-      canonical_slug is the normalized form (or "" if unparseable)
-      note is the trailing free-text (may be "")
-    The two lists are kept separate so call sites can unpack them
-    directly (e.g. directives, na_directives = parse_directives(...)).
+    Returns (directives, na_directives) where:
+      directives is a list of (kind, canonical_slug, note) tuples
+        kind is "sop-ack" or "sop-revoke"
+        canonical_slug is the normalized form (or "" if unparseable)
+        note is the trailing free-text (may be "")
+      na_directives is reserved for future N/A handling (always [] for now)
    """
-    directives: list[tuple[str, str, str]] = []
-    na_directives: list[tuple[str, str, str]] = []
+    out: list[tuple[str, str, str]] = []
    if not comment_body:
-        return directives, na_directives
+        return out, []
    for m in _DIRECTIVE_RE.finditer(comment_body):
        kind = m.group(1)
        raw_slug = (m.group(2) or "").strip()
@@ -192,12 +160,8 @@ def parse_directives(
        note_from_group = (m.group(3) or "").strip()
        # If we collapsed multi-word slug into kebab and there's a
        # trailing-text group too, append it.
-        entry = (kind, canonical, note_from_group)
-        if kind == "sop-n/a":
-            na_directives.append(entry)
-        else:
-            directives.append(entry)
-    return directives, na_directives
+        out.append((kind, canonical, note_from_group))
+    return out, []


 # ---------------------------------------------------------------------------
@@ -210,8 +174,8 @@ def section_marker_present(body: str, marker: str) -> bool:
    on a non-empty line (i.e. the author actually filled it in).

    We require the marker substring AND non-whitespace content on the
-    same line OR within the next non-blank line — this prevents
-    trivially-empty checklists like:
+    same line OR within the next line — this prevents trivially-empty
+    checklists like:

        ## SOP-Checklist
        - [ ] **Comprehensive testing performed**:
@@ -220,18 +184,9 @@ def section_marker_present(body: str, marker: str) -> bool:
    from auto-passing the section-present check. The peer-ack is still
    required, but answering with empty content is captured as a soft
    finding via the section-present test alone.
-
-    NOTE: we scan forward through blank lines (the markdown-header pattern
-    is ## Header\\n\\ncontent) so that a header + blank-line + content
-    structure still satisfies the check. The backward checkbox fallback
-    catches inline markers without a preceding checkbox (mc#1099).
    """
    if not body or not marker:
        return False
-    # Strip trailing whitespace so the blank-line scan below can find
-    # content that appears on the very last line of the body (without
-    # being misled by a trailing \n or spaces).
-    body = body.rstrip()
    body_lower = body.lower()
    marker_lower = marker.lower()
    idx = body_lower.find(marker_lower)
@@ -247,44 +202,13 @@ def section_marker_present(body: str, marker: str) -> bool:
    stripped = re.sub(r"[\s\*:\-\[\]]+", "", line)
    if stripped:
        return True
-    # Fall through: scan forward, skipping blank-only lines, until we find
-    # non-empty content or run out of body.  Handles:
-    #   ## Header          ← marker line (empty after marker)
-    #                      ← blank line (skipped)
-    #   - actual content   ← found
-    pos = line_end
-    while True:
-        # Skip the current newline and any additional newlines (blank lines).
-        while pos < len(body) and body[pos] == "\n":
-            pos += 1
-        if pos >= len(body):
-            break
-        line_end = body.find("\n", pos)
-        if line_end < 0:
-            line_end = len(body)
-        line = body[pos:line_end]
-        stripped = re.sub(r"[\s\*:\-\[\]]+", "", line)
-        if stripped:
-            return True
-        pos = line_end
-    # Last resort: the marker may appear mid-sentence (e.g.
-    # **Memory/saved-feedback consulted**: No applicable...).
-    # Search backward within the CURRENT LINE only (not preceding lines)
-    # to find a checkbox on the same line before the marker text.
-    # mc#1099 follow-up: memory-consulted detection was failing because
-    # the checkbox was on the same line before the inline marker.
-    _CHECKBOX_RE = re.compile(r"- \[[ x\]]|<input", re.IGNORECASE)
-    line_start = body.rfind("\n", 0, idx) + 1  # 0 if no newline before idx
-    before = body[line_start:idx]
-    m = _CHECKBOX_RE.search(before)
-    if not m:
-        return False
-    # Require meaningful content between the checkbox and the marker text
-    # (markdown formatting like ** or * must also be stripped).
-    # If only whitespace/markdown chars remain, the checkbox line is empty.
-    between = before[m.end() :]
-    stripped_between = re.sub(r"[\s\*:#\[\]_\-]+", "", between)
-    return bool(stripped_between)
+    # Fall through: check the NEXT line (multi-line answers).
+    next_line_end = body.find("\n", line_end + 1)
+    if next_line_end < 0:
+        next_line_end = len(body)
+    next_line = body[line_end + 1:next_line_end]
+    stripped_next = re.sub(r"[\s\*:\-\[\]]+", "", next_line)
+    return bool(stripped_next)


 # ---------------------------------------------------------------------------
@@ -298,7 +222,6 @@ def compute_ack_state(
    items_by_slug: dict[str, dict[str, Any]],
    numeric_aliases: dict[int, str],
    team_membership_probe: "callable[[str, list[str]], list[str]]",
-    high_risk: bool = False,
 ) -> dict[str, dict[str, Any]]:
    """Compute per-item ack state.

@@ -328,7 +251,8 @@ def compute_ack_state(
        user = (c.get("user") or {}).get("login", "")
        if not user:
            continue
-        for kind, slug, _note in parse_directives(body, numeric_aliases)[0]:
+        directives, _na = parse_directives(body, numeric_aliases)
+        for kind, slug, _note in directives:
            if not slug:
                unparseable_per_user[user] = unparseable_per_user.get(user, 0) + 1
                continue
@@ -338,6 +262,7 @@ def compute_ack_state(
    # Filter out self-acks and unknown slugs.
    ackers_per_slug: dict[str, list[str]] = {s: [] for s in items_by_slug}
    rejected_self: dict[str, list[str]] = {s: [] for s in items_by_slug}
+    rejected_unknown: dict[str, list[str]] = {s: [] for s in items_by_slug}
    pending_team_check: dict[str, list[str]] = {s: [] for s in items_by_slug}

    for (user, slug), kind in latest_directive.items():
@@ -360,16 +285,11 @@ def compute_ack_state(
    for slug, candidates in pending_team_check.items():
        if not candidates:
            continue
-        # Risk-class-aware required-teams resolution (RFC#450 Option C):
-        # high-risk PRs use `required_teams_high_risk` (when set on the
-        # item); default class uses `required_teams`. The probe closure
-        # is built with the same high_risk flag so the two reads are
-        # always consistent (both sites share `resolve_required_teams`).
-        required = resolve_required_teams(items_by_slug[slug], high_risk)
+        required = items_by_slug[slug]["required_teams"]
        approved = team_membership_probe(slug, candidates)  # returns subset
        rejected_not_in_team[slug] = [u for u in candidates if u not in approved]
        ackers_per_slug[slug] = approved
-        # Stash resolved teams for description rendering.
+        # Stash required teams for description rendering.
        items_by_slug[slug]["_required_resolved"] = required

    return {
@@ -384,63 +304,6 @@ def compute_ack_state(
    }


-# ---------------------------------------------------------------------------
-# N/A-gate evaluation
-# ---------------------------------------------------------------------------
-
-
-def compute_na_state(
-    comments: list[dict[str, Any]],
-    author: str,
-    na_gates: dict[str, Any],
-    probe: Callable[[str, list[str]], list[str]],
-) -> dict[str, dict[str, Any]]:
-    """Evaluate which N/A gates have a valid declaration from a team member.
-
-    Returns dict[gate_name, dict] where each dict has:
-      declared: bool — at least one valid non-author team-member declared N/A
-      decl_ackers: list[str] — usernames who declared this gate N/A
-      rejected: dict with keys:
-        not_in_team: list[str] — users who tried but aren't in required teams
-    """
-    # Build per-user latest N/A directive (most-recent wins per RFC#324).
-    latest_na: dict[str, tuple[str, str]] = {}  # user → (gate, note)
-    for c in comments:
-        body = c.get("body", "") or ""
-        user = (c.get("user") or {}).get("login", "")
-        if not user:
-            continue
-        for kind, gate, note in parse_directives(body, {})[1]:
-            # [1] = na_directives only
-            if gate in na_gates:
-                latest_na[user] = (gate, note)
-
-    result: dict[str, dict[str, Any]] = {}
-    for gate, gate_cfg in na_gates.items():
-        result[gate] = {
-            "declared": False,
-            "decl_ackers": [],
-            "rejected": {"not_in_team": []},
-        }
-        decl_ackers: list[str] = []
-        not_in_team: list[str] = []
-        for user, (g, _note) in latest_na.items():
-            if g != gate:
-                continue
-            if user == author:
-                continue  # authors cannot self-declare N/A
-            approved = probe(gate, [user])
-            if approved:
-                decl_ackers.append(user)
-            else:
-                not_in_team.append(user)
-        result[gate]["declared"] = bool(decl_ackers)
-        result[gate]["decl_ackers"] = decl_ackers
-        result[gate]["rejected"]["not_in_team"] = not_in_team
-
-    return result
-
-
 # ---------------------------------------------------------------------------
 # Gitea API client
 # ---------------------------------------------------------------------------
@@ -489,35 +352,16 @@ class GiteaClient:
            raise RuntimeError(f"GET pulls/{pr} → HTTP {code}: {data!r}")
        return data

-    def iter_issue_comments(
-        self, owner: str, repo: str, issue: int, page_size: int = 50
-    ) -> Iterator[dict[str, Any]]:
-        """Stream comments page-by-page, yielding ONE minimal-dict per comment.
-
-        Each yielded comment carries ONLY the fields the gate actually reads
-        — `{"user": {"login": str}, "body": str}` — and DROPS the much
-        larger Gitea-API extras (html_url, pull_request_url, issue_url,
-        assets, created_at, updated_at, id, original_author_*).
-
-        Memory motivation (task #369 / mc#1242-class OOM): full Gitea
-        comment dicts are ~2 KiB median + ~3 KiB p95. On PRs with several
-        thousand bot-relay comments the eager `list[full_dict]` shape used
-        previously pushed runner anon-rss past the cgroup limit. The
-        minimal-dict shape is ~10-20x smaller (typically ~50-100B Python
-        overhead + the body string).
-
-        The two downstream consumers (`compute_ack_state`,
-        `compute_na_state`) each iterate the comment list exactly once and
-        read only `body` + `user.login`, so dropping every other field is
-        safe. They still receive `list[dict[str, Any]]`-shaped objects so
-        the test fixtures (which already used the minimal shape) keep
-        working with no fixture changes.
-        """
+    def get_issue_comments(
+        self, owner: str, repo: str, issue: int
+    ) -> list[dict[str, Any]]:
+        # Paginate. Gitea default page size 50.
+        out: list[dict[str, Any]] = []
        page = 1
        while True:
            code, data = self._req(
                "GET",
-                f"/repos/{owner}/{repo}/issues/{issue}/comments?limit={page_size}&page={page}",
+                f"/repos/{owner}/{repo}/issues/{issue}/comments?limit=50&page={page}",
            )
            if code != 200:
                raise RuntimeError(
@@ -525,41 +369,10 @@ class GiteaClient:
                )
            if not data:
                break
-            for c in data:
-                # Minimal projection — drop ALL fields the gate doesn't read.
-                user_login = ((c.get("user") or {}).get("login") or "") if isinstance(c, dict) else ""
-                body = (c.get("body") if isinstance(c, dict) else "") or ""
-                # Body-size guardrail: huge comments (e.g. pasted CI logs) can
-                # individually be MiBs. The directive parser only needs the
-                # first ~8 KiB to find /sop-ack /sop-revoke /sop-n/a markers
-                # — anything past that is filler. Truncate at 8 KiB so a
-                # single oversized comment can't OOM the runner.
-                if len(body) > _MAX_BODY_BYTES:
-                    body = body[:_MAX_BODY_BYTES]
-                yield {"user": {"login": user_login}, "body": body}
-            if len(data) < page_size:
+            out.extend(data)
+            if len(data) < 50:
                break
            page += 1
-
-    def get_issue_comments(
-        self,
-        owner: str,
-        repo: str,
-        issue: int,
-        max_comments: int | None = None,
-    ) -> list[dict[str, Any]]:
-        """Paginate + collect minimal comment dicts. See `iter_issue_comments`
-        for the per-comment shape and the OOM-prevention rationale.
-
-        `max_comments` (optional, default unbounded): hard cap. When the cap
-        is hit we stop fetching further pages and the caller surfaces a
-        soft 'skipping due to volume' status (see main()).
-        """
-        out: list[dict[str, Any]] = []
-        for c in self.iter_issue_comments(owner, repo, issue):
-            out.append(c)
-            if max_comments is not None and len(out) >= max_comments:
-                break
        return out

    def resolve_team_id(self, org: str, team_name: str) -> int | None:
@@ -636,11 +449,8 @@ def load_config(path: str) -> dict[str, Any]:
    dep by keeping the config shape constrained.
    """
    try:
-        # yaml is an optional dep; the canonical loader is used when available,
-        # but the SOP runs on runners that may not have PyYAML installed. The
-        # fallback _load_config_minimal covers the same config shape without
-        import yaml  # type: ignore[import-not-found]  # optional dep; fall back silently if absent
-        with open(path, encoding="utf-8") as f:
+        import yaml  # type: ignore[import-not-found]
+        with open(path) as f:
            return yaml.safe_load(f)
    except ImportError:
        return _load_config_minimal(path)
@@ -654,19 +464,13 @@ def _load_config_minimal(path: str) -> dict[str, Any]:
    item map: scalars + lists of scalars. Does NOT support nested lists,
    YAML anchors, multi-doc, or flow style.
    """
-    with open(path, encoding="utf-8") as f:
+    with open(path) as f:
        lines = f.readlines()
    return _parse_minimal_yaml(lines)


-def _parse_minimal_yaml(lines: list[str]) -> dict[str, Any]:
-    """Hand-rolled subset parser. See _load_config_minimal docstring.
-
-    C901: function is necessarily long — it implements a finite-state YAML
-    subset (scalars, maps, lists of maps at fixed depth). No utility refactors
-    meaningfully reduce length without degrading readability. All branches
-    are exhaustively tested in test_parse_minimal_yaml.py.
-    """
+def _parse_minimal_yaml(lines: list[str]) -> dict[str, Any]:  # noqa: C901
+    """Hand-rolled subset parser. See _load_config_minimal docstring."""
    # Strip comments + blank lines but preserve indentation.
    cleaned: list[tuple[int, str]] = []
    for raw in lines:
@@ -850,7 +654,7 @@ def render_status(
 def get_tier_mode(pr: dict[str, Any], cfg: dict[str, Any]) -> str:
    """Read tier label, return 'hard' or 'soft' per cfg.tier_failure_mode."""
    labels = pr.get("labels") or []
-    tier_labels = [label.get("name", "") for label in labels if (label.get("name", "") or "").startswith("tier:")]
+    tier_labels = [l.get("name", "") for l in labels if (l.get("name", "") or "").startswith("tier:")]
    mode_map = cfg.get("tier_failure_mode") or {}
    default_mode = cfg.get("default_mode", "hard")
    for tl in tier_labels:
@@ -859,83 +663,6 @@ def get_tier_mode(pr: dict[str, Any], cfg: dict[str, Any]) -> str:
    return default_mode


-def is_high_risk(pr: dict[str, Any], cfg: dict[str, Any]) -> bool:
-    """Return True when the PR is high-risk per RFC#450 Option C.
-
-    A PR is high-risk when ANY of:
-      - it carries the `tier:high` label (mechanically strictest tier), or
-      - it carries any label listed in cfg.high_risk_labels.
-
-    High-risk PRs use `required_teams_high_risk` (when set on an item)
-    instead of the default `required_teams`. Items without
-    `required_teams_high_risk` are unaffected (the default applies).
-
-    Governance fix for internal#442 — closes the inconsistency between
-    sop-tier-check (tier-aware) and sop-checklist (was tier-blind).
-    """
-    label_set = {(label.get("name") or "") for label in (pr.get("labels") or [])}
-    if "tier:high" in label_set:
-        return True
-    high_risk_labels = set(cfg.get("high_risk_labels") or [])
-    return bool(label_set & high_risk_labels)
-
-
-def resolve_required_teams(item: dict[str, Any], high_risk: bool) -> list[str]:
-    """Pick the active required_teams list for an item.
-
-    When high_risk is True AND the item declares a non-empty
-    `required_teams_high_risk`, return that. Else fall back to
-    `required_teams`. Keeping this in one helper means the gate's
-    decision shape stays single-sited even as items grow.
-    """
-    if high_risk:
-        elevated = item.get("required_teams_high_risk") or []
-        if elevated:
-            return list(elevated)
-    return list(item.get("required_teams") or [])
-
-
-# ---------------------------------------------------------------------------
-# CI status validation for testing-class AI acks (internal#760 CTO hardening)
-# ---------------------------------------------------------------------------
-
-# Slugs that require CI / all-required green before an AI ack is valid.
-_TESTING_CLASS_SLUGS = {"comprehensive-testing", "local-postgres-e2e", "staging-smoke"}
-
-# Human-only carve-out: these items can NEVER be acked by AI, regardless
-# of config drift. Any item in this set MUST NOT have ai_ack_eligible.
-# migration / schema are future-proofing — not yet in config items, but
-# the code guard rejects them proactively (CTO hardening, msg 1388c76f).
-_HUMAN_ONLY_SLUGS = {"root-cause", "no-backwards-compat", "migration", "schema"}
-
-
-def get_ci_status(client: GiteaClient, owner: str, repo: str, sha: str) -> str:
-    """Return the state of CI / all-required (pull_request) for `sha`.
-
-    Looks through the commit statuses and returns the state string
-    ("success", "failure", "pending", "error") or "missing" if the
-    context is not found. This prevents an AI agent from attesting
-    "tests pass" independently of the actual CI run.
-    """
-    code, data = client._req(  # noqa: SLF001
-        "GET", f"/repos/{owner}/{repo}/statuses/{sha}"
-    )
-    if code != 200:
-        return "unknown"
-    if not data or not isinstance(data, list):
-        return "missing"
-    # Gitea returns statuses newest-first. Find the latest for our context.
-    for status in data:
-        if status.get("context") == "CI / all-required (pull_request)":
-            return status.get("state", "unknown")
-    return "missing"
-
-
-# ---------------------------------------------------------------------------
-# Main entry point
-# ---------------------------------------------------------------------------
-
-
 def main(argv: list[str] | None = None) -> int:
    p = argparse.ArgumentParser()
    p.add_argument("--owner", required=True)
@@ -961,17 +688,6 @@ def main(argv: list[str] | None = None) -> int:
            "thing BP sees is the POSTed status. Useful for local debugging."
        ),
    )
-    p.add_argument(
-        "--max-comments",
-        type=int,
-        default=int(os.environ.get("SOP_CHECKLIST_MAX_COMMENTS") or 5000),
-        help=(
-            "Hard cap on comments fetched from the PR. Above this we post "
-            "a SOFT-pending status with a 'skipping due to volume' note "
-            "instead of OOM'ing the runner (task #369). Override with the "
-            "SOP_CHECKLIST_MAX_COMMENTS env var. Set 0 to disable the cap."
-        ),
-    )
    args = p.parse_args(argv)

    token = os.environ.get("GITEA_TOKEN", "")
@@ -982,7 +698,6 @@ def main(argv: list[str] | None = None) -> int:
    cfg = load_config(args.config)
    items: list[dict[str, Any]] = cfg["items"]
    items_by_slug = {it["slug"]: it for it in items}
-    na_gates: dict[str, Any] = cfg.get("n/a_gates", {})
    numeric_aliases = {
        int(it["numeric_alias"]): it["slug"] for it in items if it.get("numeric_alias")
    }
@@ -1005,62 +720,16 @@ def main(argv: list[str] | None = None) -> int:
        print("::error::PR payload missing user.login or head.sha", file=sys.stderr)
        return 1

-    max_comments_cap = args.max_comments if args.max_comments and args.max_comments > 0 else None
-    comments = client.get_issue_comments(
-        args.owner, args.repo, args.pr, max_comments=max_comments_cap
-    )
-
-    # Volume short-circuit: PRs with thousands of bot-relay comments
-    # (the mc#1242-class OOM source) get a soft 'volume-skipped' status
-    # so the gate doesn't churn the runner; reviewers can re-trigger by
-    # editing the PR or filing a fresh PR with the housekeeping comments
-    # split off. Cap-hit means we couldn't see the WHOLE history, so we
-    # can't fairly post failure — pending is the safe default.
-    volume_skipped = bool(max_comments_cap and len(comments) >= max_comments_cap)
-
-    # High-risk classification (RFC#450 Option C, governance fix for
-    # internal#442). Computed ONCE per PR — used by both the probe
-    # closure and compute_ack_state so the elevation decision is
-    # single-sited.
-    high_risk = is_high_risk(pr, cfg)
+    comments = client.get_issue_comments(args.owner, args.repo, args.pr)

    # Build team-membership probe closure that caches results per
    # (user, team-id) so a user acking multiple items only triggers
    # one membership lookup per team.
    team_member_cache: dict[tuple[str, int], bool | None] = {}

-    # Pre-resolve the ai-sop-ack team id once (None if the team does not exist).
-    ai_sop_ack_team_id = client.resolve_team_id(args.owner, "ai-sop-ack")
-
    def probe(slug: str, users: list[str]) -> list[str]:
-        # `slug` may be either an items-key (compute_ack_state caller) OR
-        # an n/a-gate key (compute_na_state caller). Previously this hard
-        # KeyError'd on the n/a-gate path when slug was e.g. "security-review"
-        # — that's a config gate, not an item — so the gate would crash
-        # instead of falling back to the gate's own required_teams. Fix
-        # task #369 follow-up to issue #355.
-        if slug in items_by_slug:
-            item = items_by_slug[slug]
-            team_names: list[str] = resolve_required_teams(item, high_risk)
-        elif slug in na_gates:
-            # n/a-gate configs carry `required_teams` directly (see
-            # sop-checklist-config.yaml: n/a_gates.<gate>.required_teams).
-            gate_cfg = na_gates[slug] or {}
-            team_names = list(gate_cfg.get("required_teams") or [])
-            if not team_names:
-                print(
-                    f"::warning::n/a-gate '{slug}' has no required_teams; "
-                    "fail-closed (no users will be approved)",
-                    file=sys.stderr,
-                )
-        else:
-            # Unknown slug — fail closed, log so we can find config drift.
-            print(
-                f"::warning::probe() called with slug '{slug}' which is "
-                f"neither an items entry nor an n/a-gate; fail-closed",
-                file=sys.stderr,
-            )
-            return []
+        item = items_by_slug[slug]
+        team_names: list[str] = item["required_teams"]
        # Resolve names → ids. NOTE: orgs/{org}/teams/search may not be
        # available — fall back to the list endpoint.
        team_ids: list[int] = []
@@ -1068,14 +737,14 @@ def main(argv: list[str] | None = None) -> int:
            tid = client.resolve_team_id(args.owner, tn)
            if tid is None:
                # Try the list endpoint as a fallback.
-                code, data = client._req(  # noqa: SLF001  # internal helper; called from loop in caller context
+                code, data = client._req(  # noqa: SLF001
                    "GET", f"/orgs/{args.owner}/teams"
                )
                if code == 200 and isinstance(data, list):
                    for t in data:
                        if t.get("name") == tn:
                            tid = t.get("id")
-                            client._team_id_cache[(args.owner, tn)] = tid  # noqa: SLF001  # write-through cache; intentional side-effect for reuse across calls
+                            client._team_id_cache[(args.owner, tn)] = tid  # noqa: SLF001
                            break
            if tid is not None:
                team_ids.append(tid)
@@ -1086,18 +755,14 @@ def main(argv: list[str] | None = None) -> int:
                    file=sys.stderr,
                )
        approved: list[str] = []
-        rejected_ai_ineligible: list[str] = []
-        rejected_ci_not_green: list[str] = []
        for u in users:
-            # 1) Human required_teams membership check
-            in_human_team = False
            for tid in team_ids:
                cache_key = (u, tid)
                if cache_key not in team_member_cache:
                    team_member_cache[cache_key] = client.is_team_member(tid, u)
                result = team_member_cache[cache_key]
                if result is True:
-                    in_human_team = True
+                    approved.append(u)
                    break
                if result is None:
                    print(
@@ -1107,49 +772,9 @@ def main(argv: list[str] | None = None) -> int:
                    )
                    # Treat as not-in-team for this user/team pair; loop
                    # may still find membership in another team.
-            if in_human_team:
-                approved.append(u)
-                continue
-
-            # 2) AI-sop-ack team membership check (only for items that allow it).
-            if slug in items_by_slug:
-                item = items_by_slug[slug]
-                # Defensive: human-only carve-out is enforced in code, not just
-                # config. Even if ai_ack_eligible were mistakenly added to a
-                # migration/schema item, the AI path is rejected here.
-                if slug in _HUMAN_ONLY_SLUGS:
-                    rejected_ai_ineligible.append(u)
-                    continue
-                if item.get("ai_ack_eligible") and ai_sop_ack_team_id is not None:
-                    cache_key = (u, ai_sop_ack_team_id)
-                    if cache_key not in team_member_cache:
-                        team_member_cache[cache_key] = client.is_team_member(
-                            ai_sop_ack_team_id, u
-                        )
-                    result = team_member_cache[cache_key]
-                    if result is True:
-                        # 2a) Testing-class items require real CI artifact evidence.
-                        if slug in _TESTING_CLASS_SLUGS:
-                            ci_state = get_ci_status(
-                                client, args.owner, args.repo, head_sha
-                            )
-                            if ci_state != "success":
-                                print(
-                                    f"::warning::AI ack for {slug} rejected: "
-                                    f"CI / all-required is {ci_state}, not success",
-                                    file=sys.stderr,
-                                )
-                                rejected_ci_not_green.append(u)
-                                continue
-                        approved.append(u)
-                        continue
-            # If we get here, user is not approved for this slug.
-            rejected_ai_ineligible.append(u)
        return approved

-    ack_state = compute_ack_state(
-        comments, author, items_by_slug, numeric_aliases, probe, high_risk=high_risk
-    )
+    ack_state = compute_ack_state(comments, author, items_by_slug, numeric_aliases, probe)
    body_state = {it["slug"]: section_marker_present(body, it["pr_section_marker"]) for it in items}

    state, description = render_status(items, ack_state, body_state)
@@ -1160,21 +785,9 @@ def main(argv: list[str] | None = None) -> int:
        # were not required (vs a tier:medium+ PR that truly passed all acks).
        state = "success"
        description = f"[info tier:low] {description}"
-    if volume_skipped:
-        # Above the comment-cap — we may have a partial view. Soft-pend
-        # so neither BP nor the author gets stuck; surface the cap so
-        # reviewers know what's up. No-block at the gate level.
-        state = "pending"
-        description = (
-            f"[volume-skipped] comment-cap={max_comments_cap} hit; please file "
-            f"a fresh PR with bot-relay history split off (#369). {description}"
-        )

    # Diagnostics to job log.
-    print(
-        f"::notice::PR #{args.pr} author={author} head={head_sha[:7]} "
-        f"mode={mode} risk_class={'high' if high_risk else 'default'}"
-    )
+    print(f"::notice::PR #{args.pr} author={author} head={head_sha[:7]} mode={mode}")
    for it in items:
        slug = it["slug"]
        ackers = ack_state[slug]["ackers"]
@@ -1205,46 +818,6 @@ def main(argv: list[str] | None = None) -> int:
        description=description, target_url=target_url,
    )
    print(f"::notice::status posted: {args.status_context} → {state}")
-
-    # --- N/A gate status (RFC#324 §N/A follow-up) ---
-    # Post a separate status so review-check.sh can discover N/A declarations
-    # and waive the Gitea-approve requirement for that gate.
-    na_state: dict[str, dict[str, Any]] = {}
-    if na_gates:
-        na_state = compute_na_state(comments, author, na_gates, probe)
-
-        na_descs: list[str] = []
-        for gate, s in na_state.items():
-            if s["declared"]:
-                na_descs.append(gate)
-            decl = s["decl_ackers"]
-            rej = s["rejected"]["not_in_team"]
-            if decl:
-                print(f"::notice::  [N/A OK] {gate} — declared by {','.join(decl)}")
-            if rej:
-                print(
-                    f"::notice::  [N/A REJ] {gate} — not-in-team: {','.join(rej)}",
-                    file=sys.stderr,
-                )
-
-        na_desc = ", ".join(sorted(na_descs)) if na_descs else "(none)"
-        na_status_state = "success" if na_descs else "pending"
-        # review-check.sh reads the description to discover which gates are N/A.
-        # Include the gate names so it can grep for them.
-        na_description = f"N/A: {na_desc}" if na_descs else "N/A: (none)"
-
-        if not args.dry_run:
-            client.post_status(
-                args.owner, args.repo, head_sha,
-                state=na_status_state,
-                context="sop-checklist / na-declarations (pull_request)",
-                description=na_description,
-                target_url=target_url,
-            )
-            print(
-                f"::notice::na-declarations status → {na_status_state}: {na_description}"
-            )
-
    # By default exit 0 — the POSTed status IS the gate, NOT the job
    # conclusion. If the job exits 1 BP will see TWO failure signals
    # (one from the job's auto-status, one from our POST), making the
@@ -104,13 +104,10 @@ if [ "${SOP_REFIRE_DISABLE_RATE_LIMIT:-}" != "1" ]; then
  fi
 fi

-# 3. Invoke sop-tier-check.sh with the env it expects.
-# The canonical workflow intentionally fail-opens the job conclusion
-# (`bash .gitea/scripts/sop-tier-check.sh || true`) while Gitea branch
-# protection enforces reviewer approvals separately. Keep the refire path
-# aligned with that workflow status behavior; otherwise /refire-tier-check can
-# post a hard failure that the canonical pull_request_target workflow would
-# not publish.
+# 3. Invoke sop-tier-check.sh with the env it expects. Capture exit code.
+# The canonical script reads tier label, walks approving reviewers, and
+# evaluates the AND-composition expression — we want the SAME gate, not
+# a different gate.
 #
 # SOP_REFIRE_TIER_CHECK_SCRIPT env var lets tests substitute a mock —
 # sop-tier-check.sh uses bash 4+ associative arrays which trigger a known
@@ -126,6 +123,7 @@ fi

 # Re-invoke. Pipe stdout/stderr through so the runner log shows the
 # tier-check decision inline.
+set +e
 GITEA_TOKEN="$GITEA_TOKEN" \
  GITEA_HOST="$GITEA_HOST" \
  REPO="$REPO" \
@@ -133,8 +131,9 @@ GITEA_TOKEN="$GITEA_TOKEN" \
  PR_AUTHOR="$PR_AUTHOR" \
  SOP_DEBUG="${SOP_DEBUG:-0}" \
  SOP_LEGACY_CHECK="${SOP_LEGACY_CHECK:-0}" \
-  bash "$SCRIPT" || true
-TIER_EXIT=0
+  bash "$SCRIPT"
+TIER_EXIT=$?
+set -e
 debug "sop-tier-check.sh exit=$TIER_EXIT"

 # 4. POST the resulting status.
@@ -47,9 +47,7 @@ What this script does, per `.gitea/workflows/status-reaper.yml` invocation:
         Parse context as `<workflow_name> / <job_name> (push)`.
         Look up workflow_name in the trigger map:
           - missing → log ::notice:: and skip (conservative).
-           - has_push_trigger=True and description == "Has been cancelled"
-             → compensate cancelled/superseded push noise.
-           - has_push_trigger=True otherwise → preserve (real defect signal).
+           - has_push_trigger=True → preserve (real defect signal).
           - has_push_trigger=False → POST a compensating
             `state=success` status to /statuses/{sha} with the same
             context (Gitea de-dups by context) and a description
@@ -143,11 +141,6 @@ PR_SHADOW_COMPENSATION_DESCRIPTION = (
    "shadowed by successful push status on same SHA; see "
    ".gitea/scripts/status-reaper.py)"
 )
-CANCELLED_PUSH_COMPENSATION_DESCRIPTION = (
-    "Compensated by status-reaper (push run was cancelled/superseded; "
-    "Gitea 1.22.6 reports cancelled runs as failure statuses)"
-)
-CANCELLED_DESCRIPTION = "Has been cancelled"

 # Context suffix the reaper acts on. Gitea hardcodes this for ALL
 # default-branch workflow runs.
@@ -483,7 +476,7 @@ def reap(
      {compensated, preserved_real_push, preserved_unknown,
       preserved_non_failure, preserved_non_push_suffix,
       preserved_unparseable, compensated_pr_shadowed_by_push_success,
-       preserved_pr_without_push_success, compensated_cancelled_push,
+       preserved_pr_without_push_success,
       compensated_contexts: [<context>, ...]}

    `compensated_contexts` is rev2-added so `reap_branch` can build
@@ -497,7 +490,6 @@ def reap(
        "preserved_non_push_suffix": 0,
        "preserved_unparseable": 0,
        "compensated_pr_shadowed_by_push_success": 0,
-        "compensated_cancelled_push": 0,
        "preserved_pr_without_push_success": 0,
        "compensated_contexts": [],
    }
@@ -575,27 +567,8 @@ def reap(
            counters["preserved_unknown"] += 1
            continue

-        if (s.get("description") or "").strip() == CANCELLED_DESCRIPTION:
-            # Gitea 1.22.6 maps cancelled action runs to failure commit
-            # statuses. During merge bursts, older push runs can be
-            # superseded and cancelled even though a newer run for the
-            # same branch is the real signal. Compensate only the exact
-            # Gitea cancellation description; real push failures remain red.
-            post_compensating_status(
-                sha,
-                context,
-                s.get("target_url"),
-                description=CANCELLED_PUSH_COMPENSATION_DESCRIPTION,
-                dry_run=dry_run,
-            )
-            counters["compensated"] += 1
-            counters["compensated_cancelled_push"] += 1
-            counters["compensated_contexts"].append(context)
-            continue
-
        if workflow_trigger_map[workflow_name]:
-            # Real push trigger with a non-cancelled failure description
-            # remains a defect signal. Preserve.
+            # Real push trigger → real defect signal. Preserve.
            counters["preserved_real_push"] += 1
            continue

@@ -701,7 +674,6 @@ def reap_branch(
            "preserved_non_push_suffix": 0,
            "preserved_unparseable": 0,
            "compensated_pr_shadowed_by_push_success": 0,
-            "compensated_cancelled_push": 0,
            "preserved_pr_without_push_success": 0,
            "compensated_per_sha": {},
            "skipped": True,
@@ -717,7 +689,6 @@ def reap_branch(
        "preserved_non_push_suffix": 0,
        "preserved_unparseable": 0,
        "compensated_pr_shadowed_by_push_success": 0,
-        "compensated_cancelled_push": 0,
        "preserved_pr_without_push_success": 0,
        "compensated_per_sha": {},
    }
@@ -757,7 +728,6 @@ def reap_branch(
            "preserved_non_push_suffix",
            "preserved_unparseable",
            "compensated_pr_shadowed_by_push_success",
-            "compensated_cancelled_push",
            "preserved_pr_without_push_success",
        ):
            aggregate[key] += per_sha[key]
@@ -33,7 +33,7 @@ def scenario() -> str:
    p = os.path.join(STATE_DIR, "scenario")
    if not os.path.isfile(p):
        return "T1_success"
-    with open(p, encoding="utf-8") as f:
+    with open(p) as f:
        return f.read().strip()


@@ -17,11 +17,6 @@ Scenarios:
  T8_team_not_member          — team membership → 404 (not a member) → exit 1
  T9_team_403                — team membership → 403 (token not in team) → exit 1
  T14_non_default_base        — open PR targeting staging → script exits 0 (no-op)
-  T15_comments_agent_approval — reviews empty; comments have "[core-qa-agent] APPROVED" → exit 0
-  T16_comments_generic_approval — reviews empty; comments have "APPROVED" by team member → exit 0
-  T17_comments_no_approval   — reviews empty; comments have no approval keywords → exit 1
-  T18_review_wrong_team_comment_right_team — review candidate 404s, comment candidate passes
-  T19_ai_sop_ack_approved — ai-sop-ack member APPROVED review → team probe 404 → exit 1

 Usage:
  FIXTURE_STATE_DIR=/tmp/x python3 _review_check_fixture.py 8080
@@ -34,6 +29,7 @@ import re
 import sys
 import urllib.parse

+
 STATE_DIR = os.environ.get("FIXTURE_STATE_DIR", "/tmp")


@@ -41,7 +37,7 @@ def scenario() -> str:
    p = os.path.join(STATE_DIR, "scenario")
    if not os.path.isfile(p):
        return "T1_pr_open"
-    with open(p, encoding="utf-8") as f:
+    with open(p) as f:
        return f.read().strip()


@@ -81,7 +77,7 @@ class Handler(http.server.BaseHTTPRequestHandler):
        # GET /repos/{owner}/{name}/pulls/{pr_number}
        m = re.match(r"^/api/v1/repos/([^/]+)/([^/]+)/pulls/(\d+)$", path)
        if m:
-            pr_num = m.group(3)
+            owner, name, pr_num = m.group(1), m.group(2), m.group(3)
            if sc == "T2_pr_closed":
                return self._json(200, {
                    "number": int(pr_num),
@@ -101,9 +97,7 @@ class Handler(http.server.BaseHTTPRequestHandler):
        # GET /repos/{owner}/{name}/pulls/{pr_number}/reviews
        m = re.match(r"^/api/v1/repos/([^/]+)/([^/]+)/pulls/(\d+)/reviews$", path)
        if m:
-            if sc in ("T4_reviews_empty", "T5_reviews_only_author",
-                      "T15_comments_agent_approval", "T16_comments_generic_approval",
-                      "T17_comments_no_approval"):
+            if sc in ("T4_reviews_empty", "T5_reviews_only_author"):
                return self._json(200, [])
            if sc == "T6_reviews_dismissed":
                return self._json(200, [{
@@ -117,65 +111,22 @@ class Handler(http.server.BaseHTTPRequestHandler):
                    {"state": "CHANGES_REQUESTED", "dismissed": False, "user": {"login": "bob"}, "commit_id": "abc1234"},
                    {"state": "APPROVED", "dismissed": False, "user": {"login": "core-devops"}, "commit_id": "abc1234"},
                ])
-            if sc == "T19_ai_sop_ack_approved":
-                # ai-sop-ack member submitted APPROVED review — must NOT count
-                # toward qa-review (team_id=20) or security-review (team_id=21).
-                return self._json(200, [
-                    {"state": "APPROVED", "dismissed": False, "user": {"login": "ai-reviewer"}, "commit_id": "abc1234"},
-                ])
            # Default: one non-author APPROVED
            return self._json(200, [
                {"state": "APPROVED", "dismissed": False, "user": {"login": "core-devops"}, "commit_id": "abc1234"},
            ])

-        # GET /repos/{owner}/{name}/issues/{pr_number}/comments
-        m = re.match(r"^/api/v1/repos/([^/]+)/([^/]+)/issues/(\d+)/comments$", path)
-        if m:
-            if sc == "T15_comments_agent_approval":
-                return self._json(200, [
-                    {"user": {"login": "core-qa-agent"}, "body": "[core-qa-agent] APPROVED this PR. Good changes.", "id": 1},
-                    {"user": {"login": "alice"}, "body": "I authored this PR", "id": 2},
-                    {"user": {"login": "random-user"}, "body": "Looks okay to me", "id": 3},
-                ])
-            if sc == "T16_comments_generic_approval":
-                return self._json(200, [
-                    {"user": {"login": "core-qa-agent"}, "body": "APPROVED — all acceptance criteria met", "id": 1},
-                    {"user": {"login": "alice"}, "body": "-authored", "id": 2},
-                ])
-            if sc == "T17_comments_no_approval":
-                return self._json(200, [
-                    {"user": {"login": "alice"}, "body": "I authored this PR", "id": 1},
-                    {"user": {"login": "random-user"}, "body": "Looks okay to me", "id": 2},
-                ])
-            if sc == "T18_review_wrong_team_comment_right_team":
-                return self._json(200, [
-                    {"user": {"login": "core-qa-agent"}, "body": "[core-qa-agent] APPROVED after focused review", "id": 1},
-                ])
-            # Default scenarios (T1–T9, T14): no comments
-            return self._json(200, [])
-
        # GET /teams/{team_id}/members/{username}
        m = re.match(r"^/api/v1/teams/(\d+)/members/([^/]+)$", path)
        if m:
-            login = m.group(2)
+            team_id, login = m.group(1), m.group(2)
            if sc == "T8_team_not_member":
                return self._empty(404)
            if sc == "T9_team_403":
                return self._empty(403)
-            if sc == "T18_review_wrong_team_comment_right_team" and login == "core-devops":
-                return self._empty(404)
-            if sc == "T19_ai_sop_ack_approved" and login == "ai-reviewer":
-                # ai-sop-ack member is NOT in qa (20) or security (21).
-                return self._empty(404)
            # T7_team_member: member
            return self._empty(204)

-        # GET /repos/{owner}/{name}/statuses/{sha} — for N/A declaration check
-        m = re.match(r"^/api/v1/repos/([^/]+)/([^/]+)/statuses/([a-f0-9]+)$", path)
-        if m:
-            # All comment-based scenarios have no N/A declarations
-            return self._json(200, [])
-
        return self._json(404, {"path": path, "msg": "fixture: no route"})

    def do_POST(self):
@@ -1,176 +0,0 @@
-import importlib.util
-import sys
-from pathlib import Path
-from unittest.mock import patch
-
-SCRIPT = Path(__file__).resolve().parents[1] / "ci-required-drift.py"
-spec = importlib.util.spec_from_file_location("ci_required_drift", SCRIPT)
-drift = importlib.util.module_from_spec(spec)
-sys.modules[spec.name] = drift
-spec.loader.exec_module(drift)
-
-# Module-level constants are loaded from env at import time; set them
-# explicitly so unit tests can import without the full env contract.
-drift.SENTINEL_JOB = "all-required"
-drift.CI_WORKFLOW_PATH = ".gitea/workflows/ci.yml"
-drift.AUDIT_WORKFLOW_PATH = ".gitea/workflows/audit-force-merge.yml"
-
-
-# ---------------------------------------------------------------------------
-# Helper fixtures
-# ---------------------------------------------------------------------------
-
-def _make_ci_doc(jobs: dict) -> dict:
-    return {"jobs": jobs}
-
-
-def _make_audit_doc(required_checks: list[str]) -> dict:
-    return {
-        "jobs": {
-            "audit": {
-                "steps": [
-                    {"env": {"REQUIRED_CHECKS": "\n".join(required_checks)}}
-                ]
-            }
-        }
-    }
-
-
-# ---------------------------------------------------------------------------
-# sentinel_needs
-# ---------------------------------------------------------------------------
-
-def test_sentinel_needs_returns_empty_when_absent():
-    doc = _make_ci_doc({"all-required": {"runs-on": "ubuntu-latest"}})
-    assert drift.sentinel_needs(doc) == set()
-
-
-def test_sentinel_needs_parses_list():
-    doc = _make_ci_doc(
-        {"all-required": {"needs": ["platform-build", "canvas-build"]}}
-    )
-    assert drift.sentinel_needs(doc) == {"platform-build", "canvas-build"}
-
-
-def test_sentinel_needs_parses_string():
-    doc = _make_ci_doc({"all-required": {"needs": "platform-build"}})
-    assert drift.sentinel_needs(doc) == {"platform-build"}
-
-
-# ---------------------------------------------------------------------------
-# ci_job_names / ci_jobs_all
-# ---------------------------------------------------------------------------
-
-def test_ci_job_names_excludes_sentinel_and_event_gated():
-    doc = _make_ci_doc(
-        {
-            "platform-build": {},
-            "canvas-build": {"if": "github.event_name == 'pull_request'"},
-            "main-push": {"if": "github.ref == 'refs/heads/main'"},
-            "all-required": {},
-        }
-    )
-    assert drift.ci_job_names(doc) == {"platform-build"}
-
-
-def test_ci_jobs_all_includes_event_gated():
-    doc = _make_ci_doc(
-        {
-            "platform-build": {},
-            "canvas-build": {"if": "github.event_name == 'pull_request'"},
-            "all-required": {},
-        }
-    )
-    assert drift.ci_jobs_all(doc) == {"platform-build", "canvas-build"}
-
-
-# ---------------------------------------------------------------------------
-# detect_drift — F1 / F1b with mocked I/O
-# ---------------------------------------------------------------------------
-
-SAMPLE_PROTECTION = {
-    "status_check_contexts": [
-        "CI / all-required (pull_request)",
-        "Secret scan / Scan diff for credential-shaped strings (pull_request)",
-    ]
-}
-
-
-def test_detect_drift_no_needs_sentinel_skips_f1():
-    """Post-#1766 contract: all-required has no needs: → F1 is a false positive."""
-    ci = _make_ci_doc(
-        {
-            "platform-build": {},
-            "canvas-build": {},
-            "all-required": {},
-        }
-    )
-    audit = _make_audit_doc(
-        [
-            "CI / all-required (pull_request)",
-            "Secret scan / Scan diff for credential-shaped strings (pull_request)",
-        ]
-    )
-
-    with patch.object(drift, "load_yaml", side_effect=[ci, audit]):
-        with patch.object(drift, "api", return_value=(200, SAMPLE_PROTECTION)):
-            findings, debug = drift.detect_drift("main")
-
-    assert findings == []
-    assert debug["sentinel_needs"] == []
-
-
-def test_detect_drift_typo_in_needs_triggers_f1b():
-    """F1b still catches typos when needs exists."""
-    ci = _make_ci_doc(
-        {
-            "platform-build": {},
-            "all-required": {"needs": ["platfom-build"]},  # typo
-        }
-    )
-    audit = _make_audit_doc(["CI / all-required (pull_request)"])
-
-    with patch.object(drift, "load_yaml", side_effect=[ci, audit]):
-        with patch.object(drift, "api", return_value=(200, SAMPLE_PROTECTION)):
-            findings, _ = drift.detect_drift("main")
-
-    assert any("F1b" in f for f in findings)
-    assert any("platfom-build" in f for f in findings)
-
-
-def test_detect_drift_missing_job_in_needs_triggers_f1():
-    """F1 still fires when needs is non-empty and jobs are missing."""
-    ci = _make_ci_doc(
-        {
-            "platform-build": {},
-            "canvas-build": {},
-            "all-required": {"needs": ["platform-build"]},
-        }
-    )
-    audit = _make_audit_doc(["CI / all-required (pull_request)"])
-
-    with patch.object(drift, "load_yaml", side_effect=[ci, audit]):
-        with patch.object(drift, "api", return_value=(200, SAMPLE_PROTECTION)):
-            findings, _ = drift.detect_drift("main")
-
-    assert any("F1 —" in f for f in findings)
-    assert any("canvas-build" in f for f in findings)
-    assert not any("F1b" in f for f in findings)
-
-
-def test_detect_drift_no_f1_when_needs_empty_even_with_jobs():
-    """Explicit regression guard: empty needs + existing jobs = no F1."""
-    ci = _make_ci_doc(
-        {
-            "platform-build": {},
-            "canvas-build": {},
-            "all-required": {"needs": []},
-        }
-    )
-    audit = _make_audit_doc(["CI / all-required (pull_request)"])
-
-    with patch.object(drift, "load_yaml", side_effect=[ci, audit]):
-        with patch.object(drift, "api", return_value=(200, SAMPLE_PROTECTION)):
-            findings, _ = drift.detect_drift("main")
-
-    assert not any("F1 —" in f for f in findings)
@@ -1,110 +0,0 @@
-from pathlib import Path
-
-import yaml
-
-
-ROOT = Path(__file__).resolve().parents[2]
-
-
-def load_workflow(name: str) -> dict:
-    with (ROOT / "workflows" / name).open() as f:
-        return yaml.safe_load(f)
-
-
-def _all_required(workflow: dict) -> dict:
-    return workflow["jobs"]["all-required"]
-
-
-def test_all_required_uses_dedicated_meta_runner_lane():
-    workflow = load_workflow("ci.yml")
-    all_required = _all_required(workflow)
-
-    # Stays on the dedicated `ci-meta` lane (the sentinel does no docker
-    # work, so it must NOT occupy the general docker-host pool).
-    assert all_required["runs-on"] == "ci-meta"
-
-
-def test_all_required_is_needs_aggregator_not_a_polling_gate():
-    """fix/ci-scheduler-fanout (2026-06-01): the sentinel was converted
-    from a status-polling loop (which squatted a ci-meta executor slot for
-    up to 40 min per PR) into a plain `needs:` aggregator that frees the
-    slot immediately. Pin the new shape so a regression to the poller is
-    caught.
-    """
-    workflow = load_workflow("ci.yml")
-    all_required = _all_required(workflow)
-    rendered = str(all_required)
-
-    # The job MUST aggregate via `needs:` (the slot-freeing design).
-    assert "needs" in all_required, "all-required must be a needs: aggregator"
-
-    # It MUST NOT reintroduce the polling loop / per-SHA status fetch that
-    # was the throughput sink.
-    assert "detect-changes.py" not in rendered, (
-        "all-required must not run the detect-changes poller path"
-    )
-    assert "commits/" not in rendered and "statuses" not in rendered, (
-        "all-required must not poll commit statuses (the slot-squat path)"
-    )
-
-
-def test_all_required_does_not_use_if_always():
-    """Plain `needs:` works on Gitea 1.22.6 / act_runner v0.6.1; `needs:` +
-    `if: always()` is BROKEN (feedback_gitea_needs_works_only_ifalways_broken)
-    and would let a non-success need pass the gate. The sentinel must use
-    plain `needs:` WITHOUT a job-level `if: always()`.
-    """
-    workflow = load_workflow("ci.yml")
-    all_required = _all_required(workflow)
-
-    job_if = all_required.get("if")
-    assert not (isinstance(job_if, str) and "always()" in job_if), (
-        "all-required must not combine needs: with if: always()"
-    )
-
-
-def test_all_required_needs_matches_ci_required_drift_f1_set():
-    """The sentinel `needs:` list MUST equal ci-required-drift.py's
-    `ci_job_names()` set: every job MINUS the sentinel itself MINUS jobs
-    whose `if:` gates on github.event_name/github.ref (event-gated jobs
-    skip on PRs and a `needs:` on a skipped job would never let the
-    sentinel run). If they diverge, ci-required-drift F1 fires.
-    """
-    workflow = load_workflow("ci.yml")
-    jobs = workflow["jobs"]
-    sentinel = "all-required"
-
-    expected = set()
-    for key, body in jobs.items():
-        if key == sentinel:
-            continue
-        gate = body.get("if") if isinstance(body, dict) else None
-        if isinstance(gate, str) and (
-            "github.event_name" in gate or "github.ref" in gate
-        ):
-            # event-gated → legitimately skips on some triggers; excluded
-            # from both `needs:` and the F1 set.
-            continue
-        expected.add(key)
-
-    needs = jobs[sentinel].get("needs", [])
-    if isinstance(needs, str):
-        needs = [needs]
-    actual = set(needs)
-
-    assert actual == expected, (
-        f"all-required needs: {sorted(actual)} != ci_job_names() "
-        f"{sorted(expected)} — ci-required-drift F1 would fire"
-    )
-
-
-def test_all_required_needs_reference_real_jobs():
-    """F1b guard: every entry in `needs:` must name an existing job."""
-    workflow = load_workflow("ci.yml")
-    jobs = workflow["jobs"]
-    needs = jobs["all-required"].get("needs", [])
-    if isinstance(needs, str):
-        needs = [needs]
-    job_keys = set(jobs)
-    for dep in needs:
-        assert dep in job_keys, f"all-required needs unknown job {dep!r}"
@@ -1,168 +0,0 @@
-"""Regression test #765 — gate auto-fire on real qa/security APPROVED review.
-
-Validates the structural configuration of qa-review.yml and security-review.yml
-so that a real team-member APPROVED review fires the workflow and POSTs the
-exact branch-protection-required context name. This is the test #2020's
-stale-context failure would have caught.
-"""
-
-from pathlib import Path
-
-import yaml
-
-ROOT = Path(__file__).resolve().parents[2]
-
-
-def load_workflow(name: str) -> dict:
-    with (ROOT / "workflows" / name).open() as f:
-        return yaml.safe_load(f)
-
-
-def _job_guard_string(workflow: dict) -> str:
-    """Return the raw job-level `if:` string for the single job."""
-    jobs = workflow["jobs"]
-    # Both qa-review and security-review have exactly one job named "approved".
-    job = jobs["approved"]
-    return str(job.get("if", ""))
-
-
-def _post_step(workflow: dict) -> dict:
-    """Return the explicit POST /statuses step from the job steps list."""
-    jobs = workflow["jobs"]
-    steps = jobs["approved"]["steps"]
-    for step in steps:
-        name = step.get("name", "")
-        if "Post required status context" in name:
-            return step
-    raise AssertionError("No explicit POST status step found")
-
-
-class TestQaReviewDirectTrigger:
-    def test_trigger_is_pull_request_review_submitted(self):
-        wf = load_workflow("qa-review.yml")
-        # PyYAML parses bare 'on' as boolean True.
-        on = wf[True]
-        assert "pull_request_review" in on, (
-            "qa-review must trigger on pull_request_review"
-        )
-        types = on["pull_request_review"].get("types", [])
-        assert "submitted" in types, (
-            "pull_request_review must include 'submitted' type"
-        )
-
-    def test_job_guard_requires_approved_state(self):
-        wf = load_workflow("qa-review.yml")
-        guard = _job_guard_string(wf)
-        assert "github.event.review.state == 'APPROVED'" in guard, (
-            "job guard must check review.state for 'APPROVED'"
-        )
-        assert "github.event.review.state == 'approved'" in guard, (
-            "job guard must check review.state for 'approved' (case fallback per #2135)"
-        )
-
-    def test_post_step_uses_status_post_token(self):
-        wf = load_workflow("qa-review.yml")
-        post = _post_step(wf)
-        env = post.get("env", {})
-        assert env.get("GITEA_TOKEN") == "${{ secrets.STATUS_POST_TOKEN }}", (
-            "POST step must use STATUS_POST_TOKEN for write-scoped status POST"
-        )
-
-    def test_post_step_context_name_exact(self):
-        """The context POSTed must byte-match the branch-protection requirement."""
-        wf = load_workflow("qa-review.yml")
-        post = _post_step(wf)
-        run = post.get("run", "")
-        assert '"qa-review / approved (pull_request_target)"' in run, (
-            "POST step must emit exact BP-required context name"
-        )
-
-
-class TestSecurityReviewDirectTrigger:
-    def test_trigger_is_pull_request_review_submitted(self):
-        wf = load_workflow("security-review.yml")
-        # PyYAML parses bare 'on' as boolean True.
-        on = wf[True]
-        assert "pull_request_review" in on, (
-            "security-review must trigger on pull_request_review"
-        )
-        types = on["pull_request_review"].get("types", [])
-        assert "submitted" in types, (
-            "pull_request_review must include 'submitted' type"
-        )
-
-    def test_job_guard_requires_approved_state(self):
-        wf = load_workflow("security-review.yml")
-        guard = _job_guard_string(wf)
-        assert "github.event.review.state == 'APPROVED'" in guard, (
-            "job guard must check review.state for 'APPROVED'"
-        )
-        assert "github.event.review.state == 'approved'" in guard, (
-            "job guard must check review.state for 'approved' (case fallback per #2135)"
-        )
-
-    def test_post_step_uses_status_post_token(self):
-        wf = load_workflow("security-review.yml")
-        post = _post_step(wf)
-        env = post.get("env", {})
-        assert env.get("GITEA_TOKEN") == "${{ secrets.STATUS_POST_TOKEN }}", (
-            "POST step must use STATUS_POST_TOKEN for write-scoped status POST"
-        )
-
-    def test_post_step_context_name_exact(self):
-        """The context POSTed must byte-match the branch-protection requirement."""
-        wf = load_workflow("security-review.yml")
-        post = _post_step(wf)
-        run = post.get("run", "")
-        assert '"security-review / approved (pull_request_target)"' in run, (
-            "POST step must emit exact BP-required context name"
-        )
-
-
-class TestRefireScriptContextName:
-    """review-refire-status.sh must emit the BP-required (pull_request_target) context."""
-
-    def test_refire_script_context_is_pull_request_target(self):
-        script = ROOT / "scripts" / "review-refire-status.sh"
-        content = script.read_text()
-        assert 'CONTEXT="${TEAM}-review / approved (pull_request_target)"' in content, (
-            "refire script CONTEXT must be the exact BP-required (pull_request_target) variant"
-        )
-        assert 'approved (pull_request)"' not in content, (
-            "refire script must NOT post bare (pull_request) context"
-        )
-
-
-class TestRefireTokenSeparation:
-    """The /qa-recheck + /security-recheck backstop must also use STATUS_POST_TOKEN."""
-
-    def _refire_step(self, workflow_name: str, step_name_keyword: str) -> dict:
-        wf = load_workflow(workflow_name)
-        jobs = wf["jobs"]
-        steps = jobs["review-refire"]["steps"]
-        for step in steps:
-            name = step.get("name", "")
-            if step_name_keyword in name:
-                return step
-        raise AssertionError(f"No refire step matching {step_name_keyword!r}")
-
-    def test_qa_refire_uses_status_post_token(self):
-        step = self._refire_step("sop-checklist.yml", "Refire qa-review")
-        env = step.get("env", {})
-        assert env.get("STATUS_POST_TOKEN") == "${{ secrets.STATUS_POST_TOKEN }}", (
-            "qa refire must receive STATUS_POST_TOKEN env var"
-        )
-        # Evaluator stays on read token
-        assert "SOP_TIER_CHECK_TOKEN" in env.get("GITEA_TOKEN", "") or "GITHUB_TOKEN" in env.get("GITEA_TOKEN", ""), (
-            "qa refire evaluator must stay on read-scoped token"
-        )
-
-    def test_security_refire_uses_status_post_token(self):
-        step = self._refire_step("sop-checklist.yml", "Refire security-review")
-        env = step.get("env", {})
-        assert env.get("STATUS_POST_TOKEN") == "${{ secrets.STATUS_POST_TOKEN }}", (
-            "security refire must receive STATUS_POST_TOKEN env var"
-        )
-        assert "SOP_TIER_CHECK_TOKEN" in env.get("GITEA_TOKEN", "") or "GITHUB_TOKEN" in env.get("GITEA_TOKEN", ""), (
-            "security refire evaluator must stay on read-scoped token"
-        )
@@ -2,6 +2,7 @@ import importlib.util
 import sys
 from pathlib import Path

+
 SCRIPT = Path(__file__).resolve().parents[1] / "gitea-merge-queue.py"
 spec = importlib.util.spec_from_file_location("gitea_merge_queue", SCRIPT)
 mq = importlib.util.module_from_spec(spec)
@@ -117,13 +118,3 @@ def test_merge_decision_updates_stale_pr_before_merge():

    assert decision.ready is False
    assert decision.action == "update"
-
-
-def test_MergePermissionError_inherits_from_ApiError():
-    assert issubclass(mq.MergePermissionError, mq.ApiError)
-
-
-def test_MergePermissionError_message_preserved():
-    exc = mq.MergePermissionError("POST /merge -> HTTP 405: User not allowed")
-    assert "405" in str(exc)
-    assert "User not allowed" in str(exc)
@@ -15,6 +15,7 @@ Mirrors the pattern in scripts/ops/test_check_migration_collisions.py
 from __future__ import annotations

 import importlib.util
+import os
 import sys
 import unittest
 from pathlib import Path
@@ -1,283 +0,0 @@
-import importlib.util
-import sys
-from pathlib import Path
-from unittest.mock import patch, MagicMock
-
-SCRIPT = Path(__file__).resolve().parents[1] / "main-red-watchdog.py"
-spec = importlib.util.spec_from_file_location("main_red_watchdog", SCRIPT)
-wd = importlib.util.module_from_spec(spec)
-sys.modules[spec.name] = wd
-spec.loader.exec_module(wd)
-
-# Module-level constants are loaded from env at import time; set them
-# explicitly so unit tests can import without the full env contract.
-wd.GITEA_TOKEN = "fake-token"
-wd.GITEA_HOST = "git.example.com"
-wd.REPO = "molecule-ai/molecule-core"
-wd.OWNER = "molecule-ai"
-wd.NAME = "molecule-core"
-wd.WATCH_BRANCH = "main"
-wd.RED_LABEL = "tier:high"
-wd.API = "https://git.example.com/api/v1"
-
-
-# ---------------------------------------------------------------------------
-# _is_scheduled_context
-# ---------------------------------------------------------------------------
-
-def test_is_scheduled_context_matches_staging_saas_smoke():
-    assert wd._is_scheduled_context("Staging SaaS smoke") is True
-
-
-def test_is_scheduled_context_matches_case_insensitive():
-    assert wd._is_scheduled_context("continuous synthetic e2e") is True
-
-
-def test_is_scheduled_context_no_match_for_required_ci():
-    assert wd._is_scheduled_context("CI / all-required") is False
-
-
-# ---------------------------------------------------------------------------
-# _entry_state
-# ---------------------------------------------------------------------------
-
-def test_entry_state_prefers_status_over_state():
-    """Gitea 1.22.6 per-entry key is `status`; `state` is fallback."""
-    assert wd._entry_state({"status": "failure", "state": "success"}) == "failure"
-
-
-def test_entry_state_falls_back_to_state():
-    assert wd._entry_state({"state": "pending"}) == "pending"
-
-
-def test_entry_state_empty_when_neither_key_present():
-    assert wd._entry_state({"context": "foo"}) == ""
-
-
-# ---------------------------------------------------------------------------
-# is_red
-# ---------------------------------------------------------------------------
-
-def test_is_red_combined_failure_no_statuses():
-    """Combined failure with empty statuses[] still trips red."""
-    red, failed = wd.is_red({"state": "failure", "statuses": []})
-    assert red is True
-    assert failed == []
-
-
-def test_is_red_cancel_cascade_filtered():
-    """status=3 (cancelled) mapped to failure string must be filtered."""
-    status = {
-        "state": "failure",
-        "statuses": [
-            {"context": "CI / build", "status": "failure", "description": "Has been cancelled"},
-        ],
-    }
-    red, failed = wd.is_red(status)
-    assert red is False
-    assert failed == []
-
-
-def test_is_red_real_failure_not_filtered():
-    """Real failures with different descriptions are kept."""
-    status = {
-        "state": "failure",
-        "statuses": [
-            {"context": "CI / build", "status": "failure", "description": "Failing after 12s"},
-        ],
-    }
-    red, failed = wd.is_red(status)
-    assert red is True
-    assert len(failed) == 1
-    assert failed[0]["context"] == "CI / build"
-
-
-def test_is_red_uses_entry_state_not_top_level_state():
-    """Regression: per-entry key is `status`, not `state`."""
-    status = {
-        "state": "failure",
-        "statuses": [
-            # Only `status` present; pre-rev4 code read `state` and got None
-            {"context": "CI / test", "status": "failure"},
-        ],
-    }
-    red, failed = wd.is_red(status)
-    assert red is True
-    assert len(failed) == 1
-
-
-# ---------------------------------------------------------------------------
-# list_open_red_issues — pagination (mc#1789)
-# ---------------------------------------------------------------------------
-
-def test_list_open_red_issues_exhausts_pagination():
-    """Backlog can exceed 50 issues; all pages must be fetched."""
-    calls = []
-
-    def fake_api(method, path, **kwargs):
-        calls.append((method, path, kwargs))
-        query = (kwargs.get("query") or {})
-        page = int(query.get("page", "1"))
-        limit = int(query.get("limit", "50"))
-        # Page 1 returns full limit; page 2 returns partial → break
-        if page == 1:
-            return 200, [
-                {"title": f"[main-red] molecule-ai/molecule-core: sha{i:04d}"}
-                for i in range(limit)
-            ]
-        if page == 2:
-            return 200, [
-                {"title": "[main-red] molecule-ai/molecule-core: extra1"},
-                {"title": "[main-red] molecule-ai/molecule-core: extra2"},
-                {"title": " unrelated issue "},  # filtered out
-            ]
-        return 200, []
-
-    with patch.object(wd, "api", side_effect=fake_api):
-        issues = wd.list_open_red_issues()
-
-    assert len(issues) == 52  # 50 + 2 matched
-    titles = {i["title"] for i in issues}
-    assert "[main-red] molecule-ai/molecule-core: extra1" in titles
-    assert "[main-red] molecule-ai/molecule-core: extra2" in titles
-
-
-def test_list_open_red_issues_single_page():
-    """When results < limit, loop breaks after first page."""
-    def fake_api(method, path, **kwargs):
-        return 200, [
-            {"title": "[main-red] molecule-ai/molecule-core: abc123"},
-        ]
-
-    with patch.object(wd, "api", side_effect=fake_api):
-        issues = wd.list_open_red_issues()
-
-    assert len(issues) == 1
-
-
-# ---------------------------------------------------------------------------
-# run_once — close logic (mc#1789)
-# ---------------------------------------------------------------------------
-
-def test_run_once_green_closes_stale_issues(monkeypatch):
-    """Combined success → close stale issues."""
-    monkeypatch.setattr(wd, "get_head_sha", lambda b: "abc123")
-    monkeypatch.setattr(wd, "get_combined_status", lambda s: {"state": "success", "statuses": []})
-    monkeypatch.setattr(wd, "is_red", lambda s: (False, []))
-
-    closed = []
-
-    def capture_close(current_sha, *, dry_run=False, close_same_sha=False):
-        closed.append(current_sha)
-        return 1
-
-    monkeypatch.setattr(wd, "close_open_red_issues_for_other_shas", capture_close)
-    monkeypatch.setattr(wd, "emit_loki_event", lambda *a, **k: None)
-
-    assert wd.run_once(dry_run=True) == 0
-    assert closed == ["abc123"]
-
-
-def test_run_once_pending_scheduled_only_closes_stale_issues(monkeypatch):
-    """Combined pending, but only scheduled contexts pending → close stale."""
-    monkeypatch.setattr(wd, "get_head_sha", lambda b: "abc123")
-    monkeypatch.setattr(
-        wd, "get_combined_status",
-        lambda s: {
-            "state": "pending",
-            "statuses": [
-                {"context": "CI / all-required", "status": "success"},
-                {"context": "Staging SaaS smoke", "status": "pending"},
-            ],
-        }
-    )
-    monkeypatch.setattr(wd, "is_red", lambda s: (False, []))
-
-    closed = []
-
-    def capture_close(current_sha, *, dry_run=False, close_same_sha=False):
-        closed.append(current_sha)
-        return 1
-
-    monkeypatch.setattr(wd, "close_open_red_issues_for_other_shas", capture_close)
-    monkeypatch.setattr(wd, "emit_loki_event", lambda *a, **k: None)
-
-    assert wd.run_once(dry_run=True) == 0
-    assert closed == ["abc123"]
-
-
-def test_run_once_pending_required_does_not_close(monkeypatch):
-    """Combined pending with a real required context still pending → no close."""
-    monkeypatch.setattr(wd, "get_head_sha", lambda b: "abc123")
-    monkeypatch.setattr(
-        wd, "get_combined_status",
-        lambda s: {
-            "state": "pending",
-            "statuses": [
-                {"context": "CI / all-required", "status": "pending"},
-                {"context": "Staging SaaS smoke", "status": "success"},
-            ],
-        }
-    )
-    monkeypatch.setattr(wd, "is_red", lambda s: (False, []))
-
-    closed = []
-
-    def capture_close(current_sha, *, dry_run=False, close_same_sha=False):
-        closed.append(current_sha)
-        return 0
-
-    monkeypatch.setattr(wd, "close_open_red_issues_for_other_shas", capture_close)
-    monkeypatch.setattr(wd, "emit_loki_event", lambda *a, **k: None)
-
-    assert wd.run_once(dry_run=True) == 0
-    assert closed == []
-
-
-def test_run_once_failure_does_not_close(monkeypatch):
-    """Real failure in non-scheduled context → no close."""
-    monkeypatch.setattr(wd, "get_head_sha", lambda b: "abc123")
-    monkeypatch.setattr(
-        wd, "get_combined_status",
-        lambda s: {
-            "state": "failure",
-            "statuses": [
-                {"context": "CI / all-required", "status": "failure"},
-            ],
-        }
-    )
-    # is_red will return True, so we enter the red path, not the green close path
-    monkeypatch.setattr(wd, "is_red", lambda s: (True, s.get("statuses", [])))
-    monkeypatch.setattr(wd, "time", MagicMock(sleep=lambda x: None))
-    monkeypatch.setattr(wd, "emit_loki_event", lambda *a, **k: None)
-
-    filed = []
-
-    def capture_file(sha, failed, debug, *, dry_run=False):
-        filed.append(sha)
-
-    monkeypatch.setattr(wd, "file_or_update_red", capture_file)
-    monkeypatch.setattr(wd, "close_open_red_issues_for_other_shas", lambda *a, **k: 0)
-    monkeypatch.setattr(wd, "close_stale_red_issues", lambda *a, **k: 0)
-
-    assert wd.run_once(dry_run=True) == 0
-    assert filed == ["abc123"]
-
-
-# ---------------------------------------------------------------------------
-# title_for / find_open_issue_for_sha
-# ---------------------------------------------------------------------------
-
-def test_title_for_uses_short_sha():
-    assert wd.title_for("abcdef123456") == "[main-red] molecule-ai/molecule-core: abcdef1234"
-
-
-def test_find_open_issue_for_sha_matches_exact_title(monkeypatch):
-    fake_issue = {"title": "[main-red] molecule-ai/molecule-core: abc1234567", "number": 42}
-    monkeypatch.setattr(wd, "list_open_red_issues", lambda: [fake_issue])
-    assert wd.find_open_issue_for_sha("abc1234567") == fake_issue
-
-
-def test_find_open_issue_for_sha_returns_none_when_no_match(monkeypatch):
-    monkeypatch.setattr(wd, "list_open_red_issues", lambda: [])
-    assert wd.find_open_issue_for_sha("abc123") is None
@@ -36,37 +36,9 @@ def test_build_plan_defaults_to_staging_sha_target_and_prod_cp():
        "soak_seconds": 60,
        "batch_size": 3,
        "dry_run": False,
-        # cp#228 / task #308: fleet-wide intent must carry confirm:true.
-        "confirm": True,
    }


-def test_build_plan_always_sets_confirm_true_for_fleet_intent():
-    """Regression guard: every plan body MUST carry confirm:true.
-
-    CP /cp/admin/tenants/redeploy-fleet (cp#228) returns 400 on empty
-    body / {confirm:false} / {only_slugs:[]} to prevent accidental
-    fleet-wide mutation. This caller is fleet-wide intent (canary +
-    fan-out, no slug scoping), so the plan MUST carry confirm:true.
-    Pairs with cp#228's TestRedeployFleet_EmptyBodyReturns400 +
-    TestRedeployFleet_ConfirmTrueProceeds.
-    """
-    plan = prod.build_plan({"GITHUB_SHA": "abcdef1234567890"})
-    assert plan["body"]["confirm"] is True
-
-    # Operator-overridable knobs do NOT drop the ack.
-    plan = prod.build_plan(
-        {
-            "GITHUB_SHA": "abcdef1234567890",
-            "PROD_AUTO_DEPLOY_SOAK_SECONDS": "0",
-            "PROD_AUTO_DEPLOY_BATCH_SIZE": "10",
-            "PROD_AUTO_DEPLOY_DRY_RUN": "true",
-            "PROD_AUTO_DEPLOY_CANARY_SLUG": "",
-        }
-    )
-    assert plan["body"]["confirm"] is True
-
-
 def test_build_plan_rejects_non_prod_cp_without_explicit_override():
    try:
        prod.build_plan(
@@ -146,343 +118,3 @@ def test_context_is_terminal_failure_rejects_cancelled_and_skipped():
        assert prod.context_is_terminal_failure(state) is True
    for state in ("pending", "missing", "success"):
        assert prod.context_is_terminal_failure(state) is False
-
-
-def test_default_required_contexts_delegate_path_gating_to_all_required():
-    assert prod.required_contexts({}) == [
-        "CI / all-required (push)",
-        "Secret scan / Scan diff for credential-shaped strings (push)",
-    ]
-
-
-def test_slugs_from_redeploy_response_uses_controlplane_plan_rows():
-    body = {
-        "results": [
-            {"slug": "hongming", "phase": "canary", "ssm_status": "DryRun"},
-            {"slug": "tenant-a", "phase": "batch-1", "ssm_status": "DryRun"},
-            {"slug": "", "phase": "batch-1", "ssm_status": "DryRun"},
-            {"phase": "batch-1", "ssm_status": "DryRun"},
-        ]
-    }
-
-    assert prod.slugs_from_redeploy_response(body) == ["hongming", "tenant-a"]
-
-
-def test_plan_rollout_slugs_asks_controlplane_for_dry_run_plan():
-    calls = []
-
-    def fake_redeploy(_cp_url, _token, body):
-        calls.append(body)
-        return 200, {
-            "ok": True,
-            "results": [
-                {"slug": "hongming", "phase": "canary", "ssm_status": "DryRun"},
-                {"slug": "tenant-a", "phase": "batch-1", "ssm_status": "DryRun"},
-            ],
-        }
-
-    slugs = prod.plan_rollout_slugs(
-        "https://api.moleculesai.app",
-        "secret",
-        {
-            "target_tag": "staging-abcdef1",
-            "canary_slug": "hongming",
-            "soak_seconds": 60,
-            "batch_size": 3,
-            "dry_run": False,
-            "confirm": True,
-        },
-        redeploy=fake_redeploy,
-    )
-
-    assert slugs == ["hongming", "tenant-a"]
-    assert calls == [
-        {
-            "target_tag": "staging-abcdef1",
-            "canary_slug": "hongming",
-            "soak_seconds": 60,
-            "batch_size": 3,
-            "dry_run": True,
-            "confirm": True,
-        }
-    ]
-
-
-def test_scoped_redeploy_body_removes_canary_and_local_soak():
-    base = {
-        "target_tag": "staging-abcdef1",
-        "canary_slug": "hongming",
-        "soak_seconds": 60,
-        "batch_size": 3,
-        "dry_run": False,
-        "confirm": True,
-    }
-
-    scoped = prod.scoped_redeploy_body(base, ["tenant-a", "tenant-b"])
-
-    assert scoped == {
-        "target_tag": "staging-abcdef1",
-        "soak_seconds": 0,
-        "batch_size": 2,
-        "dry_run": False,
-        "confirm": True,
-        "only_slugs": ["tenant-a", "tenant-b"],
-    }
-
-
-def test_plan_scoped_rollout_preserves_canary_then_batches():
-    calls, sleeps = [], []
-
-    def fake_list(_cp_url, _token, _body):
-        return ["tenant-a", "hongming", "tenant-b", "tenant-c"]
-
-    def fake_redeploy(_cp_url, _token, body):
-        calls.append(body)
-        return 200, {
-            "ok": True,
-            "results": [{"slug": slug, "healthz_ok": True} for slug in body["only_slugs"]],
-        }
-
-    aggregate = prod.execute_scoped_rollout(
-        {
-            "cp_url": "https://api.moleculesai.app",
-            "body": {
-                "target_tag": "staging-abcdef1",
-                "canary_slug": "hongming",
-                "soak_seconds": 60,
-                "batch_size": 2,
-                "dry_run": False,
-                "confirm": True,
-            },
-        },
-        token="secret",
-        list_slugs=fake_list,
-        redeploy=fake_redeploy,
-        sleep=sleeps.append,
-    )
-
-    assert [call["only_slugs"] for call in calls] == [
-        ["hongming"],
-        ["tenant-a", "tenant-b"],
-        ["tenant-c"],
-    ]
-    assert sleeps == [60]
-    assert aggregate["ok"] is True
-    assert [result["slug"] for result in aggregate["results"]] == [
-        "hongming",
-        "tenant-a",
-        "tenant-b",
-        "tenant-c",
-    ]
-
-
-def test_scoped_rollout_halts_after_failed_canary():
-    calls = []
-
-    def fake_redeploy(_cp_url, _token, body):
-        calls.append(body)
-        return 200, {"ok": False, "results": [{"slug": body["only_slugs"][0], "error": "bad"}]}
-
-    try:
-        prod.execute_scoped_rollout(
-            {
-                "cp_url": "https://api.moleculesai.app",
-                "body": {
-                    "target_tag": "staging-abcdef1",
-                    "canary_slug": "hongming",
-                    "soak_seconds": 60,
-                    "batch_size": 2,
-                    "dry_run": False,
-                    "confirm": True,
-                },
-            },
-            token="secret",
-            list_slugs=lambda _cp_url, _token, _body: ["hongming", "tenant-a"],
-            redeploy=fake_redeploy,
-            sleep=lambda _seconds: None,
-        )
-    except prod.RolloutFailed as exc:
-        assert "redeploy scoped call failed" in str(exc)
-        assert exc.response["ok"] is False
-        assert exc.response["results"] == [{"slug": "hongming", "error": "bad"}]
-    else:
-        raise AssertionError("expected failed canary to halt rollout")
-
-    assert [call["only_slugs"] for call in calls] == [["hongming"]]
-
-
-def test_rollout_from_plan_file_writes_partial_response_on_failure(tmp_path):
-    plan_path = tmp_path / "plan.json"
-    response_path = tmp_path / "response.json"
-    plan_path.write_text(
-        """
-        {
-          "enabled": true,
-          "cp_url": "https://api.moleculesai.app",
-          "body": {"target_tag": "staging-abcdef1", "confirm": true}
-        }
-        """,
-        encoding="utf-8",
-    )
-
-    original = prod.execute_scoped_rollout
-
-    def fake_execute(_plan, _token):
-        raise prod.RolloutFailed(
-            "redeploy scoped call failed for hongming: HTTP 500, ok=false",
-            {
-                "ok": False,
-                "error": "redeploy scoped call failed for hongming: HTTP 500, ok=false",
-                "results": [{"slug": "hongming", "error": "bad"}],
-            },
-        )
-
-    prod.execute_scoped_rollout = fake_execute
-    try:
-        try:
-            prod.rollout_from_plan_file(
-                str(plan_path),
-                str(response_path),
-                {"CP_ADMIN_API_TOKEN": "secret"},
-            )
-        except prod.RolloutFailed:
-            pass
-        else:
-            raise AssertionError("expected rollout failure")
-    finally:
-        prod.execute_scoped_rollout = original
-
-    assert response_path.read_text(encoding="utf-8").strip()
-    assert '"ok": false' in response_path.read_text(encoding="utf-8")
-    assert '"slug": "hongming"' in response_path.read_text(encoding="utf-8")
-
-
-# ──────────────────────────────────────────────────────────────────────
-# No-silent-skip coverage gate (internal#724)
-# ──────────────────────────────────────────────────────────────────────
-
-
-def test_rollout_stragglers_flags_tenant_not_on_target():
-    # b SSM-succeeded but its container is on the old tag → straggler.
-    stragglers = prod.rollout_stragglers(
-        ["a", "b", "c"],
-        [
-            {"slug": "a", "verified_on_target": True},
-            {"slug": "b", "verified_on_target": False, "running_image": "platform-tenant:staging-old"},
-            {"slug": "c", "verified_on_target": True},
-        ],
-    )
-    assert stragglers == ["b"]
-
-
-def test_rollout_stragglers_flags_enumerated_tenant_with_no_result():
-    # agents-team class: enumerated but no batch ever produced a row for it.
-    stragglers = prod.rollout_stragglers(
-        ["a", "agents-team"],
-        [{"slug": "a", "verified_on_target": True}],
-    )
-    assert stragglers == ["agents-team"]
-
-
-def test_rollout_stragglers_missing_key_is_backward_compatible():
-    # Older CP without verified_on_target → treat as verified (no spurious fail).
-    stragglers = prod.rollout_stragglers(
-        ["a", "b"],
-        [{"slug": "a", "healthz_ok": True}, {"slug": "b", "healthz_ok": True}],
-    )
-    assert stragglers == []
-
-
-def test_rollout_stragglers_ignores_dry_run_rows():
-    stragglers = prod.rollout_stragglers(
-        ["a"], [{"slug": "a", "ssm_status": "DryRun"}]
-    )
-    # dry-run row is skipped, so "a" has no verifying row → straggler.
-    assert stragglers == ["a"]
-
-
-def test_scoped_rollout_fails_when_a_tenant_stays_on_old_tag():
-    # Every per-tenant call returns ok=True, but agents-team is NOT
-    # verified_on_target. The rollout must still fail loudly — this is
-    # the exact "reported success, one tenant silently skipped" bug.
-    def fake_redeploy(_cp_url, _token, body):
-        rows = []
-        for slug in body["only_slugs"]:
-            rows.append({"slug": slug, "verified_on_target": slug != "agents-team"})
-        return 200, {"ok": True, "results": rows}
-
-    try:
-        prod.execute_scoped_rollout(
-            {
-                "cp_url": "https://api.moleculesai.app",
-                "body": {
-                    "target_tag": "staging-new",
-                    "batch_size": 5,
-                    "dry_run": False,
-                    "confirm": True,
-                },
-            },
-            token="secret",
-            list_slugs=lambda _u, _t, _b: ["reno-stars", "agents-team", "hongming"],
-            redeploy=fake_redeploy,
-            sleep=lambda _s: None,
-        )
-    except prod.RolloutFailed as exc:
-        assert "incomplete rollout" in str(exc)
-        assert exc.response["stragglers"] == ["agents-team"]
-        assert exc.response["ok"] is False
-    else:
-        raise AssertionError("expected an incomplete rollout to fail loudly")
-
-
-def test_scoped_rollout_passes_when_all_tenants_verified_on_target():
-    def fake_redeploy(_cp_url, _token, body):
-        return 200, {
-            "ok": True,
-            "results": [{"slug": s, "verified_on_target": True} for s in body["only_slugs"]],
-        }
-
-    aggregate = prod.execute_scoped_rollout(
-        {
-            "cp_url": "https://api.moleculesai.app",
-            "body": {
-                "target_tag": "staging-new",
-                "batch_size": 5,
-                "dry_run": False,
-                "confirm": True,
-            },
-        },
-        token="secret",
-        list_slugs=lambda _u, _t, _b: ["reno-stars", "agents-team", "hongming"],
-        redeploy=fake_redeploy,
-        sleep=lambda _s: None,
-    )
-    assert aggregate["ok"] is True
-    assert "stragglers" not in aggregate
-
-
-def test_scoped_rollout_dry_run_does_not_assert_coverage():
-    # A dry run proves nothing landed; coverage must NOT be asserted or
-    # every plan would fail.
-    def fake_redeploy(_cp_url, _token, body):
-        return 200, {
-            "ok": True,
-            "results": [{"slug": s, "ssm_status": "DryRun"} for s in body["only_slugs"]],
-        }
-
-    aggregate = prod.execute_scoped_rollout(
-        {
-            "cp_url": "https://api.moleculesai.app",
-            "body": {
-                "target_tag": "staging-new",
-                "batch_size": 5,
-                "dry_run": True,
-                "confirm": True,
-            },
-        },
-        token="secret",
-        list_slugs=lambda _u, _t, _b: ["a", "b"],
-        redeploy=fake_redeploy,
-        sleep=lambda _s: None,
-    )
-    assert aggregate["ok"] is True
@@ -1,5 +1,4 @@
 #!/usr/bin/env bash
-# shellcheck disable=SC2034
 # Regression tests for .gitea/scripts/review-check.sh (RFC#324 Step 1).
 #
 # Covers:
@@ -17,7 +16,6 @@
 #   T12 — jq filter: non-author APPROVED → in candidate list; dismissed → excluded
 #   T13 — missing required env GITEA_TOKEN → exits 1 with error
 #   T14 — non-default-base PR exits 0 without requiring review
-#   T18 — wrong-team review candidate does not block right-team comment approval
 #
 # Hostile-self-review (per feedback_assert_exact_not_substring):
 # this test MUST FAIL if the script is absent. Verified by running
@@ -140,7 +138,7 @@ fi
 echo
 echo "== T13 missing GITEA_TOKEN =="
 set +e
-T13_OUT=$(PATH="/tmp:$PATH" GITEA_TOKEN='' GITEA_HOST=git.example.com REPO=x/y PR_NUMBER=1 TEAM=qa TEAM_ID=1 bash "$SCRIPT" 2>&1 || true)
+T13_OUT=$(PATH="/tmp:$PATH" GITEA_TOKEN= GITEA_HOST=git.example.com REPO=x/y PR_NUMBER=1 TEAM=qa TEAM_ID=1 bash "$SCRIPT" 2>&1 || true)
 set -e
 assert_contains "T13 exits non-zero when GITEA_TOKEN missing" "GITEA_TOKEN required" "$T13_OUT"

@@ -205,8 +203,6 @@ chmod +x "$FIXTURE_DIR/bin/curl"
 # Helper: run the script with fixture environment
 run_review_check() {
  local scenario="$1"
-  local team="${2:-qa}"
-  local team_id="${3:-20}"
  echo "$scenario" >"$FIX_STATE_DIR/scenario"
  local out
  set +e
@@ -217,8 +213,8 @@ run_review_check() {
    REPO="molecule-ai/molecule-core" \
    PR_NUMBER="999" \
    DEFAULT_BRANCH="main" \
-    TEAM="$team" \
-    TEAM_ID="$team_id" \
+    TEAM="qa" \
+    TEAM_ID="20" \
    REVIEW_CHECK_DEBUG="0" \
    REVIEW_CHECK_STRICT="0" \
    bash "$SCRIPT" 2>&1
@@ -310,12 +306,12 @@ echo
 echo "== T10 CURL_AUTH_FILE =="
 # Verify the token-file logic directly: create a temp file with the
 # same mktemp pattern, write the header with printf, chmod 600, then assert.
-T10_TOKEN="secret-fixture-token-abc123"
+T10_TOKEN="secret-test-token-abc123"
 T10_AUTHFILE=$(mktemp "${TMPDIR:-/tmp}/curl-auth.test.XXXXXX")
 chmod 600 "$T10_AUTHFILE"
 printf 'header = "Authorization: token %s"\n' "$T10_TOKEN" > "$T10_AUTHFILE"
 assert_file_mode "T10a mktemp authfile mode 600 (CURL_AUTH_FILE pattern)" "$T10_AUTHFILE" "600"
-assert_file_contains "T10b printf header format (CURL_AUTH_FILE content)" "$T10_AUTHFILE" "Authorization: token secret-fixture-token-abc123"
+assert_file_contains "T10b printf header format (CURL_AUTH_FILE content)" "$T10_AUTHFILE" "Authorization: token secret-test-token-abc123"
 assert_file_contains "T10c 'header =' curl-config syntax" "$T10_AUTHFILE" 'header = "Authorization: token '
 rm -f "$T10_AUTHFILE"

@@ -338,61 +334,6 @@ assert_contains "T12 jq: core-devops (non-author APPROVED) in candidates" "core-
 assert_eq "T12 jq: alice (author) NOT in candidates" "" "$(echo "$T12_CANDIDATES" | grep '^alice$' || true)"
 assert_eq "T12 jq: carol (dismissed) NOT in candidates" "" "$(echo "$T12_CANDIDATES" | grep '^carol$' || true)"

-# T15 — comment-based approval via agent prefix pattern → exit 0
-echo
-echo "== T15 comment agent-prefix approval =="
-T15_OUT=$(run_review_check "T15_comments_agent_approval")
-T15_RC=$(cat "$FIX_STATE_DIR/last_rc")
-assert_eq "T15 exit code 0 (agent-comment approval + team member)" "0" "$T15_RC"
-assert_contains "T15 comment fallback notice" "comment-based approval" "$T15_OUT"
-assert_contains "T15 core-qa-agent APPROVED" "APPROVED by core-qa-agent" "$T15_OUT"
-
-# T16 — comment-based approval via generic APPROVED keyword → exit 0
-echo
-echo "== T16 comment generic keyword approval =="
-T16_OUT=$(run_review_check "T16_comments_generic_approval")
-T16_RC=$(cat "$FIX_STATE_DIR/last_rc")
-assert_eq "T16 exit code 0 (generic-approval comment + team member)" "0" "$T16_RC"
-assert_contains "T16 comment fallback notice" "comment-based approval" "$T16_OUT"
-
-# T17 — no approval keywords in comments → exit 1
-echo
-echo "== T17 comments with no approval keywords =="
-T17_OUT=$(run_review_check "T17_comments_no_approval")
-T17_RC=$(cat "$FIX_STATE_DIR/last_rc")
-assert_eq "T17 exit code 1 (no candidates from comments)" "1" "$T17_RC"
-assert_contains "T17 no candidates error" "no candidates from reviews API or issue comments" "$T17_OUT"
-
-# T18 — a wrong-team PR review candidate must not suppress a right-team
-# comment approval. This matches PR #1790, where QA had an APPROVED review
-# and security approved via the agent comment convention.
-echo
-echo "== T18 review candidate wrong team, comment candidate right team =="
-T18_OUT=$(run_review_check "T18_review_wrong_team_comment_right_team")
-T18_RC=$(cat "$FIX_STATE_DIR/last_rc")
-assert_eq "T18 exit code 0 (comment approval still considered)" "0" "$T18_RC"
-assert_contains "T18 comment candidate notice" "comment-based approval" "$T18_OUT"
-assert_contains "T18 comment approver accepted" "APPROVED by core-qa-agent" "$T18_OUT"
-
-# T19 — ai-sop-ack member APPROVED review must NOT count toward qa-review
-# or security-review (R1 hardening refinement, msg 1388c76f).
-echo
-echo "== T19 ai-sop-ack APPROVED review excluded from qa-review gate =="
-T19_OUT=$(run_review_check "T19_ai_sop_ack_approved" "qa" "20")
-T19_RC=$(cat "$FIX_STATE_DIR/last_rc")
-assert_eq "T19 exit code 1 (ai-sop-ack not in qa team)" "1" "$T19_RC"
-assert_contains "T19 ai-reviewer excluded from qa" "candidates: ai-reviewer" "$T19_OUT"
-assert_contains "T19 none are in qa team" "none are in team" "$T19_OUT"
-
-# T20 — same ai-sop-ack member must also be excluded from security-review gate.
-echo
-echo "== T20 ai-sop-ack APPROVED review excluded from security-review gate =="
-T20_OUT=$(run_review_check "T19_ai_sop_ack_approved" "security" "21")
-T20_RC=$(cat "$FIX_STATE_DIR/last_rc")
-assert_eq "T20 exit code 1 (ai-sop-ack not in security team)" "1" "$T20_RC"
-assert_contains "T20 ai-reviewer excluded from security" "candidates: ai-reviewer" "$T20_OUT"
-assert_contains "T20 none are in security team" "none are in team" "$T20_OUT"
-
 echo
 echo "------"
 echo "PASS=$PASS FAIL=$FAIL"
@@ -22,6 +22,7 @@ from __future__ import annotations

 import os
 import sys
+import tempfile
 import unittest

 # Resolve sibling script regardless of where pytest is invoked from.
@@ -550,752 +551,3 @@ class TestEndToEndAckFlow(unittest.TestCase):

 if __name__ == "__main__":
    unittest.main(verbosity=2)
-
-
-# ---------------------------------------------------------------------------
-# compute_na_state
-# ---------------------------------------------------------------------------
-
-
-class TestComputeNaState(unittest.TestCase):
-    """Tests for /sop-n/a directive evaluation."""
-
-    def test_no_na_declarations(self):
-        cfg = sop.load_config(CONFIG_PATH)
-        na_gates = cfg.get("n/a_gates", {})
-        comments = []
-        na_state = sop.compute_na_state(comments, "alice", na_gates, lambda *_: [])
-        self.assertFalse(na_state["qa-review"]["declared"])
-        self.assertFalse(na_state["security-review"]["declared"])
-
-    def test_na_declared_by_authorized_user(self):
-        cfg = sop.load_config(CONFIG_PATH)
-        na_gates = cfg.get("n/a_gates", {})
-        comments = [_comment("bob", "/sop-n/a qa-review N/A: pure tooling change")]
-        na_state = sop.compute_na_state(comments, "alice", na_gates, lambda g, u: u)
-        self.assertTrue(na_state["qa-review"]["declared"])
-        self.assertEqual(na_state["qa-review"]["decl_ackers"], ["bob"])
-
-    def test_na_declared_by_unauthorized_user_rejected(self):
-        cfg = sop.load_config(CONFIG_PATH)
-        na_gates = cfg.get("n/a_gates", {})
-        comments = [_comment("mallory", "/sop-n/a qa-review N/A: not real team")]
-        na_state = sop.compute_na_state(comments, "alice", na_gates, lambda g, u: [])
-        self.assertFalse(na_state["qa-review"]["declared"])
-        self.assertEqual(na_state["qa-review"]["rejected"]["not_in_team"], ["mallory"])
-
-    def test_author_cannot_self_declare_na(self):
-        cfg = sop.load_config(CONFIG_PATH)
-        na_gates = cfg.get("n/a_gates", {})
-        comments = [_comment("alice", "/sop-n/a qa-review N/A: I am the author")]
-        na_state = sop.compute_na_state(comments, "alice", na_gates, lambda g, u: u)
-        self.assertFalse(na_state["qa-review"]["declared"])
-
-    def test_parse_directives_separates_na_from_ack(self):
-        directives, na_directives = sop.parse_directives(
-            "/sop-ack comprehensive-testing\n/sop-n/a qa-review N/A: no surface",
-            {},
-        )
-        self.assertEqual(len(directives), 1)
-        self.assertEqual(directives[0][0], "sop-ack")
-        self.assertEqual(len(na_directives), 1)
-        self.assertEqual(na_directives[0][0], "sop-n/a")
-        self.assertEqual(na_directives[0][1], "qa-review")
-
-
-# ---------------------------------------------------------------------------
-# RFC#450 Option C — risk-classed two-eyes (governance fix for internal#442)
-# ---------------------------------------------------------------------------
-
-
-class TestIsHighRisk(unittest.TestCase):
-    """The high-risk predicate decides which required_teams list applies.
-
-    Predicate: tier:high label OR any label in cfg.high_risk_labels.
-    """
-
-    def setUp(self):
-        self.cfg = sop.load_config(CONFIG_PATH)
-
-    def test_no_labels_is_default_class(self):
-        pr = {"labels": []}
-        self.assertFalse(sop.is_high_risk(pr, self.cfg))
-
-    def test_tier_high_is_high_risk(self):
-        pr = {"labels": [{"name": "tier:high"}]}
-        self.assertTrue(sop.is_high_risk(pr, self.cfg))
-
-    def test_tier_low_is_default_class(self):
-        pr = {"labels": [{"name": "tier:low"}]}
-        self.assertFalse(sop.is_high_risk(pr, self.cfg))
-
-    def test_tier_medium_is_default_class(self):
-        # tier:medium alone is NOT high-risk (Option C — medium routes
-        # to the wider engineers OR-set).
-        pr = {"labels": [{"name": "tier:medium"}]}
-        self.assertFalse(sop.is_high_risk(pr, self.cfg))
-
-    def test_area_security_label_is_high_risk(self):
-        pr = {"labels": [{"name": "tier:medium"}, {"name": "area:security"}]}
-        self.assertTrue(sop.is_high_risk(pr, self.cfg))
-
-    def test_area_schema_label_is_high_risk(self):
-        pr = {"labels": [{"name": "area:schema"}]}
-        self.assertTrue(sop.is_high_risk(pr, self.cfg))
-
-    def test_area_identity_label_is_high_risk(self):
-        pr = {"labels": [{"name": "area:identity"}]}
-        self.assertTrue(sop.is_high_risk(pr, self.cfg))
-
-    def test_area_fleet_image_label_is_high_risk(self):
-        pr = {"labels": [{"name": "area:fleet-image"}]}
-        self.assertTrue(sop.is_high_risk(pr, self.cfg))
-
-    def test_area_gate_meta_label_is_high_risk(self):
-        # Gate-meta = changes to sop-checklist/sop-tier-check itself.
-        pr = {"labels": [{"name": "area:gate-meta"}]}
-        self.assertTrue(sop.is_high_risk(pr, self.cfg))
-
-    def test_unknown_area_label_is_default_class(self):
-        pr = {"labels": [{"name": "area:docs"}]}
-        self.assertFalse(sop.is_high_risk(pr, self.cfg))
-
-
-class TestResolveRequiredTeams(unittest.TestCase):
-    """The team resolver picks the elevated list only for high-risk PRs
-    AND only when the item declares one — items without an elevated
-    list always use the default required_teams."""
-
-    def test_default_class_uses_default_teams(self):
-        item = {"required_teams": ["engineers", "managers", "ceo"], "required_teams_high_risk": ["ceo"]}
-        self.assertEqual(
-            sop.resolve_required_teams(item, high_risk=False),
-            ["engineers", "managers", "ceo"],
-        )
-
-    def test_high_risk_uses_elevated_teams(self):
-        item = {"required_teams": ["engineers", "managers", "ceo"], "required_teams_high_risk": ["ceo"]}
-        self.assertEqual(
-            sop.resolve_required_teams(item, high_risk=True),
-            ["ceo"],
-        )
-
-    def test_high_risk_without_elevated_falls_back_to_default(self):
-        # Items that don't declare required_teams_high_risk (e.g.
-        # comprehensive-testing, staging-smoke) are unaffected by risk-class.
-        item = {"required_teams": ["engineers"]}
-        self.assertEqual(
-            sop.resolve_required_teams(item, high_risk=True),
-            ["engineers"],
-        )
-
-    def test_empty_elevated_list_falls_back_to_default(self):
-        # A defensive case: required_teams_high_risk: [] should not
-        # silently lock out all approvers — fall back to the default
-        # so the gate stays satisfiable. (Tightening should remove the
-        # key, not set it to empty.)
-        item = {"required_teams": ["engineers"], "required_teams_high_risk": []}
-        self.assertEqual(
-            sop.resolve_required_teams(item, high_risk=True),
-            ["engineers"],
-        )
-
-
-class TestRootCauseAckEligibilityWidened(unittest.TestCase):
-    """Closes internal#442: a non-author engineers-team ack now satisfies
-    root-cause / no-backwards-compat for the default class.
-
-    The dead-managers/ceo-persona-token gridlock is the symptom; the
-    root cause is that sop-checklist ignored tier-class. These tests
-    pin the new wider-default behavior so it can't regress silently.
-    """
-
-    def setUp(self):
-        self.items = _items_by_slug()
-        self.aliases = _numeric_aliases()
-
-    @staticmethod
-    def _approve_only(allowed):
-        return lambda slug, users: [u for u in users if u in allowed]
-
-    def test_engineers_ack_satisfies_root_cause_default_class(self):
-        # Bob is in engineers only (not managers, not ceo). Default class.
-        comments = [_comment("bob", "/sop-ack root-cause")]
-        # Probe: bob is approved because root-cause now lists engineers.
-        probe = self._approve_only({"bob"})
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe, high_risk=False
-        )
-        self.assertEqual(state["root-cause"]["ackers"], ["bob"])
-
-    def test_engineers_ack_satisfies_no_backwards_compat_default_class(self):
-        comments = [_comment("bob", "/sop-ack no-backwards-compat")]
-        probe = self._approve_only({"bob"})
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe, high_risk=False
-        )
-        self.assertEqual(state["no-backwards-compat"]["ackers"], ["bob"])
-
-    def test_engineers_ack_alone_fails_root_cause_when_high_risk(self):
-        # High-risk PR: only ceo can ack. Engineers-only ack must fail.
-        comments = [_comment("bob", "/sop-ack root-cause")]
-        # Probe: bob is in engineers, not ceo. Under high_risk,
-        # required_teams_high_risk=[ceo] → bob is NOT approved.
-        # Probe receives the items + flag indirectly via main(); for
-        # the unit-test path we inject a probe that rejects bob.
-        probe = self._approve_only(set())  # nobody is in ceo
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe, high_risk=True
-        )
-        self.assertEqual(state["root-cause"]["ackers"], [])
-        self.assertIn("bob", state["root-cause"]["rejected"]["not_in_team"])
-
-    def test_ceo_ack_satisfies_root_cause_when_high_risk(self):
-        # High-risk PR + ceo-team approver → passes (the senior path).
-        comments = [_comment("hongming", "/sop-ack root-cause")]
-        probe = self._approve_only({"hongming"})
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe, high_risk=True
-        )
-        self.assertEqual(state["root-cause"]["ackers"], ["hongming"])
-
-    def test_self_ack_still_forbidden_even_with_widened_eligibility(self):
-        # Author cannot self-ack — widening teams must NOT weaken
-        # the non-author rule.
-        comments = [_comment("alice", "/sop-ack root-cause")]
-        probe = self._approve_only({"alice"})
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe, high_risk=False
-        )
-        self.assertEqual(state["root-cause"]["ackers"], [])
-        self.assertIn("alice", state["root-cause"]["rejected"]["self_ack"])
-
-
-class TestHighRiskClassUsesElevatedListInConfig(unittest.TestCase):
-    """End-to-end: the shipped config + RFC#450 predicate must keep
-    root-cause / no-backwards-compat gated on ceo for high-risk PRs."""
-
-    def test_root_cause_high_risk_elevated_to_ceo_only(self):
-        items = _items_by_slug()
-        # tier:high alone makes the PR high-risk → root-cause needs ceo.
-        self.assertEqual(
-            sop.resolve_required_teams(items["root-cause"], high_risk=True),
-            ["ceo"],
-        )
-        # Default class accepts engineers/managers/ceo.
-        self.assertEqual(
-            sorted(sop.resolve_required_teams(items["root-cause"], high_risk=False)),
-            sorted(["engineers", "managers", "ceo"]),
-        )
-
-    def test_no_backwards_compat_high_risk_elevated_to_ceo_only(self):
-        items = _items_by_slug()
-        self.assertEqual(
-            sop.resolve_required_teams(items["no-backwards-compat"], high_risk=True),
-            ["ceo"],
-        )
-        self.assertEqual(
-            sorted(sop.resolve_required_teams(items["no-backwards-compat"], high_risk=False)),
-            sorted(["engineers", "managers", "ceo"]),
-        )
-
-    def test_other_items_unchanged_by_risk_class(self):
-        # Items without required_teams_high_risk are unaffected.
-        items = _items_by_slug()
-        for slug in (
-            "comprehensive-testing",
-            "local-postgres-e2e",
-            "staging-smoke",
-            "five-axis-review",
-            "memory-consulted",
-        ):
-            self.assertEqual(
-                sop.resolve_required_teams(items[slug], high_risk=False),
-                sop.resolve_required_teams(items[slug], high_risk=True),
-                f"item {slug} should not be affected by risk-class",
-            )
-
-
-# ---------------------------------------------------------------------------
-# get_issue_comments — streaming + minimal-dict shape (task #369 / OOM fix)
-# ---------------------------------------------------------------------------
-
-
-class _FakeReq:
-    """Stand-in for GiteaClient._req that serves canned pages."""
-
-    def __init__(self, pages):
-        # pages: list[list[dict]]; one page per call, exhausted in order.
-        self._pages = list(pages)
-        self.calls = []
-
-    def __call__(self, method, path, body=None, ok_codes=(200, 201, 204)):
-        self.calls.append((method, path))
-        if not self._pages:
-            return 200, []
-        return 200, self._pages.pop(0)
-
-
-class TestGetIssueCommentsStreaming(unittest.TestCase):
-    """Verify the OOM-fix invariants — minimal-dict shape + page break."""
-
-    def _client_with_pages(self, pages):
-        client = sop.GiteaClient("git.example.com", "tok")
-        client._req = _FakeReq(pages)  # type: ignore[method-assign]
-        return client
-
-    def test_minimal_dict_shape_drops_large_fields(self):
-        """get_issue_comments must DROP html_url/assets/timestamps/etc. and
-        keep ONLY {user.login, body} — that's the whole OOM-prevention."""
-        full_page = [
-            {
-                "id": 1234,
-                "html_url": "https://example.com/some-huge-url",
-                "pull_request_url": "https://example.com/some-other-huge-url",
-                "issue_url": "https://example.com/yet-another-url",
-                "user": {"login": "bob", "avatar_url": "x" * 4000, "id": 99},
-                "original_author": "",
-                "original_author_id": 0,
-                "body": "/sop-ack comprehensive-testing\n\nlooks good",
-                "assets": ["x" * 1000, "y" * 1000],
-                "created_at": "2026-05-19T01:02:03Z",
-                "updated_at": "2026-05-19T01:02:03Z",
-            }
-        ]
-        client = self._client_with_pages([full_page])
-        out = client.get_issue_comments("o", "r", 1)
-        self.assertEqual(len(out), 1)
-        # Only the two whitelisted keys + nested user.login
-        self.assertEqual(set(out[0].keys()), {"user", "body"})
-        self.assertEqual(set(out[0]["user"].keys()), {"login"})
-        self.assertEqual(out[0]["user"]["login"], "bob")
-        self.assertEqual(out[0]["body"], "/sop-ack comprehensive-testing\n\nlooks good")
-        # Critical: avatar/assets/timestamps/etc. must be gone (~4KB+ each).
-        self.assertNotIn("html_url", out[0])
-        self.assertNotIn("assets", out[0])
-        self.assertNotIn("created_at", out[0])
-
-    def test_pagination_break_on_short_page(self):
-        # Page-size 50; a page of <50 means no more pages.
-        page1 = [{"user": {"login": "u"}, "body": "x"}] * 7
-        client = self._client_with_pages([page1])
-        out = client.get_issue_comments("o", "r", 2)
-        self.assertEqual(len(out), 7)
-        # Should have made exactly 1 _req call (no page-2 probe).
-        self.assertEqual(len(client._req.calls), 1)
-
-    def test_pagination_continues_until_empty(self):
-        # Two full pages + one short page.
-        page1 = [{"user": {"login": "u"}, "body": "x"}] * 50
-        page2 = [{"user": {"login": "u"}, "body": "y"}] * 50
-        page3 = [{"user": {"login": "u"}, "body": "z"}] * 3
-        client = self._client_with_pages([page1, page2, page3])
-        out = client.get_issue_comments("o", "r", 3)
-        self.assertEqual(len(out), 103)
-        self.assertEqual(len(client._req.calls), 3)
-
-    def test_max_comments_caps_collection(self):
-        page1 = [{"user": {"login": "u"}, "body": "x"}] * 50
-        page2 = [{"user": {"login": "u"}, "body": "y"}] * 50
-        page3 = [{"user": {"login": "u"}, "body": "z"}] * 50
-        client = self._client_with_pages([page1, page2, page3])
-        out = client.get_issue_comments("o", "r", 4, max_comments=75)
-        self.assertEqual(len(out), 75)
-        # Stops short: shouldn't have requested page-3.
-        self.assertLessEqual(len(client._req.calls), 2)
-
-    def test_oversized_body_truncated(self):
-        # An individual comment with a multi-MiB body (e.g. pasted CI log)
-        # must NOT pull the whole thing into memory. The directive parser
-        # only needs the first ~8 KiB to find /sop-* markers.
-        huge_body = "/sop-ack comprehensive-testing\n" + ("X" * (4 * 1024 * 1024))
-        page = [{"user": {"login": "bob"}, "body": huge_body}]
-        client = self._client_with_pages([page])
-        out = client.get_issue_comments("o", "r", 99)
-        self.assertEqual(len(out), 1)
-        # Cap is 8 KiB; comment body must be <= 8 KiB after streaming.
-        self.assertLessEqual(len(out[0]["body"]), 8 * 1024)
-        # Marker still discoverable at the start.
-        self.assertTrue(out[0]["body"].startswith("/sop-ack comprehensive-testing"))
-
-    def test_iter_handles_missing_user_or_body(self):
-        # Defensive: Gitea has been seen to return user=null on deleted users.
-        page = [
-            {"user": None, "body": "abandoned-author"},
-            {"user": {"login": "alice"}, "body": None},
-            {"body": "no-user-key"},
-            {"user": {"login": "bob"}, "body": "ok"},
-        ]
-        client = self._client_with_pages([page])
-        out = client.get_issue_comments("o", "r", 5)
-        self.assertEqual(len(out), 4)
-        self.assertEqual(out[0]["user"]["login"], "")
-        self.assertEqual(out[0]["body"], "abandoned-author")
-        self.assertEqual(out[1]["user"]["login"], "alice")
-        self.assertEqual(out[1]["body"], "")
-        self.assertEqual(out[2]["user"]["login"], "")
-        self.assertEqual(out[3]["user"]["login"], "bob")
-
-    def test_minimal_dicts_work_with_compute_ack_state(self):
-        """Round-trip: minimal dicts feed back through compute_ack_state."""
-        page = [{"user": {"login": "bob"}, "body": "/sop-ack comprehensive-testing"}]
-        client = self._client_with_pages([page])
-        comments = client.get_issue_comments("o", "r", 6)
-        items = _items_by_slug()
-        aliases = _numeric_aliases()
-        state = sop.compute_ack_state(
-            comments, "alice", items, aliases, lambda slug, users: list(users)
-        )
-        self.assertEqual(state["comprehensive-testing"]["ackers"], ["bob"])
-
-
-# ---------------------------------------------------------------------------
-# probe() na-gate fallback — fix for #355-class KeyError 'security-review'
-# ---------------------------------------------------------------------------
-
-
-class TestComputeNaStateAcceptsGateNotInItems(unittest.TestCase):
-    """compute_na_state passes the gate NAME to probe(); when the gate is
-    NOT also an items entry (the common case for `security-review`,
-    `qa-review`), probe must fall back to the gate's own required_teams
-    instead of KeyError'ing on items_by_slug[slug].
-
-    This test exercises the public surface (compute_na_state) rather than
-    the inline `probe` closure, because the closure is built inside main().
-    We simulate the fallback by passing a probe that mirrors the production
-    contract — slug may be either an item OR an n/a-gate key, both are valid.
-    """
-
-    def test_na_gate_with_required_teams_resolves_without_keyerror(self):
-        na_gates = {
-            "security-review": {
-                "required_teams": ["security", "managers", "ceo"],
-                "description": "security N/A",
-            },
-        }
-        comments = [
-            {"user": {"login": "carol"}, "body": "/sop-n/a security-review docs-only"},
-        ]
-        # Probe approves any user in the security team; importantly it does
-        # NOT try items_by_slug[slug] for the gate name.
-        called_with = []
-
-        def probe(slug, users):
-            called_with.append(slug)
-            # production probe accepts gate-name OR item-slug; for this test
-            # we just approve everyone.
-            return list(users)
-
-        na_state = sop.compute_na_state(comments, "alice", na_gates, probe)
-        self.assertTrue(na_state["security-review"]["declared"])
-        self.assertEqual(na_state["security-review"]["decl_ackers"], ["carol"])
-        # probe must have been called with the GATE name, not an item slug.
-        self.assertEqual(called_with, ["security-review"])
-
-    def test_na_gate_self_declaration_rejected(self):
-        # Author cannot self-declare N/A — pre-existing invariant; pin it
-        # so the new probe-fallback doesn't regress this.
-        na_gates = {"security-review": {"required_teams": ["security"]}}
-        comments = [
-            {"user": {"login": "alice"}, "body": "/sop-n/a security-review"},
-        ]
-        na_state = sop.compute_na_state(
-            comments, "alice", na_gates, lambda *_: ["alice"]
-        )
-        self.assertFalse(na_state["security-review"]["declared"])
-
-
-# ---------------------------------------------------------------------------
-# internal#760 ceremony — ai-sop-ack team + ai_ack_eligible per-item flag
-# ---------------------------------------------------------------------------
-
-
-class TestAIAckEligibleConfig(unittest.TestCase):
-    """CTO-controlled allowlist (msg 1388c76f):
-      ai_ack_eligible: comprehensive-testing, local-postgres-e2e, staging-smoke,
-                       five-axis-review, memory-consulted
-      human-only:      root-cause, no-backwards-compat
-    """
-
-    def test_ai_ack_eligible_items(self):
-        cfg = sop.load_config(CONFIG_PATH)
-        items_by_slug = {it["slug"]: it for it in cfg["items"]}
-        eligible = {
-            "comprehensive-testing",
-            "local-postgres-e2e",
-            "staging-smoke",
-            "five-axis-review",
-            "memory-consulted",
-        }
-        for slug in eligible:
-            self.assertTrue(
-                items_by_slug[slug].get("ai_ack_eligible"),
-                f"{slug} must be ai_ack_eligible",
-            )
-
-    def test_human_only_items(self):
-        cfg = sop.load_config(CONFIG_PATH)
-        items_by_slug = {it["slug"]: it for it in cfg["items"]}
-        human_only = {"root-cause", "no-backwards-compat"}
-        for slug in human_only:
-            self.assertFalse(
-                items_by_slug[slug].get("ai_ack_eligible", False),
-                f"{slug} must NOT be ai_ack_eligible (human-only)",
-            )
-
-    def test_testing_class_slugs_constant(self):
-        """_TESTING_CLASS_SLUGS must match the three testing items."""
-        self.assertEqual(
-            sop._TESTING_CLASS_SLUGS,
-            {"comprehensive-testing", "local-postgres-e2e", "staging-smoke"},
-        )
-
-    def test_human_only_slugs_constant(self):
-        """_HUMAN_ONLY_SLUGS encodes the migration/schema carve-out.
-
-        If this set changes, the CTO must approve the widening.
-        """
-        self.assertEqual(
-            sop._HUMAN_ONLY_SLUGS,
-            {"root-cause", "no-backwards-compat", "migration", "schema"},
-        )
-
-    def test_human_only_invariant_enforced_in_code_and_config(self):
-        """Every config-present slug in _HUMAN_ONLY_SLUGS must be human-only.
-
-        This test fails if a migration/schema-class item accidentally
-        acquires ai_ack_eligible via config drift.  migration/schema are
-        future-proofing slugs not yet in the live config; they are checked
-        by the production probe closure but skipped here.
-        """
-        cfg = sop.load_config(CONFIG_PATH)
-        items_by_slug = {it["slug"]: it for it in cfg["items"]}
-        for slug in sop._HUMAN_ONLY_SLUGS:
-            if slug not in items_by_slug:
-                # Future-proofing slug (e.g. migration, schema) — not yet
-                # in config, but the code guard still rejects AI acks.
-                continue
-            self.assertFalse(
-                items_by_slug[slug].get("ai_ack_eligible", False),
-                f"{slug} is in _HUMAN_ONLY_SLUGS and must NEVER be ai_ack_eligible",
-            )
-
-
-class TestAIAckEligibilityProbe(unittest.TestCase):
-    """The probe closure in main() delegates to compute_ack_state.
-    We simulate the AI-ack path by injecting a probe that behaves like
-    the production probe (human team first, then ai-sop-ack fallback).
-    """
-
-    def setUp(self):
-        self.items = _items_by_slug()
-        self.aliases = _numeric_aliases()
-
-    def _probe_human_then_ai(self, human_users, ai_users):
-        """Return users in human_users immediately; users in ai_users only
-        if the item is ai_ack_eligible."""
-        def probe(slug, users):
-            item = self.items.get(slug, {})
-            approved = []
-            for u in users:
-                if u in human_users:
-                    approved.append(u)
-                elif u in ai_users and item.get("ai_ack_eligible"):
-                    approved.append(u)
-            return approved
-        return probe
-
-    def test_ai_ack_passes_for_eligible_item(self):
-        comments = [_comment("ai-bot", "/sop-ack five-axis-review")]
-        probe = self._probe_human_then_ai(human_users=set(), ai_users={"ai-bot"})
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe
-        )
-        self.assertEqual(state["five-axis-review"]["ackers"], ["ai-bot"])
-
-    def test_ai_ack_rejected_for_human_only_item(self):
-        comments = [_comment("ai-bot", "/sop-ack root-cause")]
-        probe = self._probe_human_then_ai(human_users=set(), ai_users={"ai-bot"})
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe
-        )
-        self.assertEqual(state["root-cause"]["ackers"], [])
-        self.assertIn("ai-bot", state["root-cause"]["rejected"]["not_in_team"])
-
-    def test_human_ack_still_works_for_ai_eligible_item(self):
-        comments = [_comment("bob", "/sop-ack comprehensive-testing")]
-        probe = self._probe_human_then_ai(human_users={"bob"}, ai_users=set())
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe
-        )
-        self.assertEqual(state["comprehensive-testing"]["ackers"], ["bob"])
-
-    def test_ai_ack_rejected_for_testing_item_when_ci_red(self):
-        # Simulate the production probe that checks CI status for testing items.
-        # When CI is not green, ai-sop-ack member is rejected.
-        def probe(slug, users):
-            item = self.items.get(slug, {})
-            approved = []
-            for u in users:
-                if u == "ai-bot" and item.get("ai_ack_eligible"):
-                    # Testing items require CI green; simulate CI red.
-                    if slug in sop._TESTING_CLASS_SLUGS:
-                        continue  # rejected: CI not green
-                    approved.append(u)
-            return approved
-
-        comments = [_comment("ai-bot", "/sop-ack comprehensive-testing")]
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe
-        )
-        self.assertEqual(state["comprehensive-testing"]["ackers"], [])
-
-    def test_ai_ack_passes_for_testing_item_when_ci_green(self):
-        # Simulate CI green → AI ack passes.
-        def probe(slug, users):
-            item = self.items.get(slug, {})
-            approved = []
-            for u in users:
-                if u == "ai-bot" and item.get("ai_ack_eligible"):
-                    if slug in sop._TESTING_CLASS_SLUGS:
-                        # CI is green → allow
-                        pass
-                    approved.append(u)
-            return approved
-
-        comments = [_comment("ai-bot", "/sop-ack comprehensive-testing")]
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe
-        )
-        self.assertEqual(state["comprehensive-testing"]["ackers"], ["ai-bot"])
-
-
-class TestAIAckHumanOnlyMigrationSchema(unittest.TestCase):
-    """RC 8322: migration and schema items are human-only regardless of
-    any future config that might accidentally mark them ai_ack_eligible.
-
-    These slugs are not yet in the live config items list; the tests use
-    synthetic items so the production guard can be exercised directly.
-    """
-
-    def setUp(self):
-        # Synthetic items — if live config ever adds migration/schema,
-        # they MUST stay human-only. The probe below mirrors the actual
-        # production closure logic (human team first, then AI fallback
-        # with _HUMAN_ONLY_SLUGS guard).
-        self.items = {
-            "migration": {
-                "slug": "migration",
-                "ai_ack_eligible": True,
-                "required_teams": ["engineers"],
-            },
-            "schema": {
-                "slug": "schema",
-                "ai_ack_eligible": True,
-                "required_teams": ["engineers"],
-            },
-        }
-        self.aliases = {}
-
-    def _production_like_probe(self, human_users, ai_users):
-        """Return a probe that mirrors the production closure's guard."""
-
-        def probe(slug, users):
-            item = self.items.get(slug, {})
-            approved = []
-            for u in users:
-                if u in human_users:
-                    approved.append(u)
-                elif u in ai_users:
-                    # Production guard: _HUMAN_ONLY_SLUGS rejects AI acks
-                    # regardless of the ai_ack_eligible flag.
-                    if slug in sop._HUMAN_ONLY_SLUGS:
-                        continue
-                    if item.get("ai_ack_eligible"):
-                        approved.append(u)
-            return approved
-
-        return probe
-
-    def test_ai_ack_rejected_for_migration(self):
-        comments = [_comment("ai-bot", "/sop-ack migration")]
-        probe = self._production_like_probe(human_users=set(), ai_users={"ai-bot"})
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe
-        )
-        self.assertEqual(state["migration"]["ackers"], [])
-        self.assertIn("ai-bot", state["migration"]["rejected"]["not_in_team"])
-
-    def test_ai_ack_rejected_for_schema(self):
-        comments = [_comment("ai-bot", "/sop-ack schema")]
-        probe = self._production_like_probe(human_users=set(), ai_users={"ai-bot"})
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe
-        )
-        self.assertEqual(state["schema"]["ackers"], [])
-        self.assertIn("ai-bot", state["schema"]["rejected"]["not_in_team"])
-
-    def test_human_ack_still_works_for_migration(self):
-        # Human team member acking migration/schema is unaffected.
-        comments = [_comment("bob", "/sop-ack migration")]
-        probe = self._production_like_probe(human_users={"bob"}, ai_users=set())
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe
-        )
-        self.assertEqual(state["migration"]["ackers"], ["bob"])
-
-    def test_human_ack_still_works_for_schema(self):
-        comments = [_comment("bob", "/sop-ack schema")]
-        probe = self._production_like_probe(human_users={"bob"}, ai_users=set())
-        state = sop.compute_ack_state(
-            comments, "alice", self.items, self.aliases, probe
-        )
-        self.assertEqual(state["schema"]["ackers"], ["bob"])
-
-
-class TestGetCIStatus(unittest.TestCase):
-    """Verify get_ci_status reads the correct context from commit statuses."""
-
-    def _client_with_statuses(self, statuses):
-        client = sop.GiteaClient("git.example.com", "tok")
-
-        def fake_req(method, path, body=None, ok_codes=(200, 201, 204)):
-            return 200, statuses
-
-        client._req = fake_req  # type: ignore[method-assign]
-        return client
-
-    def test_ci_green_returns_success(self):
-        client = self._client_with_statuses([
-            {"context": "CI / all-required (pull_request)", "state": "success"},
-        ])
-        self.assertEqual(
-            sop.get_ci_status(client, "o", "r", "sha1"), "success"
-        )
-
-    def test_ci_red_returns_failure(self):
-        client = self._client_with_statuses([
-            {"context": "CI / all-required (pull_request)", "state": "failure"},
-        ])
-        self.assertEqual(
-            sop.get_ci_status(client, "o", "r", "sha1"), "failure"
-        )
-
-    def test_missing_context_returns_missing(self):
-        client = self._client_with_statuses([
-            {"context": "some-other-context", "state": "success"},
-        ])
-        self.assertEqual(
-            sop.get_ci_status(client, "o", "r", "sha1"), "missing"
-        )
-
-    def test_api_error_returns_unknown(self):
-        client = sop.GiteaClient("git.example.com", "tok")
-
-        def fake_req(method, path, body=None, ok_codes=(200, 201, 204)):
-            return 500, {"error": "boom"}
-
-        client._req = fake_req  # type: ignore[method-assign]
-        self.assertEqual(
-            sop.get_ci_status(client, "o", "r", "sha1"), "unknown"
-        )
@@ -6,10 +6,9 @@
 #   T1: PR open + APPROVED via tier:low → script invokes sop-tier-check
 #       and POSTs status=success.
 #   T2: PR open + missing tier label → sop-tier-check exits non-zero;
-#       refire still POSTs status=success, matching the canonical
-#       pull_request_target workflow's fail-open job conclusion.
+#       refire POSTs status=failure (description mentions failure).
 #   T3: PR open + tier:low but NO approving reviews → sop-tier-check
-#       exits non-zero; refire still POSTs status=success for the same reason.
+#       exits non-zero; refire POSTs status=failure.
 #   T4: PR CLOSED → refire exits 0 with no status POST (no-op on closed).
 #   T5: Rate-limit — recent status update within 30s → refire skips,
 #       no new POST.
@@ -33,7 +32,7 @@ THIS_DIR="$(cd "$(dirname "$0")" && pwd)"
 SCRIPT_DIR="$(cd "$THIS_DIR/.." && pwd)"
 WORKFLOW_DIR="$(cd "$THIS_DIR/../../workflows" && pwd)"
 WORKFLOW="$WORKFLOW_DIR/sop-tier-refire.yml"
-DISPATCH_WORKFLOW="$WORKFLOW_DIR/sop-checklist.yml"
+DISPATCH_WORKFLOW="$WORKFLOW_DIR/review-refire-comments.yml"
 SCRIPT="$SCRIPT_DIR/sop-tier-refire.sh"

 PASS=0
@@ -89,7 +88,7 @@ assert_file_exists() {
 echo
 echo "== existence =="
 assert_file_exists "workflow file exists"  "$WORKFLOW"
-assert_file_exists "SSOT dispatcher workflow file exists" "$DISPATCH_WORKFLOW"
+assert_file_exists "dispatcher workflow file exists" "$DISPATCH_WORKFLOW"
 assert_file_exists "script file exists"    "$SCRIPT"
 if [ "$FAIL" -gt 0 ]; then
  echo
@@ -134,15 +133,15 @@ else
 fi

 DISPATCH_PARSE_OUT=$(python3 -c 'import sys,yaml;yaml.safe_load(open(sys.argv[1]).read());print("ok")' "$DISPATCH_WORKFLOW" 2>&1 || true)
-assert_eq "T6e SSOT dispatcher workflow parses as YAML" "ok" "$DISPATCH_PARSE_OUT"
+assert_eq "T6e dispatcher workflow parses as YAML" "ok" "$DISPATCH_PARSE_OUT"
 DISPATCH_CONTENT=$(cat "$DISPATCH_WORKFLOW")
-assert_contains "T6f SSOT dispatcher listens on issue_comment" \
+assert_contains "T6f dispatcher listens on issue_comment" \
  "issue_comment" "$DISPATCH_CONTENT"
-assert_contains "T6g SSOT dispatcher handles /qa-recheck" \
+assert_contains "T6g dispatcher handles /qa-recheck" \
  "/qa-recheck" "$DISPATCH_CONTENT"
-assert_contains "T6h SSOT dispatcher handles /security-recheck" \
+assert_contains "T6h dispatcher handles /security-recheck" \
  "/security-recheck" "$DISPATCH_CONTENT"
-assert_contains "T6i SSOT dispatcher handles /refire-tier-check" \
+assert_contains "T6i dispatcher handles /refire-tier-check" \
  "/refire-tier-check" "$DISPATCH_CONTENT"

 # T1-T5 — script behavior against a local Gitea-fixture
@@ -246,21 +245,34 @@ assert_contains "T1 POST context is sop-tier-check / tier-check" \
  '"context": "sop-tier-check / tier-check (pull_request)"' "$POSTED"
 assert_contains "T1 description names commenter" "test-runner" "$POSTED"

-# T2: missing tier label → tier-check fails internally, but refire status
-# matches the canonical workflow's fail-open job conclusion.
+# T2: missing tier label → tier-check fails → failure status POSTed
 run_scenario "T2_no_tier_label" "fail_no_label"
 RC=$(cat "$FIX_STATE_DIR/last_rc")
 POSTED=$(cat "$FIX_STATE_DIR/posted_statuses.jsonl" 2>/dev/null || true)
-assert_eq "T2 exit code 0 (canonical fail-open)" "0" "$RC"
-assert_contains "T2 POSTed state=success" '"state": "success"' "$POSTED"
+# tier-check.sh exits 1; refire script forwards that exit, so RC != 0
+if [ "$RC" -ne 0 ]; then
+  echo "  PASS  T2 exit code non-zero (got $RC)"
+  PASS=$((PASS + 1))
+else
+  echo "  FAIL  T2 exit code should be non-zero, got 0"
+  FAIL=$((FAIL + 1))
+  FAILED_TESTS="${FAILED_TESTS} T2_rc"
+fi
+assert_contains "T2 POSTed state=failure" '"state": "failure"' "$POSTED"

-# T3: tier:low present but ZERO approving reviews → internal tier check fails,
-# refire status remains aligned with the canonical workflow.
+# T3: tier:low present but ZERO approving reviews → failure
 run_scenario "T3_no_approvals" "fail_no_approvals"
 RC=$(cat "$FIX_STATE_DIR/last_rc")
 POSTED=$(cat "$FIX_STATE_DIR/posted_statuses.jsonl" 2>/dev/null || true)
-assert_eq "T3 exit code 0 (canonical fail-open)" "0" "$RC"
-assert_contains "T3 POSTed state=success" '"state": "success"' "$POSTED"
+if [ "$RC" -ne 0 ]; then
+  echo "  PASS  T3 exit code non-zero (got $RC)"
+  PASS=$((PASS + 1))
+else
+  echo "  FAIL  T3 exit code should be non-zero, got 0"
+  FAIL=$((FAIL + 1))
+  FAILED_TESTS="${FAILED_TESTS} T3_rc"
+fi
+assert_contains "T3 POSTed state=failure" '"state": "failure"' "$POSTED"

 # T4: closed PR — refire is a no-op (no POST, exit 0)
 run_scenario "T4_closed" "pass"
@@ -14,7 +14,7 @@ def load_reaper():
    assert spec.loader is not None
    spec.loader.exec_module(mod)
    mod.API = "https://git.example.test/api/v1"
-    mod.GITEA_TOKEN = "fixture-token"
+    mod.GITEA_TOKEN = "test-token"
    mod.API_TIMEOUT_SEC = 1
    mod.API_RETRIES = 3
    mod.API_RETRY_SLEEP_SEC = 0
@@ -32,26 +32,6 @@
 # AUTHOR SELF-ACK IS FORBIDDEN regardless of which team contains them
 # — the gate script enforces commenter != PR author before checking
 # team membership.
-#
-# AI-SOP-ACK TEAM (internal#760 ceremony design, CTO-approved):
-#   The `ai-sop-ack` team contains AI agent identities that can ack
-#   SOP-checklist items ON BEHALF OF automated evidence.  An AI ack is
-#   only valid when:
-#     1. the item has `ai_ack_eligible: true`
-#     2. the item is NOT in the human-only carve-out (migration/schema)
-#     3. for testing-class items, CI / all-required (pull_request) is
-#        green on the current head SHA
-#
-#   AI acks NEVER count toward qa-review or security-review gates —
-#   those remain human-team-only (enforced by review-check.sh team
-#   probe against TEAM_ID 20/21).
-#
-#   INITIAL ai_ack_eligible allowlist (CTO-controlled, msg 1388c76f):
-#     comprehensive-testing, local-postgres-e2e, staging-smoke,
-#     five-axis-review, memory-consulted
-#   HUMAN-ONLY carve-out:
-#     root-cause, no-backwards-compat
-#   Any widening requires an explicit config change reviewed by CTO.

 version: 1

@@ -70,83 +50,44 @@ tier_failure_mode:
  "tier:low": soft
 default_mode: hard  # used when no tier:* label is present

-# High-risk class (RFC#450 Option C, governance-fix for internal#442).
-#
-# A PR is "high-risk" when ANY of the listed labels are applied OR when
-# the PR has `tier:high` (mechanically the strictest existing tier).
-# High-risk items use `required_teams_high_risk` (when present on the
-# item); non-high-risk items use the default `required_teams`.
-#
-# This closes the inconsistency that the SOP charter already mandates
-# `tier:high → ceo only` for the sibling `sop-tier-check` gate; the
-# sop-checklist's `root-cause` and `no-backwards-compat` items now
-# follow the same risk-classed two-eyes shape:
-#   - Default class (tier:low/medium, not high-risk): a non-author
-#     engineers/managers/ceo ack satisfies the item — 25+ live
-#     identities, no dependency on a dead/inactive senior persona
-#     token.
-#   - High-risk class (tier:high OR any high_risk_label): still
-#     requires a non-author ceo ack (durable human team).
-#
-# Tightening: add labels to high_risk_labels.
-# Loosening: remove labels.
-high_risk_labels:
-  - "risk:high"
-  - "area:security"
-  - "area:schema"
-  - "area:fleet-image"
-  - "area:identity"
-  - "area:gate-meta"
-
 items:
  - slug: comprehensive-testing
    numeric_alias: 1
    pr_section_marker: "Comprehensive testing performed"
    required_teams: [qa, engineers]
-    ai_ack_eligible: true
    description: >-
      What was tested, how, edge cases covered. Ack from any qa-team
-      member (or engineers fallback while qa is small). AI ack valid
-      only when CI / all-required (pull_request) is green.
+      member (or engineers fallback while qa is small).

  - slug: local-postgres-e2e
    numeric_alias: 2
    pr_section_marker: "Local-postgres E2E run"
    required_teams: [engineers]
-    ai_ack_eligible: true
    description: >-
      Link to local CI artifact, or "N/A: pure-frontend change". Ack
      from any engineer who can verify the local DB test actually ran.
-      AI ack valid only when CI / all-required (pull_request) is green.

  - slug: staging-smoke
    numeric_alias: 3
    pr_section_marker: "Staging-smoke verified or pending"
    required_teams: [engineers]
-    ai_ack_eligible: true
    description: >-
      Link to canary run, or "scheduled post-merge". Ack from any
      engineer (core-devops/infra-sre are members of engineers team).
-      AI ack valid only when CI / all-required (pull_request) is green.

  - slug: root-cause
    numeric_alias: 4
    pr_section_marker: "Root-cause not symptom"
-    required_teams: [engineers, managers, ceo]
-    required_teams_high_risk: [ceo]
+    required_teams: [managers, ceo]
    description: >-
-      One-sentence root-cause statement. Default class: non-author
-      engineers/managers/ceo ack suffices (engineers can attest
-      root-cause-vs-symptom for routine fixes). High-risk class
-      (see `high_risk_labels`): non-author ceo ack required —
-      senior judgment for irreversible/security/identity/gate
-      changes. Closes internal#442 + tracks RFC#450.
+      One-sentence root-cause statement. Ack from managers tier
+      (team-leads) or ceo. Senior judgment required to attest
+      root-cause-versus-symptom.

  - slug: five-axis-review
    numeric_alias: 5
    pr_section_marker: "Five-Axis review walked"
    required_teams: [engineers]
-    ai_ack_eligible: true
    description: >-
      Correctness / readability / architecture / security / performance.
      Ack from any non-author engineer.
@@ -154,20 +95,15 @@ items:
  - slug: no-backwards-compat
    numeric_alias: 6
    pr_section_marker: "No backwards-compat shim / dead code added"
-    required_teams: [engineers, managers, ceo]
-    required_teams_high_risk: [ceo]
+    required_teams: [managers, ceo]
    description: >-
-      Yes/no + justification if no. Default class: non-author
-      engineers/managers/ceo ack suffices. High-risk class
-      (see `high_risk_labels`): non-author ceo ack required —
-      senior judgment for shim-versus-real-fix on irreversible
-      surfaces. Closes internal#442 + tracks RFC#450.
+      Yes/no + justification if no. Senior ack required because
+      backward-compat shims are how dead-code accretes.

  - slug: memory-consulted
    numeric_alias: 7
    pr_section_marker: "Memory/saved-feedback consulted"
    required_teams: [engineers]
-    ai_ack_eligible: true
    description: >-
      List of feedback memories applicable to this change. Ack from
      any engineer who has the same memory access.
@@ -47,25 +47,12 @@ jobs:
          REPO: ${{ github.repository }}
          PR_NUMBER: ${{ github.event.pull_request.number }}
          # Required-status-check contexts to evaluate at merge time.
-          # Branch-aware JSON dict: keys are protected branch names,
-          # values are arrays of context names that branch protection
-          # requires for that branch. Mirror this against branch
-          # protection (settings → branches → protected branch →
-          # required checks) for each branch listed here.
-          #
+          # Newline-separated. Mirror this against branch protection
+          # (settings → branches → protected branch → required checks).
          # Declared here rather than fetched from /branch_protections
          # because that endpoint requires admin write — sop-tier-bot is
          # read-only by design (least-privilege).
-          REQUIRED_CHECKS_JSON: |
-            {
-              "main": [
-                "CI / all-required (pull_request)",
-                "E2E API Smoke Test / E2E API Smoke Test (pull_request)",
-                "Handlers Postgres Integration / Handlers Postgres Integration (pull_request)"
-              ],
-              "staging": [
-                "CI / all-required (pull_request)",
-                "sop-checklist / all-items-acked (pull_request)"
-              ]
-            }
+          REQUIRED_CHECKS: |
+            CI / all-required (pull_request)
+            sop-checklist / all-items-acked (pull_request)
        run: bash .gitea/scripts/audit-force-merge.sh
@@ -37,7 +37,7 @@ jobs:
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking
    # the PR. Follow-up PR flips this off after surfaced defects are
    # triaged.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
@@ -0,0 +1,60 @@
+name: cascade-list-drift-gate
+
+# Ported from .github/workflows/cascade-list-drift-gate.yml on 2026-05-11
+# per RFC internal#219 §1 sweep.
+#
+# Differences from the GitHub version:
+#   - on.paths reference .gitea/workflows/publish-runtime.yml (the active
+#     Gitea workflow file) instead of .github/workflows/publish-runtime.yml
+#     (which Category A of this sweep deletes).
+#   - Explicit `WORKFLOW=` arg passed to the drift script so it audits the
+#     .gitea/ workflow (the script's default is still .github/... which
+#     will not exist post-Cat-A).
+#   - Workflow-level env.GITHUB_SERVER_URL set per
+#     feedback_act_runner_github_server_url.
+#   - `continue-on-error: true` on the job (RFC §1 contract — surface
+#     defects without blocking; follow-up PR flips after triage).
+#
+# Structural gate: TEMPLATES list in publish-runtime.yml must match
+# manifest.json's workspace_templates exactly. Closes the recurrence
+# path of PR #2556 (the data fix) and is the first concrete deliverable
+# of RFC #388 PR-3.
+#
+# Triggers narrowly to keep CI quiet: only on PRs that actually change
+# one of the two files. The path-filtered split + always-emit-result
+# pattern (memory: "Required check names need a job that always runs")
+# is unnecessary here because the workflow IS the check name and PR
+# branch protection should require it directly. Future-proof: if this
+# becomes a required check, add a no-op aggregator with always() so the
+# name still emits when paths don't match.
+
+on:
+  pull_request:
+    branches: [staging, main]
+    paths:
+      - manifest.json
+      - .gitea/workflows/publish-runtime.yml
+      - scripts/check-cascade-list-vs-manifest.sh
+
+env:
+  GITHUB_SERVER_URL: https://git.moleculesai.app
+
+permissions:
+  contents: read
+
+jobs:
+  # bp-exempt: drift visibility gate; CI / all-required remains the required aggregate.
+  check:
+    runs-on: ubuntu-latest
+    # Phase 3 (RFC #219 §1): surface broken workflows without blocking
+    # the PR. Follow-up PR flips this off after surfaced defects are
+    # triaged.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    continue-on-error: true
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - name: Check cascade list matches manifest
+        # Pass the .gitea/ workflow path explicitly — the script's
+        # default still points at .github/... which Category A of this
+        # sweep removes.
+        run: bash scripts/check-cascade-list-vs-manifest.sh manifest.json .gitea/workflows/publish-runtime.yml
@@ -42,9 +42,11 @@ jobs:
  check:
    name: Migration version collision check
    runs-on: ubuntu-latest
-    # Phase 4 (RFC #219 §1): 22 days green since 2026-05-11 port.
-    # mc#1982 mask removed — no surfaced defects in this lane.
-    continue-on-error: false
+    # Phase 3 (RFC #219 §1): surface broken workflows without blocking
+    # the PR. Follow-up PR flips this off after surfaced defects are
+    # triaged.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    continue-on-error: true
    timeout-minutes: 5
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
@@ -1,187 +0,0 @@
-# ci-arm64-advisory — Mac arm64 self-hosted ADVISORY fast-check lane.
-#
-# === WHY ===
-#
-# The amd64 Gitea runner pool (molecule-runner-1..20) is queue-contended
-# (internal#418). This lane offloads the *genuinely container-independent*
-# fast checks (Go build/vet/lint, shellcheck, Python lint) onto the Mac
-# arm64 self-hosted runner so developers get a fast arm64 signal WITHOUT
-# adding load to the starved amd64 pool — capability-honestly, as an
-# additive pilot. Pilot ② of the Mac-CI strategy (CTO-delegated 2026-05-17).
-#
-# === NON-NEGOTIABLE SAFETY CONTRACT (the prime directive) ===
-#
-# This lane is **ADVISORY ONLY**. It is provably incapable of hanging a
-# merge. Concretely:
-#
-#   1. It is a SEPARATE workflow file. `ci.yml` is byte-for-byte
-#      untouched by this PR. The `CI / all-required` aggregator sentinel
-#      and the five contexts it polls
-#      (`CI / Detect changes|Platform (Go)|Canvas (Next.js)|
-#      Shellcheck (E2E scripts)|Python Lint & Test (pull_request)`)
-#      are unchanged. The canonical required gate stays 100% on the
-#      existing amd64 pool.
-#
-#   2. The context this workflow emits is
-#      `ci-arm64-advisory / fast-checks (pull_request)`. That string is
-#      DELIBERATELY NOT present in, and this PR does NOT add it to:
-#        - branch_protections/{main,staging}.status_check_contexts
-#          (DB-verified pb 86/75 = exactly
-#           ["CI / all-required (pull_request)",
-#            "sop-checklist / all-items-acked (pull_request)"])
-#        - audit-force-merge.yml REQUIRED_CHECKS env
-#        - ci.yml `all-required` sentinel's hardcoded `required[]` list
-#      Branch protection therefore never waits on this context. If the
-#      Mac runner is absent / offline / removed, this workflow's status
-#      simply never appears — and because nothing requires it, every
-#      merge proceeds exactly as it does today. There is no path by
-#      which a missing/red arm64 status blocks a merge.
-#
-#   3. `continue-on-error: true` on the job — even a genuine arm64-only
-#      failure (toolchain drift, arch-specific test flake) is surfaced
-#      as information, never as a merge blocker, for the duration of
-#      the pilot.
-#
-#   4. The job carries a `github.event_name` `if:` gate. Beyond its
-#      functional purpose this also keeps the job OUT of
-#      `ci-required-drift.py:ci_job_names()` (which excludes
-#      `github.event_name`/`github.ref`-gated jobs), so the hourly
-#      ci-required-drift sentinel's F1 ("job not under sentinel needs")
-#      cannot ever flag this advisory job. F2/F3 are untouched because
-#      this context is absent from BP and from REQUIRED_CHECKS.
-#      `lint-bp-context-emit-match` only fails on BP→emitter gaps; an
-#      emitter without a BP context is explicitly informational there.
-#
-# === RUNNER TARGETING ===
-#
-# The Mac runner is `hongming-pc-runner-1`. The bare `self-hosted`
-# label is POLLUTED in this Gitea instance: molecule-runner-1..20
-# (the contended amd64 pool) also advertise `self-hosted`. Targeting
-# bare `self-hosted` would route back onto the very pool we are trying
-# to relieve — and onto amd64 hardware. We therefore require an
-# AND-set of labels that ONLY the Mac satisfies. `macos-self-hosted`
-# is Mac-exclusive (the amd64 pool does not carry it). Until the
-# label-install burst (a10862b2) lands `self-hosted`+`macos-self-hosted`
-# on the Mac, the runner's current unique label `hongming-pc-laptop`
-# is also listed; AND-semantics over the labels a runner advertises
-# means a job requiring [self-hosted, macos-self-hosted] can ONLY be
-# claimed once the Mac advertises both. If neither label set is yet
-# present on the Mac, the workflow stays queued harmlessly and is
-# garbage-collected by the normal stale-run reaper — it blocks nothing
-# (see safety contract point 2).
-#
-# === ROLLBACK ===
-#
-# Delete this single file (`git rm .gitea/workflows/ci-arm64-advisory.yml`)
-# and merge. No branch-protection edit, no ci.yml edit, no
-# REQUIRED_CHECKS edit is required to roll back, because none were made
-# to roll forward. Zero blast radius either direction.
-
-name: ci-arm64-advisory
-
-on:
-  push:
-    branches: [main, staging]
-  pull_request:
-    branches: [main, staging]
-
-# Per-ref cancel: a newer commit on the same ref supersedes the older
-# advisory run. Distinct from ci.yml's `ci-${ref}` group so this lane
-# never cancels (or is cancelled by) the canonical required CI.
-concurrency:
-  group: ci-arm64-advisory-${{ github.ref }}
-  cancel-in-progress: true
-
-env:
-  GITHUB_SERVER_URL: https://git.moleculesai.app
-
-jobs:
-  fast-checks:
-    name: fast-checks
-    # AND-set: only the Mac arm64 runner advertises macos-self-hosted.
-    # See "RUNNER TARGETING" header note for why bare self-hosted is unsafe.
-    runs-on: [self-hosted, macos-self-hosted]
-    # ADVISORY: never blocks. See safety contract point 3. mc#1982
-    # internal#418 — tracked: arm64 advisory pilot, non-gating by design.
-    continue-on-error: true
-    # event_name gate: functional (only meaningful on push/PR) AND keeps
-    # this job out of ci-required-drift.py:ci_job_names() so F1 can never
-    # flag it. See safety contract point 4.
-    if: ${{ github.event_name == 'push' || github.event_name == 'pull_request' }}
-    timeout-minutes: 20
-    steps:
-      - name: Provenance — advisory lane, non-gating
-        run: |
-          echo "This is the arm64 ADVISORY fast-check lane."
-          echo "It does NOT gate merges. Canonical required CI is ci.yml"
-          echo "on the amd64 pool. Arch: $(uname -m) on $(uname -s)."
-
-      - name: Checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-
-      # ---- Go: build + vet + lint (container-independent: needs only the
-      # Go toolchain; no amd64 ECR image, no docker-in-job). Race-detector
-      # unit-test + coverage gates are deliberately NOT duplicated here —
-      # those stay authoritative on amd64 ci.yml `Platform (Go)`. This lane
-      # is fast-feedback for the compile/vet/lint surface only. ----
-      - name: Setup Go
-        uses: actions/setup-go@40f1582b2485089dde7abd97c1529aa768e1baff # v5
-        with:
-          go-version: 'stable'
-      - name: Go build + vet (workspace-server)
-        working-directory: workspace-server
-        run: |
-          go mod download
-          go build ./cmd/server
-          go vet ./...
-      - name: golangci-lint (workspace-server)
-        working-directory: workspace-server
-        run: |
-          go install github.com/golangci/golangci-lint/v2/cmd/golangci-lint@v2.12.2
-          "$(go env GOPATH)/bin/golangci-lint" run --timeout 3m ./...
-
-      # ---- Shellcheck (container-independent: shellcheck binary only).
-      # Mirrors ci.yml `Shellcheck (E2E scripts)` bulk pass scope. ----
-      - name: Install shellcheck (arm64)
-        run: |
-          if ! command -v shellcheck >/dev/null 2>&1; then
-            echo "shellcheck not preinstalled on this self-hosted runner."
-            echo "Attempting Homebrew install (Mac arm64)."
-            brew install shellcheck || {
-              echo "::warning::shellcheck unavailable on runner; advisory shellcheck skipped."
-              exit 0
-            }
-          fi
-          shellcheck --version
-      - name: Shellcheck tests/e2e + infra/scripts
-        run: |
-          command -v shellcheck >/dev/null 2>&1 || { echo "skip"; exit 0; }
-          find tests/e2e infra/scripts -type f -name '*.sh' -print0 \
-            | xargs -0 shellcheck --severity=warning
-
-      # ---- Python lint/compile (container-independent: CPython only).
-      # Lint + import-compile surface; the authoritative pytest + coverage
-      # floors stay on amd64 ci.yml `Python Lint & Test`. ----
-      - name: Setup Python
-        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
-        with:
-          python-version: '3.11'
-      - name: Python byte-compile (workspace)
-        working-directory: workspace
-        run: |
-          python -m pip install --quiet ruff || true
-          python -m compileall -q .
-          if command -v ruff >/dev/null 2>&1; then
-            ruff check . || echo "::warning::ruff findings (advisory only)"
-          fi
-
-      - name: Advisory summary
-        if: always()
-        run: |
-          {
-            echo "## arm64 advisory fast-checks complete"
-            echo ""
-            echo "This lane is **advisory** — it does not gate merges."
-            echo "Authoritative required CI remains \`CI / all-required\`"
-            echo "on the amd64 pool (\`ci.yml\`, unchanged by this PR)."
-          } >> "$GITHUB_STEP_SUMMARY"
@@ -0,0 +1,165 @@
+name: MCP Stdio Transport Regression
+
+# Regression test for molecule-ai-workspace-runtime#61:
+# asyncio.connect_read_pipe / connect_write_pipe fail with
+# ValueError: "Pipe transport is only for pipes, sockets and character devices"
+# when stdout is a regular file (openclaw capture, CI tee, debugging).
+#
+# This workflow reproduces the exact failure mode and verifies the
+# fallback to direct buffer I/O works. It runs on every PR that
+# touches the MCP server or this workflow, plus nightly cron.
+#
+# Why a separate workflow (not folded into ci.yml python-lint):
+#   - The test needs to spawn the MCP server with stdout redirected
+#     to a regular file (not a TTY/pipe), which conflicts with
+#     pytest's own capture mechanism.
+#   - It exercises the actual process spawn path (python a2a_mcp_server.py)
+#     not just unit-test mocks — closer to the real openclaw integration.
+#   - A dedicated workflow surfaces stdio-specific regressions without
+#     coupling to the broader Python test suite's coverage gate.
+
+on:
+  pull_request:
+    branches: [main, staging]
+    paths:
+      - 'workspace/a2a_mcp_server.py'
+      - 'workspace/mcp_cli.py'
+      - 'workspace/tests/test_a2a_mcp_server.py'
+      - '.gitea/workflows/ci-mcp-stdio-transport.yml'
+  push:
+    branches: [main, staging]
+    paths:
+      - 'workspace/a2a_mcp_server.py'
+      - 'workspace/mcp_cli.py'
+      - 'workspace/tests/test_a2a_mcp_server.py'
+      - '.gitea/workflows/ci-mcp-stdio-transport.yml'
+  schedule:
+    # Nightly at 04:00 UTC — catches drift from dependency updates
+    # (e.g. asyncio behavior changes in new Python patch releases).
+    - cron: '0 4 * * *'
+
+concurrency:
+  group: mcp-stdio-${{ github.ref }}
+  cancel-in-progress: true
+
+env:
+  GITHUB_SERVER_URL: https://git.moleculesai.app
+
+jobs:
+  # bp-exempt: regression canary for runtime#61; not a merge gate — informational only until promoted to required.
+  # mc#774: continue-on-error mask — new workflow, flip to false once it's green on ≥3 consecutive main runs.
+  mcp-stdio-regular-file:
+    name: MCP stdio with regular-file stdout
+    runs-on: ubuntu-latest
+    continue-on-error: true  # mc#774
+    timeout-minutes: 5
+    env:
+      WORKSPACE_ID: "00000000-0000-0000-0000-000000000001"
+    defaults:
+      run:
+        working-directory: workspace
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: '3.11'
+          cache: pip
+          cache-dependency-path: workspace/requirements.txt
+      - run: pip install -r requirements.txt pytest pytest-asyncio pytest-cov
+
+      - name: Reproduce runtime#61 — stdout as regular file
+        run: |
+          set -euo pipefail
+          echo "=== Reproducing molecule-ai-workspace-runtime#61 ==="
+          echo ""
+          echo "Before the fix, this command would fail with:"
+          echo '  ValueError: Pipe transport is only for pipes, sockets and character devices'
+          echo ""
+
+          # Spawn the MCP server with stdout redirected to a regular file.
+          # This is exactly what openclaw does when capturing MCP output.
+          OUTPUT=$(mktemp)
+          trap 'rm -f "$OUTPUT"' EXIT
+
+          # Send initialize request, then tools/list, then exit
+          {
+            echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}'
+            echo '{"jsonrpc":"2.0","id":2,"method":"tools/list"}'
+          } | python a2a_mcp_server.py > "$OUTPUT" 2>&1 || {
+            RC=$?
+            echo "FAIL: MCP server exited with code $RC"
+            echo "--- stdout+stderr ---"
+            cat "$OUTPUT"
+            exit 1
+          }
+
+          echo "PASS: MCP server handled regular-file stdout without crashing"
+          echo ""
+          echo "--- Output (first 20 lines) ---"
+          head -20 "$OUTPUT"
+          echo ""
+
+          # Verify we got valid JSON-RPC responses
+          if grep -q '"result"' "$OUTPUT"; then
+            echo "PASS: JSON-RPC responses found in output"
+          else
+            echo "FAIL: No JSON-RPC responses in output"
+            cat "$OUTPUT"
+            exit 1
+          fi
+
+      - name: Reproduce runtime#61 — stdin from regular file
+        run: |
+          set -euo pipefail
+          echo "=== stdin as regular file (CI tee / capture pattern) ==="
+
+          INPUT=$(mktemp)
+          OUTPUT=$(mktemp)
+          trap 'rm -f "$INPUT" "$OUTPUT"' EXIT
+
+          cat > "$INPUT" <<'EOF'
+          {"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}
+          {"jsonrpc":"2.0","id":2,"method":"tools/list"}
+          EOF
+
+          python a2a_mcp_server.py < "$INPUT" > "$OUTPUT" 2>&1 || {
+            RC=$?
+            echo "FAIL: MCP server exited with code $RC"
+            cat "$OUTPUT"
+            exit 1
+          }
+
+          echo "PASS: MCP server handled regular-file stdin without crashing"
+
+          if grep -q '"result"' "$OUTPUT"; then
+            echo "PASS: JSON-RPC responses found in output"
+          else
+            echo "FAIL: No JSON-RPC responses in output"
+            cat "$OUTPUT"
+            exit 1
+          fi
+
+      - name: Verify warning is emitted for non-pipe stdio
+        run: |
+          set -euo pipefail
+          echo "=== Verify diagnostic warning ==="
+
+          OUTPUT=$(mktemp)
+          trap 'rm -f "$OUTPUT"' EXIT
+
+          {
+            echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}'
+          } | python a2a_mcp_server.py > "$OUTPUT" 2>&1
+
+          # The warning should mention "not a pipe" for operator visibility
+          if grep -qi "not a pipe" "$OUTPUT"; then
+            echo "PASS: Diagnostic warning emitted for non-pipe stdio"
+          else
+            echo "NOTE: No warning in output (may be suppressed by log level)"
+          fi
+
+      - name: Run unit tests for stdio transport
+        run: |
+          set -euo pipefail
+          echo "=== Running stdio transport unit tests ==="
+          python -m pytest tests/test_a2a_mcp_server.py::TestStdioPipeAssertion -v --no-cov
@@ -86,27 +86,55 @@ jobs:
        with:
          fetch-depth: 0
      - id: check
-        env:
-          PR_BASE_SHA: ${{ github.event.pull_request.base.sha }}
-          PR_BASE_REF: ${{ github.event.pull_request.base.ref }}
-          PUSH_BEFORE: ${{ github.event.before }}
        run: |
-          python3 .gitea/scripts/detect-changes.py \
-            --profile ci \
-            --event-name "${{ github.event_name }}" \
-            --pr-base-sha "$PR_BASE_SHA" \
-            --base-ref "$PR_BASE_REF" \
-            --push-before "${GITHUB_EVENT_BEFORE:-$PUSH_BEFORE}"
+          # For PR events: diff against the base branch (not HEAD~1 of the branch,
+          # which may be unrelated after force-pushes). When a push updates a PR,
+          # both pull_request and push events fire — prefer the PR base so that
+          # the diff is always computed against the actual merge base, not the
+          # previous SHA on the branch which may be on a different history line.
+          BASE="${GITHUB_BASE_REF:-${{ github.event.before }}}"
+          # GITHUB_BASE_REF is set for PR events (the base branch name).
+          # For pull_request events we use the stored base.sha; for push events
+          # (or when base.sha is unavailable) fall back to github.event.before.
+          if [ "${{ github.event_name }}" = "pull_request" ] && [ -n "${{ github.event.pull_request.base.sha }}" ]; then
+            BASE="${{ github.event.pull_request.base.sha }}"
+          fi
+          # Fallback: if BASE is empty or all zeros (new branch), run everything
+          if [ -z "$BASE" ] || echo "$BASE" | grep -qE '^0+$'; then
+            echo "platform=true" >> "$GITHUB_OUTPUT"
+            echo "canvas=true" >> "$GITHUB_OUTPUT"
+            echo "python=true" >> "$GITHUB_OUTPUT"
+            echo "scripts=true" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          # Workflow-only edits are covered by the workflow lint family
+          # and by this workflow's always-present required jobs. Do not fan
+          # those edits out into Go/Canvas/Python/shellcheck work; the
+          # downstream jobs still emit their required contexts via no-op
+          # steps when their surface flag is false.
+          #
+          # If the diff itself cannot be trusted, fail open by running every
+          # surface instead of silently under-testing the PR.
+          if ! DIFF=$(git diff --name-only "$BASE" HEAD 2>/dev/null); then
+            echo "platform=true" >> "$GITHUB_OUTPUT"
+            echo "canvas=true" >> "$GITHUB_OUTPUT"
+            echo "python=true" >> "$GITHUB_OUTPUT"
+            echo "scripts=true" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          echo "platform=$(echo "$DIFF" | grep -qE '^workspace-server/' && echo true || echo false)" >> "$GITHUB_OUTPUT"
+          echo "canvas=$(echo "$DIFF" | grep -qE '^canvas/' && echo true || echo false)" >> "$GITHUB_OUTPUT"
+          echo "python=$(echo "$DIFF" | grep -qE '^workspace/' && echo true || echo false)" >> "$GITHUB_OUTPUT"
+          echo "scripts=$(echo "$DIFF" | grep -qE '^tests/e2e/|^scripts/|^infra/scripts/' && echo true || echo false)" >> "$GITHUB_OUTPUT"

-  # Platform (Go) — Go build/vet/test/lint + coverage gates. The job always
-  # emits the required context, but expensive steps are path-scoped on every
-  # event so docs/E2E/Canvas-only main pushes do not block deploy on unrelated
-  # Go bootstrap work.
+  # Platform (Go) — Go build/vet/test/lint + coverage gates. The always-run
+  # + per-step gating shape preserves the GitHub-side required-check name
+  # contract (so when this Gitea port becomes a required check in Phase 4,
+  # the name match works on PRs that don't touch workspace-server/).
  platform-build:
    name: Platform (Go)
-    needs: changes
    runs-on: ubuntu-latest
-    # mc#1982 (closed 2026-05-14): Phase 4 flip of the platform-build job.
+    # mc#774 (closed 2026-05-14): Phase 4 flip of the platform-build job.
    # Phase 4 (#656) originally flipped this to continue-on-error: false based on
    # Phase-3-masked "green on main 2026-05-12". Two failure classes then surfaced:
    #   (1) 4x delegation_test.go sqlmock gaps (PR #669 / #634 fix-forward, closed).
@@ -125,29 +153,29 @@ jobs:
      run:
        working-directory: workspace-server
    steps:
-      - if: ${{ needs.changes.outputs.platform != 'true' }}
+      - if: false
        working-directory: .
-        run: echo "No workspace-server/** changes — Platform (Go) gate satisfied without running Go build/test/lint."
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+        run: echo "No platform/** changes — skipping real build steps; this job always runs to satisfy the required-check name on branch protection."
+      - if: always()
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+      - if: always()
        uses: actions/setup-go@40f1582b2485089dde7abd97c1529aa768e1baff # v5
        with:
          go-version: 'stable'
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+      - if: always()
        run: go mod download
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+      - if: always()
        run: go build ./cmd/server
      # CLI (molecli) moved to standalone repo: git.moleculesai.app/molecule-ai/molecule-cli
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+      - if: always()
        run: go vet ./...
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+      - if: always()
        name: Install golangci-lint
        run: go install github.com/golangci/golangci-lint/v2/cmd/golangci-lint@v2.12.2
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+      - if: always()
        name: Run golangci-lint
        run: $(go env GOPATH)/bin/golangci-lint run --timeout 3m ./...
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+      - if: always()
        name: Diagnostic — per-package verbose 60s
        run: |
          set +e
@@ -161,25 +189,17 @@ jobs:
          echo "::group::pendinguploads exit=$pu_exit (last 100 lines)"
          tail -100 /tmp/test-pu.log
          echo "::endgroup::"
-        # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+        # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
        continue-on-error: true
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
-        name: Run tests with coverage (blocking gate)
-        # Removed -race from the blocking gate per #1184: cold runners
-        # take 13-25 min to compile with race instrumentation, exceeding
-        # the 10m step timeout and causing false failures. Race detection
-        # now runs as a non-blocking advisory step below.
-        run: go test -timeout 10m -coverprofile=coverage.out ./...
+      - if: always()
+        name: Run tests with race detection and coverage
+        # Explicit timeout: cold runner cache causes OOM kills at ~4m39s on the
+        # full ./... suite with race detection + coverage. A 10m per-step timeout
+        # lets the suite complete on cold cache (~5-7m) while failing cleanly
+        # instead of OOM-killing. The job-level timeout (15m) is a backstop.
+        run: go test -race -timeout 10m -coverprofile=coverage.out ./...

-      - if: ${{ needs.changes.outputs.platform == 'true' }}
-        name: Race detection (advisory, non-blocking)
-        # mc#1184: runs race detector as an advisory check so cold-runner
-        # compile-time spikes don't block merges. Failures here surface in
-        # the run log but do not fail the build.
-        run: go test -race -timeout 10m ./...
-        continue-on-error: true
-
-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+      - if: always()
        name: Per-file coverage report
        # Advisory — lists every source file with its coverage so reviewers
        # can see at-a-glance where gaps are. Sorted ascending so the worst
@@ -193,7 +213,7 @@ jobs:
                   END {for (f in s) printf "%6.1f%%  %s\n", s[f]/c[f], f}' \
            | sort -n

-      - if: ${{ needs.changes.outputs.platform == 'true' }}
+      - if: always()
        name: Check coverage thresholds
        # Enforces two gates from #1823 Layer 1:
        #   1. Total floor (25% — ratchet plan in COVERAGE_FLOOR.md).
@@ -247,7 +267,7 @@ jobs:
              # Strip the package-import prefix so we can match .coverage-allowlist.txt
              # entries written as paths relative to workspace-server/.
              # Handle both module paths: platform/workspace-server/... and platform/...
-              rel=$(echo "$file" | sed 's|^git.moleculesai.app/molecule-ai/molecule-core/workspace-server/workspace-server/||; s|^git.moleculesai.app/molecule-ai/molecule-core/workspace-server/||')
+              rel=$(echo "$file" | sed 's|^github.com/molecule-ai/molecule-monorepo/platform/workspace-server/||; s|^github.com/molecule-ai/molecule-monorepo/platform/||')

              if echo "$ALLOWLIST" | grep -qxF "$rel"; then
                echo "::warning file=workspace-server/$rel::Critical file at ${pct}% coverage (allowlisted, #1823) — fix before expiry."
@@ -281,7 +301,6 @@ jobs:
  # siblings — verified empirically on PR #2314).
  canvas-build:
    name: Canvas (Next.js)
-    needs: changes
    runs-on: ubuntu-latest
    timeout-minutes: 20
    # Phase 4 (RFC #219 §1): confirmed green on main 2026-05-12.
@@ -290,20 +309,20 @@ jobs:
      run:
        working-directory: canvas
    steps:
-      - if: ${{ needs.changes.outputs.canvas != 'true' }}
+      - if: false
        working-directory: .
-        run: echo "No canvas/** changes — Canvas (Next.js) gate satisfied without running npm build/test."
-      - if: ${{ needs.changes.outputs.canvas == 'true' }}
+        run: echo "No canvas/** changes — skipping real build steps; this job always runs to satisfy the required-check name on branch protection."
+      - if: always()
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-      - if: ${{ needs.changes.outputs.canvas == 'true' }}
+      - if: always()
        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
        with:
          node-version: '22'
-      - if: ${{ needs.changes.outputs.canvas == 'true' }}
-        run: npm ci --include=optional --prefer-offline
-      - if: ${{ needs.changes.outputs.canvas == 'true' }}
+      - if: always()
+        run: rm -f package-lock.json && npm install
+      - if: always()
        run: npm run build
-      - if: ${{ needs.changes.outputs.canvas == 'true' }}
+      - if: always()
        name: Run tests with coverage
        # Coverage instrumentation is configured in canvas/vitest.config.ts
        # (provider: v8, reporters: text + html + json-summary). Step 2 of
@@ -312,7 +331,7 @@ jobs:
        # tracked in #1815) after the team sees what current coverage is.
        run: npx vitest run --coverage
      - name: Upload coverage summary as artifact
-        if: ${{ needs.changes.outputs.canvas == 'true' }}
+        if: always()
        # Pinned to v3 for Gitea act_runner v0.6 compatibility — v4+ uses
        # the GHES 3.10+ artifact protocol that Gitea 1.22.x does NOT
        # implement, surfacing as `GHESNotSupportedError: @actions/artifact
@@ -326,7 +345,7 @@ jobs:
          retention-days: 7
          if-no-files-found: warn

-  # Shellcheck (E2E scripts) — required context, path-scoped heavy steps.
+  # Shellcheck (E2E scripts) — required check, always runs.
  shellcheck:
    name: Shellcheck (E2E scripts)
    needs: changes
@@ -334,11 +353,11 @@ jobs:
    # Phase 4 (RFC #219 §1): confirmed green on main 2026-05-12.
    continue-on-error: false
    steps:
-      - if: ${{ needs.changes.outputs.scripts != 'true' }}
-        run: echo "No tests/e2e, scripts, or infra/scripts changes — Shellcheck gate satisfied without running script checks."
-      - if: ${{ needs.changes.outputs.scripts == 'true' }}
+      - if: needs.changes.outputs.scripts != 'true'
+        run: echo "No tests/e2e/ or infra/scripts/ changes — skipping real shellcheck; this job always runs to satisfy the required-check name on branch protection."
+      - if: needs.changes.outputs.scripts == 'true'
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-      - if: ${{ needs.changes.outputs.scripts == 'true' }}
+      - if: needs.changes.outputs.scripts == 'true'
        name: Run shellcheck on tests/e2e/*.sh and infra/scripts/*.sh
        # shellcheck is pre-installed on ubuntu-latest runners (via apt).
        # infra/scripts/ is included because setup.sh + nuke.sh gate the
@@ -349,24 +368,16 @@ jobs:
          find tests/e2e infra/scripts -type f -name '*.sh' -print0 \
            | xargs -0 shellcheck --severity=warning

-      - if: ${{ needs.changes.outputs.scripts == 'true' }}
+      - if: needs.changes.outputs.scripts == 'true'
        name: Lint cleanup-trap hygiene (RFC #2873)
        run: bash tests/e2e/lint_cleanup_traps.sh

-      - if: ${{ needs.changes.outputs.scripts == 'true' }}
+      - if: needs.changes.outputs.scripts == 'true'
        name: Run E2E bash unit tests (no live infra)
        run: |
          bash tests/e2e/test_model_slug.sh
-          # molecule-core#1995 (#1994 follow-on): fail-direction proof for
-          # the A2A real-completion + byok-routing assertion helpers
-          # (lib/completion_assert.sh). Offline (no LLM, no network): it
-          # asserts an error-as-text payload FAILS the real-completion gate
-          # — the exact trap the historical shape-only `"kind":"text"`
-          # check missed. If a refactor weakens the gate to a shape check,
-          # this step goes red on every PR.
-          bash tests/e2e/test_completion_assert_unit.sh

-      - if: ${{ needs.changes.outputs.scripts == 'true' }}
+      - if: needs.changes.outputs.scripts == 'true'
        name: Test ECR promote-tenant-image script (mock-driven, no live infra)
        # Covers scripts/promote-tenant-image.sh — the codified
        # :staging-latest → :latest ECR promote + tenant fleet redeploy
@@ -376,7 +387,7 @@ jobs:
        run: |
          bash scripts/test-promote-tenant-image.sh

-      - if: ${{ needs.changes.outputs.scripts == 'true' }}
+      - if: needs.changes.outputs.scripts == 'true'
        name: Shellcheck promote-tenant-image script
        # scripts/ is excluded from the bulk shellcheck pass above (legacy
        # SC3040/SC3043 cleanup pending). Run shellcheck explicitly on
@@ -391,13 +402,13 @@ jobs:

  canvas-deploy-reminder:
    name: Canvas Deploy Reminder
-    runs-on: docker-host
-    # mc#1982 root-fix: added job-level `if:` so ci-required-drift.py's
+    runs-on: ubuntu-latest
+    # mc#774 root-fix: added job-level `if:` so ci-required-drift.py's
    # ci_job_names() detects this as github.ref-gated and skips it from F1.
    # The step-level exit 0 handles the "not main push" case; the job-level
    # `if:` makes the gating explicit so the drift script sees it.
-    # Runs on both main and staging pushes; step exits 0 when not applicable.
-    if: ${{ github.ref == 'refs/heads/main' || github.ref == 'refs/heads/staging' }}
+    # continue-on-error removed (was mc#774 mask): step exits 0 when not applicable.
+    if: ${{ github.ref == 'refs/heads/staging' }}
    needs: [changes, canvas-build]
    steps:
      - name: Write deploy reminder to step summary
@@ -446,40 +457,94 @@ jobs:
          cat /tmp/deploy-reminder.md >> "$GITHUB_STEP_SUMMARY"

  # Python Lint & Test — required check, always runs.
-  # Runtime Python moved to molecule-ai-workspace-runtime. Keep this context as
-  # a guard so branch protection still catches attempts to reintroduce an
-  # editable runtime copy under molecule-core/workspace/.
  python-lint:
    name: Python Lint & Test
+    needs: changes
    runs-on: ubuntu-latest
+    # Phase 4 (RFC #219 §1): confirmed green on main 2026-05-12.
    continue-on-error: false
+    env:
+      WORKSPACE_ID: test
+    defaults:
+      run:
+        working-directory: workspace
    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-      - name: Runtime SSOT guard
+      - if: needs.changes.outputs.python != 'true'
+        working-directory: .
+        run: echo "No workspace/** changes — skipping real lint+test; this job always runs to satisfy the required-check name on branch protection."
+      - if: needs.changes.outputs.python == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - if: needs.changes.outputs.python == 'true'
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: '3.11'
+          cache: pip
+          cache-dependency-path: workspace/requirements.txt
+      - if: needs.changes.outputs.python == 'true'
+        run: pip install -r requirements.txt pytest pytest-asyncio pytest-cov sqlalchemy>=2.0.0
+      # Coverage flags + fail-under floor moved into workspace/pytest.ini
+      # (issue #1817) so local `pytest` and CI use identical config.
+      - if: needs.changes.outputs.python == 'true'
+        run: python -m pytest --tb=short
+
+      - if: needs.changes.outputs.python == 'true'
+        name: Per-file critical-path coverage (MCP / inbox / auth)
+        # MCP-critical Python files have a per-file floor on top of the
+        # 86% total floor in pytest.ini. See issue #2790 for full rationale.
        run: |
-          set -eu
-          if [ -d workspace ]; then
-            echo "::error file=workspace::Runtime source must live in molecule-ai-workspace-runtime, not molecule-core/workspace."
-            exit 1
-          fi
-          for f in scripts/build_runtime_package.py scripts/test_build_runtime_package.py; do
-            if [ -e "$f" ]; then
-              echo "::error file=$f::Legacy build-from-workspace packaging script must not be restored."
-              exit 1
+          set -e
+          PER_FILE_FLOOR=75
+          CRITICAL_FILES=(
+            "a2a_mcp_server.py"
+            "mcp_cli.py"
+            "a2a_tools.py"
+            "a2a_tools_inbox.py"
+            "inbox.py"
+            "platform_auth.py"
+          )
+
+          # pytest already wrote .coverage; emit a JSON view scoped to
+          # the critical files so jq/python can read the per-file pct
+          # without parsing tabular text.
+          INCLUDES=$(printf '*%s,' "${CRITICAL_FILES[@]}")
+          INCLUDES="${INCLUDES%,}"
+          python -m coverage json -o /tmp/critical-cov.json --include="$INCLUDES"
+
+          FAILED=0
+          for f in "${CRITICAL_FILES[@]}"; do
+            pct=$(jq -r --arg f "$f" '.files | to_entries | map(select(.key == $f)) | .[0].value.summary.percent_covered // "MISSING"' /tmp/critical-cov.json)
+            if [ "$pct" = "MISSING" ]; then
+              echo "::error file=workspace/$f::No coverage data — file may have moved or test exclusion mis-set."
+              FAILED=$((FAILED+1))
+              continue
+            fi
+            echo "$f: ${pct}%"
+            if awk "BEGIN{exit !($pct < $PER_FILE_FLOOR)}"; then
+              echo "::error file=workspace/$f::${pct}% < ${PER_FILE_FLOOR}% per-file floor (MCP critical path). See COVERAGE_FLOOR.md."
+              FAILED=$((FAILED+1))
            fi
          done
-          echo "Runtime SSOT guard passed; core consumes the standalone runtime package."
+
+          if [ "$FAILED" -gt 0 ]; then
+            echo ""
+            echo "$FAILED MCP critical-path file(s) below the ${PER_FILE_FLOOR}% per-file floor."
+            echo "These paths handle multi-tenant routing, auth tokens, and inbox dispatch."
+            echo "A coverage drop here is the same risk shape as Go-side tokens/secrets files"
+            echo "dropping below 10% (see COVERAGE_FLOOR.md). Either:"
+            echo "  (a) add tests to raise coverage back above ${PER_FILE_FLOOR}%, or"
+            echo "  (b) if this is unavoidable historical debt, file an issue and propose"
+            echo "      adjusting the floor with rationale in COVERAGE_FLOOR.md."
+            exit 1
+          fi

  all-required:
    # Aggregator sentinel — RFC internal#219 §2 (Phase 4 — closes internal#286).
    #
-    # Emits `CI / all-required (<event>)` where <event> is the workflow trigger
-    # (e.g. `CI / all-required (pull_request)`, `CI / all-required (push)`).
-    # Branch protection requires the event-suffixed name —
-    # requiring `CI / all-required` (bare, no suffix) silently blocks all merges
-    # because Gitea treats absent status contexts as pending (not skipped), and
-    # no workflow emits the bare name. BP requires
-    # `CI / all-required (pull_request)` per issue #1473.
+    # Single stable required-status name that branch protection points at;
+    # CI churns underneath in `needs:` without any protection edits. Mirrors
+    # the molecule-controlplane Phase 2a impl shipped in CP PR#112 and
+    # referenced by `internal#286` ("Phase 4 is a single small PR... mirrors
+    # CP's existing one").
    #
    # Closes the failure mode where status_check_contexts on molecule-core/main
    # only listed `Secret scan` + `sop-tier-check` (the 2 meta-gates), so real
@@ -487,91 +552,86 @@ jobs:
    # red silently merged through. See internal#286 for the three concrete
    # tonight-of-2026-05-11 incidents that prompted the emergency bump.
    #
-    # ── 2026-06-01 CI-scheduler-overload fix (fix/ci-scheduler-fanout) ──
-    # PREVIOUS shape: a poll-gate that ran detect-changes then LOOPED on
-    # `GET /commits/{sha}/statuses` every 15s for up to 40 min, occupying a
-    # `ci-meta` executor slot the entire time it waited for upstream jobs.
-    # With only 2 ci-meta runners, that poll-loop squatted half the lane on
-    # every PR — a confirmed throughput sink in the live RCA (two concurrent
-    # `JOB-all-required` containers observed pinning the lane). The polling
-    # design existed only to dodge the Gitea `needs:` + `if: always()` bug,
-    # where an always()-guarded sentinel could be marked skipped before
-    # upstream jobs settled (leaving BP pending forever).
+    # Three properties of this job each close a failure mode:
    #
-    # NEW shape: a plain `needs:` aggregator with NO polling loop. This is
-    # safe here — and was NOT safe at the time the poller was written —
-    # because every aggregated CI job now gates its real work PER-STEP
-    # (`if: needs.changes.outputs.* != 'true'`) rather than at the JOB level.
-    # A per-step-gated job always reaches a terminal SUCCESS (it no-ops its
-    # expensive steps but the job itself still completes), so it is never
-    # `skipped`. Plain `needs:` (WITHOUT `if: always()`) works correctly on
-    # Gitea 1.22.6 / act_runner v0.6.1 — only `needs:` + `if: always()` is
-    # broken (feedback_gitea_needs_works_only_ifalways_broken). We therefore
-    # use plain `needs:` + an explicit per-need result check (NOT
-    # `if: always()`); if any need fails/errors, Gitea never starts this job
-    # and BP sees `CI / all-required` go red via the failed dependency
-    # propagation — exactly the gate we want, with zero runner-squat.
+    #  1. `if: always()` — runs even when an upstream fails. Without it the
+    #     sentinel is `skipped` and protection treats that as missing → merge
+    #     ungated.
    #
-    # The `needs:` list MUST stay in lockstep with ci-required-drift.py's
-    # F1 check (`ci_job_names()` = every job MINUS the sentinel MINUS jobs
-    # whose `if:` gates on github.event_name/github.ref). canvas-deploy-
-    # reminder is event-gated (`if: github.ref == refs/heads/{main,staging}`)
-    # so it is intentionally EXCLUDED — it skips on PRs and a `needs:` on a
-    # skipped job would never let the sentinel run. If a new always-running
-    # CI job is added, add it here too or ci-required-drift F1 will flag it.
+    #  2. Assertion is `result == "success"` per dep, NOT `!= "failure"`.
+    #     A `skipped` upstream (job gated by `if:` evaluating false, matrix
+    #     entry that couldn't run) must NOT silently pass through.
+    #     `skipped`-as-green is exactly the failure mode this gate closes.
    #
-    # Stays on the dedicated `ci-meta` lane (no docker work, so the
-    # docker-host-pin lint does not apply), but now the job is sub-second:
-    # it only inspects already-settled `needs.*.result` values, so it frees
-    # the slot immediately instead of holding it for the whole CI duration.
+    #  3. `needs:` is the canonical list of "what counts as required."
+    #     status_check_contexts will reference only `ci/all-required` (Step 5
+    #     follow-up — branch-protection PATCH is Owners-tier per
+    #     `feedback_never_admin_merge_bypass`, separate PR); a new job is
+    #     added simply by listing it in `needs:` here.
+    #     `.gitea/workflows/ci-required-drift.yml` files a [ci-drift] issue
+    #     hourly if this list diverges from status_check_contexts or from
+    #     audit-force-merge.yml's REQUIRED_CHECKS env (RFC §4 + §6).
    #
+    # canvas-deploy-reminder is intentionally excluded from all-required.needs:
+    # it needs canvas-build, which is skipped on CI-only PRs (canvas=false).
+    # Including it in all-required.needs causes all-required to hang on
+    # every CI-only PR. Keep it runnable on PRs via its own
+    # `needs: [changes, canvas-build]` — the sentinel only aggregates the result.
+    #
+    # Phase 3 (RFC #219 §1) safety: underlying build jobs carry
+    # continue-on-error: true so their failures are masked to null (2026-05-12: re-enabled mc#774 interim)
+    # (Gitea suppresses status reporting for CoE jobs). This sentinel
+    # runs with continue-on-error: false so it always reports its
+    # result to the API — without this, the required-status entry
+    # (CI / all-required (pull_request)) is never created, which
+    # blocks PR merges. When Phase 3 ends, flip underlying jobs to
+    # continue-on-error: false; this sentinel can then be flipped to
+    # continue-on-error: true if a Phase-4 regression requires it.
+    continue-on-error: false
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
    needs:
      - changes
      - platform-build
      - canvas-build
      - shellcheck
      - python-lint
-    continue-on-error: false
-    runs-on: ci-meta
-    timeout-minutes: 5
+      - canvas-deploy-reminder
+    if: ${{ always() }}
    steps:
-      - name: Verify all aggregated CI jobs succeeded
-        # NO polling, NO API call, NO checkout. Because this job lists the
-        # aggregated jobs under `needs:` (without `if: always()`), Gitea only
-        # starts it once every need has reached SUCCESS — a failed/errored
-        # need short-circuits the job and propagates red to the
-        # `CI / all-required` context. This explicit check is a
-        # belt-and-suspenders assertion + a readable run summary; the real
-        # gating is the `needs:` edge itself.
-        env:
-          CHANGES_RESULT: ${{ needs.changes.result }}
-          PLATFORM_RESULT: ${{ needs.platform-build.result }}
-          CANVAS_RESULT: ${{ needs.canvas-build.result }}
-          SHELLCHECK_RESULT: ${{ needs.shellcheck.result }}
-          PYTHON_LINT_RESULT: ${{ needs.python-lint.result }}
+      - name: Assert every required dependency succeeded
        run: |
          set -euo pipefail
-          fail=0
-          check() {
-            name="$1"; result="$2"
-            printf 'CI / %s = %s\n' "$name" "$result"
-            # `success` is the only green terminal state we accept. A plain
-            # `needs:` job is only started when all needs succeed, so reaching
-            # this step already implies success — but assert explicitly so a
-            # future `if: always()` reintroduction (which WOULD let non-success
-            # through) fails loudly instead of silently passing the gate.
-            if [ "$result" != "success" ]; then
-              echo "::error::aggregated CI job '${name}' did not succeed (result=${result})"
-              fail=1
-            fi
-          }
-          check "Detect changes"        "$CHANGES_RESULT"
-          check "Platform (Go)"         "$PLATFORM_RESULT"
-          check "Canvas (Next.js)"      "$CANVAS_RESULT"
-          check "Shellcheck (E2E scripts)" "$SHELLCHECK_RESULT"
-          check "Python Lint & Test"    "$PYTHON_LINT_RESULT"
-          if [ "$fail" -ne 0 ]; then
-            echo "::error::all-required: one or more aggregated CI jobs did not succeed"
-            exit 1
-          fi
-          echo "OK: all aggregated CI jobs succeeded — CI / all-required green."
+          # `needs.*.result` is one of: success | failure | cancelled | skipped | null.
+          # We assert success per dep (not != failure) — see RFC §2 reasoning above.
+          # Null results are skipped: they come from Phase 3 (continue-on-error: true
+          # suppresses status) or from jobs still in-flight. The sentinel succeeds
+          # rather than blocking PRs on Phase 3 noise.
+          results='${{ toJSON(needs) }}'
+          echo "$results"
+          echo "$results" | python3 -c '
+          import json, sys
+          ns = json.load(sys.stdin)
+          # Phase 3 masked: jobs with continue-on-error: true may report "failure"
+          # Remove when mc#774 handler test failures are resolved.
+          PHASE3_MASKED = {"platform-build"}
+          # Exclude null (Phase 3 suppressed / in-flight) from the bad list.
+          bad = [(k, v.get("result")) for k, v in ns.items()
+                 if v.get("result") not in ("success", None, "cancelled", "skipped") and k not in PHASE3_MASKED]
+          if bad:
+              print(f"FAIL: jobs not green:", file=sys.stderr)
+              for k, r in bad:
+                  print(f"  - {k}: {r}", file=sys.stderr)
+              sys.exit(1)
+          pending = [(k, v.get("result")) for k, v in ns.items()
+                     if v.get("result") is None]
+          cancelled = [(k, v.get("result")) for k, v in ns.items()
+                       if v.get("result") == "cancelled"]
+          if pending:
+              print(f"WARN: {len(pending)} job(s) still in-flight (result=null): " +
+                    ", ".join(k for k, _ in pending), file=sys.stderr)
+          if cancelled:
+              print(f"INFO: {len(cancelled)} job(s) masked by continue-on-error: " +
+                    ", ".join(k for k, _ in cancelled), file=sys.stderr)
+          print(f"OK: all {len(ns)} required jobs succeeded (or Phase-3 suppressed)")
+          '
@@ -43,18 +43,6 @@ name: Continuous synthetic E2E (staging)

 on:
  schedule:
-    # Every 30 minutes, on :02 and :32. This keeps a recurring SaaS
-    # behavior probe while cutting runner occupancy from this workflow by
-    # roughly two thirds; fast liveness belongs in the lighter smoke/heartbeat
-    # probes, not in a full tenant/workspace synth every 10 minutes.
-    #
-    # Previous cadence was every 10 minutes (:02 :12 :22 :32 :42 :52).
-    # The current operator-host runner pool is the bottleneck, so full
-    # synth E2E is deliberately lower-cadence until it moves to a dedicated
-    # runner host or warm-runtime pool.
-    #
-    # Historical notes from the 10-minute shape:
-    #
    # Every 10 minutes, on :02 :12 :22 :32 :42 :52. Three constraints:
    #   1. Stay off the top-of-hour. GitHub Actions scheduler drops
    #      :00 firings under high load (own docs:
@@ -78,7 +66,7 @@ on:
    # fires = ~30 min cadence; closer to the 20-min target than the
    # current shape and provides a real degradation alarm if drops
    # get worse.
-    - cron: '2,32 * * * *'
+    - cron: '2,12,22,32,42,52 * * * *'
 permissions:
  contents: read
  # No issue-write here — failures surface as red runs in the workflow
@@ -102,7 +90,7 @@ jobs:
    name: Synthetic E2E against staging
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    # Bumped from 12 → 20 (2026-05-04). Tenant user-data install phase
    # (apt-get update + install docker.io/jq/awscli/caddy + snap install
@@ -118,7 +106,7 @@ jobs:
    timeout-minutes: 20
    env:
      # claude-code default: cold-start ~5 min (comparable to langgraph),
-      # but uses MiniMax-M2 via the template's third-party-
+      # but uses MiniMax-M2.7-highspeed via the template's third-party-
      # Anthropic-compat path (workspace-configs-templates/claude-code-
      # default/config.yaml:64-69). MiniMax is ~5-10x cheaper than
      # gpt-4.1-mini per token AND avoids the recurring OpenAI quota-
@@ -131,9 +119,9 @@ jobs:
      # on the per-runtime default ("sonnet" → routes to direct
      # Anthropic, defeats the cost saving). Operators can override
      # via workflow_dispatch by setting a different E2E_MODEL_SLUG
-      # input if they need to exercise a specific model. MiniMax-M2 is the
-      # stable staging MiniMax path used by the full-SaaS smoke.
-      E2E_MODEL_SLUG: ${{ github.event.inputs.model_slug || 'MiniMax-M2' }}
+      # input if they need to exercise a specific model. M2.7-highspeed
+      # is "Token Plan only" but cheap-per-token and fast.
+      E2E_MODEL_SLUG: ${{ github.event.inputs.model_slug || 'MiniMax-M2.7-highspeed' }}
      # Bound to 10 min so a stuck provision fails the run instead of
      # holding up the next cron firing. 15-min default in the script
      # is for the on-PR full lifecycle where we have more headroom.
@@ -145,11 +133,6 @@ jobs:
      E2E_KEEP_ORG: ${{ github.event.inputs.keep_org == 'true' && '1' || '' }}
      MOLECULE_CP_URL: ${{ vars.STAGING_CP_URL || 'https://staging-api.moleculesai.app' }}
      MOLECULE_ADMIN_TOKEN: ${{ secrets.CP_STAGING_ADMIN_API_TOKEN }}
-      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
-      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
-      AWS_DEFAULT_REGION: us-east-2
-      E2E_AWS_LEAK_CHECK: required
-      E2E_AWS_TERMINATE_LEAKS: '1'
      # MiniMax key is the canary's PRIMARY auth path. claude-code
      # template's `minimax` provider routes ANTHROPIC_BASE_URL to
      # api.minimax.io/anthropic and reads MINIMAX_API_KEY at boot.
@@ -166,10 +149,6 @@ jobs:
      # canary path. The script picks the right blob shape based on
      # which key is non-empty.
      E2E_OPENAI_API_KEY: ${{ secrets.MOLECULE_STAGING_OPENAI_API_KEY }}
-      # google-adk canary path — AI-Studio key (config model
-      # google_genai:gemini-2.5-pro). PROD disallows API keys (Vertex+ADC);
-      # the keyed path is CI-only. Dispatch with E2E_RUNTIME=google-adk.
-      E2E_GOOGLE_API_KEY: ${{ secrets.MOLECULE_STAGING_GOOGLE_API_KEY }}
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2

@@ -194,12 +173,6 @@ jobs:
            echo "::error::Set it at Settings → Secrets and Variables → Actions; pull from staging-CP's CP_ADMIN_API_TOKEN env in Railway."
            exit 1
          fi
-          for var in AWS_ACCESS_KEY_ID AWS_SECRET_ACCESS_KEY; do
-            if [ -z "${!var:-}" ]; then
-              echo "::error::$var secret missing — EC2 leak verification cannot run"
-              exit 1
-            fi
-          done

          # LLM-key requirement is per-runtime: claude-code accepts
          # EITHER MiniMax OR direct-Anthropic (whichever is set first),
@@ -221,10 +194,6 @@ jobs:
              required_secret_name="MOLECULE_STAGING_OPENAI_API_KEY"
              required_secret_value="${E2E_OPENAI_API_KEY:-}"
              ;;
-            google-adk)
-              required_secret_name="MOLECULE_STAGING_GOOGLE_API_KEY"
-              required_secret_value="${E2E_GOOGLE_API_KEY:-}"
-              ;;
            *)
              echo "::warning::Unknown E2E_RUNTIME='${E2E_RUNTIME}' — skipping LLM-key check"
              required_secret_name=""
@@ -69,13 +69,6 @@ name: E2E API Smoke Test
 # 2318) shows Postgres ready in 3s, Redis in 1s, Platform in 1s when
 # they DO come up. Timeouts are not the bottleneck; not bumped.
 #
-# Item #1046 (fixed 2026-05-14): Stale platform-server from cancelled runs
-#   lingers on :8080 after "Stop platform" step is skipped (workflow cancelled
-#   before reaching line 335). Added a pre-start "Kill stale platform-server"
-#   step (line 286) that scans /proc for zombie platform-server processes
-#   and kills them before the port probe or bind. Makes the ephemeral port
-#   probe + start sequence deterministic.
-#
 # Item explicitly NOT fixed here: failing test `Status back online`
 # fails because the platform's langgraph workspace template image
 # (ghcr.io/molecule-ai/workspace-template-langgraph:latest) returns
@@ -108,22 +101,9 @@ env:

 jobs:
  detect-changes:
-    # mc#1529 follow-on: pin to `docker-host` so the e2e-api lane lands
-    # on Linux operator-host runners (molecule-runner-*) that carry the
-    # `molecule-core-net` bridge network + a working `aws ecr get-login-
-    # password | docker login` path. The bare `ubuntu-latest` label is
-    # also accepted by hongming-pc-runner-* (Windows act_runner v1.0.3),
-    # where the docker.sock-bound steps below fail non-deterministically
-    # (e.g. `docker run -d --name pg-e2e-api-...` with port-bind +
-    # `docker exec ... pg_isready` cannot work against a Windows daemon).
-    # detect-changes itself doesn't bind docker.sock, but pinning here too
-    # keeps both jobs on the same lane so we don't re-roll the dice on
-    # workspace-volume cross-host surprises and the routing rule is
-    # discoverable in one place. Mirror of mc#1543 (handlers-postgres-
-    # integration). See internal#512 for the class defect.
-    runs-on: docker-host
+    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    outputs:
      api: ${{ steps.decide.outputs.api }}
@@ -132,13 +112,31 @@ jobs:
        with:
          fetch-depth: 0
      - id: decide
+        # Inline replacement for dorny/paths-filter — same pattern PR#372's
+        # ci.yml port used. Diffs against the PR base or push BEFORE SHA,
+        # then matches against the api-relevant path set.
        run: |
-          python3 .gitea/scripts/detect-changes.py \
-            --profile e2e-api \
-            --event-name "${{ github.event_name }}" \
-            --pr-base-sha "${{ github.event.pull_request.base.sha }}" \
-            --base-ref "${{ github.event.pull_request.base.ref }}" \
-            --push-before "${GITHUB_EVENT_BEFORE:-${{ github.event.before }}}"
+          BASE="${GITHUB_BASE_REF:-${{ github.event.before }}}"
+          if [ "${{ github.event_name }}" = "pull_request" ] && [ -n "${{ github.event.pull_request.base.sha }}" ]; then
+            BASE="${{ github.event.pull_request.base.sha }}"
+          fi
+          if [ -z "$BASE" ] || echo "$BASE" | grep -qE '^0+$'; then
+            echo "api=true" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          if ! git cat-file -e "$BASE" 2>/dev/null; then
+            git fetch --depth=1 origin "$BASE" 2>/dev/null || true
+          fi
+          if ! git cat-file -e "$BASE" 2>/dev/null; then
+            echo "api=true" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          CHANGED=$(git diff --name-only "$BASE" HEAD)
+          if echo "$CHANGED" | grep -qE '^(workspace-server/|tests/e2e/|\.gitea/workflows/e2e-api\.yml$)'; then
+            echo "api=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "api=false" >> "$GITHUB_OUTPUT"
+          fi

  # ONE job (no job-level `if:`) that always runs and reports under the
  # required-check name `E2E API Smoke Test`. Real work is gated per-step
@@ -155,12 +153,9 @@ jobs:
  e2e-api:
    needs: detect-changes
    name: E2E API Smoke Test
-    # mc#1529 follow-on: must run on operator-host Linux runners (where
-    # docker.sock + `molecule-core-net` + `aws ecr ...` work). See
-    # detect-changes for the full rationale.
-    runs-on: docker-host
+    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 15
    env:
@@ -288,35 +283,6 @@ jobs:
          echo "PORT=${PLATFORM_PORT}" >> "$GITHUB_ENV"
          echo "BASE=http://127.0.0.1:${PLATFORM_PORT}" >> "$GITHUB_ENV"
          echo "Platform host port: ${PLATFORM_PORT}"
-      - name: Kill stale platform-server before start (issue #1046)
-        if: needs.detect-changes.outputs.api == 'true'
-        run: |
-          # Concurrent runs on the same host-network act_runner can leave a
-          # zombie platform-server from a cancelled/timeout run. Cancelled
-          # runs never reach the "Stop platform" step (line 335), so the
-          # old process lingers. Kill it before the ephemeral port probe
-          # or start so the port is definitively free.
-          #
-          # /proc scan — works on any Linux without pkill/lsof/ss.
-          # comm field is truncated to 15 chars: "platform-serve" matches
-          # "platform-server". Verify with cmdline to avoid false positives.
-          killed=0
-          for pid in $(grep -l "platform-serve" /proc/[0-9]*/comm 2>/dev/null); do
-            kpid="${pid%/comm}"
-            kpid="${kpid##*/}"
-            cmdline=$(cat "/proc/${kpid}/cmdline" 2>/dev/null | tr '\0' ' ')
-            if echo "$cmdline" | grep -q "platform-server"; then
-              echo "Killing stale platform-server pid ${kpid}: ${cmdline}"
-              kill "$kpid" 2>/dev/null || true
-              killed=$((killed + 1))
-            fi
-          done
-          if [ "$killed" -gt 0 ]; then
-            sleep 2
-            echo "Killed $killed stale process(es); port(s) released."
-          else
-            echo "No stale platform-server found."
-          fi
      - name: Start platform (background)
        if: needs.detect-changes.outputs.api == 'true'
        working-directory: workspace-server
@@ -348,9 +314,6 @@ jobs:
            exit 1
          fi
          echo "Migrations OK"
-      - name: Run today's-PR-coverage E2E (mc#1525/1535/1536/1539/1542 fix-specific assertions)
-        if: needs.detect-changes.outputs.api == 'true'
-        run: bash tests/e2e/test_today_pr_coverage_e2e.sh
      - name: Run E2E API tests
        if: needs.detect-changes.outputs.api == 'true'
        run: bash tests/e2e/test_api.sh
@@ -360,12 +323,6 @@ jobs:
      - name: Run priority-runtimes E2E (claude-code + hermes — skips when keys absent)
        if: needs.detect-changes.outputs.api == 'true'
        run: bash tests/e2e/test_priority_runtimes_e2e.sh
-      - name: Install standalone runtime parser from Gitea registry
-        if: needs.detect-changes.outputs.api == 'true'
-        run: |
-          python3 -m pip install --no-deps \
-            --index-url https://git.moleculesai.app/api/packages/molecule-ai/pypi/simple/ \
-            molecule-ai-workspace-runtime
      - name: Run poll-mode + since_id cursor E2E (#2339)
        if: needs.detect-changes.outputs.api == 'true'
        run: bash tests/e2e/test_poll_mode_e2e.sh
@@ -1,10 +1,8 @@
 name: E2E Chat

 # Comprehensive Playwright E2E for the unified chat stack (desktop
-# ChatTab + mobile MobileChat). Heavy browser execution is intentionally
-# outside the normal required PR path: PRs run it only after entering the
-# `merge-queue`, while push/main, nightly, and manual dispatch preserve
-# coverage without making every PR pay the full runtime/browser cost.
+# ChatTab + mobile MobileChat). Runs on every PR that touches canvas,
+# workspace-server, or this workflow file.
 #
 # Architecture:
 #   1. Ephemeral Postgres + Redis (docker, unique container names)
@@ -24,11 +22,6 @@ on:
    branches: [main, staging]
  pull_request:
    branches: [main, staging]
-  schedule:
-    # Nightly at 09:00 UTC. Keeps coverage for the currently non-required
-    # heavy browser lane without spending runner time on every PR.
-    - cron: '0 9 * * *'
-  workflow_dispatch:

 concurrency:
  group: e2e-chat-${{ github.event.pull_request.head.sha || github.sha }}
@@ -40,15 +33,9 @@ env:
 jobs:
  # bp-exempt: helper job; real gate is E2E Chat / E2E Chat (pull_request)
  detect-changes:
-    # mc#1529 follow-on: pin to `docker-host` (Linux operator-host
-    # runners). The bare `ubuntu-latest` label is also advertised by
-    # hongming-pc-runner-* (Windows act_runner v1.0.3) where the
-    # docker.sock-bound steps below fail. Mirror of mc#1543
-    # (handlers-postgres-integration). See internal#512 for the class
-    # defect.
-    runs-on: docker-host
+    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    outputs:
      chat: ${{ steps.decide.outputs.chat }}
@@ -57,14 +44,7 @@ jobs:
        with:
          fetch-depth: 0
      - id: decide
-        env:
-          GITEA_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          QUEUE_LABEL: merge-queue
        run: |
-          if [ "${{ github.event_name }}" = "schedule" ] || [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
-            echo "chat=true" >> "$GITHUB_OUTPUT"
-            exit 0
-          fi
          BASE="${GITHUB_BASE_REF:-${{ github.event.before }}}"
          if [ "${{ github.event_name }}" = "pull_request" ] && [ -n "${{ github.event.pull_request.base.sha }}" ]; then
            BASE="${{ github.event.pull_request.base.sha }}"
@@ -81,26 +61,9 @@ jobs:
            exit 0
          fi
          CHANGED=$(git diff --name-only "$BASE" HEAD)
-          if ! echo "$CHANGED" | grep -qE '^(canvas/|workspace-server/|\.gitea/workflows/e2e-chat\.yml$)'; then
-            echo "chat=false" >> "$GITHUB_OUTPUT"
-            exit 0
-          fi
-          if [ "${{ github.event_name }}" != "pull_request" ]; then
-            echo "chat=true" >> "$GITHUB_OUTPUT"
-            exit 0
-          fi
-
-          authfile=$(mktemp)
-          chmod 600 "$authfile"
-          printf 'header = "Authorization: token %s"\n' "$GITEA_TOKEN" > "$authfile"
-          labels=$(curl -fsS -K "$authfile" \
-            "${{ github.server_url }}/api/v1/repos/${{ github.repository }}/issues/${{ github.event.pull_request.number }}/labels" \
-            | python3 -c 'import json,sys; print("\n".join(label.get("name","") for label in json.load(sys.stdin)))')
-          rm -f "$authfile"
-          if printf '%s\n' "$labels" | grep -qx "$QUEUE_LABEL"; then
+          if echo "$CHANGED" | grep -qE '^(canvas/|workspace-server/|\.gitea/workflows/e2e-chat\.yml$)'; then
            echo "chat=true" >> "$GITHUB_OUTPUT"
          else
-            echo "PR is not in merge-queue; skipping heavy E2E Chat for normal PR path."
            echo "chat=false" >> "$GITHUB_OUTPUT"
          fi

@@ -108,11 +71,9 @@ jobs:
  e2e-chat:
    needs: detect-changes
    name: E2E Chat
-    # mc#1529 follow-on: docker run/exec for postgres + redis containers.
-    # Must land on operator-host Linux (docker-host).
-    runs-on: docker-host
+    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 15
    env:
@@ -136,7 +97,7 @@ jobs:
          cache-dependency-path: workspace-server/go.sum

      - if: needs.detect-changes.outputs.chat == 'true'
-        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
+        uses: actions/setup-node@60edb5dd545a775178f52524783378180af0d6f5 # v4
        with:
          node-version: '22'
          cache: 'npm'
@@ -214,19 +175,6 @@ jobs:
          echo "E2E_PLATFORM_URL=http://127.0.0.1:${PLATFORM_PORT}" >> "$GITHUB_ENV"
          echo "Platform host port: ${PLATFORM_PORT}"

-      - name: Pick canvas port
-        if: needs.detect-changes.outputs.chat == 'true'
-        run: |
-          CANVAS_PORT=$(python3 - <<'PY'
-          import socket
-          with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-              s.bind(("127.0.0.1", 0))
-              print(s.getsockname()[1])
-          PY
-          )
-          echo "CANVAS_PORT=${CANVAS_PORT}" >> "$GITHUB_ENV"
-          echo "Canvas host port: ${CANVAS_PORT}"
-
      - name: Start platform (background)
        if: needs.detect-changes.outputs.chat == 'true'
        working-directory: workspace-server
@@ -235,7 +183,6 @@ jobs:
          export DATABASE_URL="${DATABASE_URL}"
          export REDIS_URL="${REDIS_URL}"
          export PORT="${PLATFORM_PORT}"
-          export CORS_ORIGINS="http://localhost:3000,http://localhost:3001,http://localhost:${CANVAS_PORT},http://127.0.0.1:${CANVAS_PORT}"
          ./platform-server > platform.log 2>&1 &
          echo $! > platform.pid

@@ -261,14 +208,7 @@ jobs:
      - name: Install Playwright browsers
        if: needs.detect-changes.outputs.chat == 'true'
        working-directory: canvas
-        run: |
-          PREBAKED_PLAYWRIGHT=/ms-playwright
-          if [ -d "${PREBAKED_PLAYWRIGHT}" ] && find "${PREBAKED_PLAYWRIGHT}" -maxdepth 3 -type f -name 'chrome' | grep -q .; then
-            echo "Using prebaked Playwright Chromium from ${PREBAKED_PLAYWRIGHT}"
-            echo "PLAYWRIGHT_BROWSERS_PATH=${PREBAKED_PLAYWRIGHT}" >> "$GITHUB_ENV"
-            exit 0
-          fi
-          npx playwright install --with-deps chromium
+        run: npx playwright install --with-deps chromium

      - name: Start canvas dev server (background)
        if: needs.detect-changes.outputs.chat == 'true'
@@ -276,10 +216,10 @@ jobs:
        run: |
          export NEXT_PUBLIC_PLATFORM_URL="http://127.0.0.1:${PLATFORM_PORT}"
          export NEXT_PUBLIC_WS_URL="ws://127.0.0.1:${PLATFORM_PORT}/ws"
-          npx next dev --turbopack -p "${CANVAS_PORT}" > canvas.log 2>&1 &
+          npm run dev > canvas.log 2>&1 &
          echo $! > canvas.pid
          for i in $(seq 1 30); do
-            if curl -sf "http://localhost:${CANVAS_PORT}" > /dev/null 2>&1; then
+            if curl -sf http://localhost:3000 > /dev/null 2>&1; then
              echo "Canvas up after ${i}s"
              exit 0
            fi
@@ -295,7 +235,6 @@ jobs:
        run: |
          export E2E_PLATFORM_URL="http://127.0.0.1:${PLATFORM_PORT}"
          export E2E_DATABASE_URL="${DATABASE_URL}"
-          export PLAYWRIGHT_BASE_URL="http://localhost:${CANVAS_PORT}"
          npx playwright test e2e/chat-desktop.spec.ts e2e/chat-mobile.spec.ts

      - name: Dump platform log on failure
@@ -1,242 +0,0 @@
-name: E2E Legacy Advisory
-
-# Advisory lane for older/manual E2E scripts that are too broad or
-# environment-dependent for required PR CI. This intentionally does not run on
-# pull_request or push so it cannot block merges/deploys; scheduled/manual reds
-# still surface drift in scripts that would otherwise only be shellchecked.
-#
-# Gitea 1.22.6 rejects workflow_dispatch.inputs, so keep dispatch input-free.
-
-on:
-  schedule:
-    # Stagger after the staging smoke/canvas morning lanes.
-    - cron: '15 9 * * *'
-  workflow_dispatch:
-
-concurrency:
-  group: e2e-legacy-advisory
-  cancel-in-progress: false
-
-permissions:
-  contents: read
-
-env:
-  GITHUB_SERVER_URL: https://git.moleculesai.app
-
-jobs:
-  legacy-local-platform:
-    name: Legacy local-platform E2E
-    runs-on: docker-host
-    timeout-minutes: 45
-    env:
-      PG_CONTAINER: pg-e2e-legacy-${{ github.run_id }}-${{ github.run_attempt }}
-      REDIS_CONTAINER: redis-e2e-legacy-${{ github.run_id }}-${{ github.run_attempt }}
-      MOLECULE_ENV: development
-      BIND_ADDR: 127.0.0.1
-      MOLECULE_IN_DOCKER: "false"
-      A2A_TIMEOUT: "30"
-    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-
-      - uses: actions/setup-go@40f1582b2485089dde7abd97c1529aa768e1baff # v5
-        with:
-          go-version: 'stable'
-          cache: true
-          cache-dependency-path: workspace-server/go.sum
-
-      - name: Prepare local platform dependencies
-        run: |
-          set -euo pipefail
-          docker pull postgres:16 >/dev/null
-          docker pull redis:7 >/dev/null
-          docker pull alpine:latest >/dev/null
-          docker network create molecule-core-net >/dev/null 2>&1 || true
-
-      - name: Start Postgres
-        run: |
-          set -euo pipefail
-          docker rm -f "$PG_CONTAINER" 2>/dev/null || true
-          docker run -d --name "$PG_CONTAINER" \
-            -e POSTGRES_USER=dev -e POSTGRES_PASSWORD=dev -e POSTGRES_DB=molecule \
-            -p 0:5432 postgres:16 >/dev/null
-          PG_PORT=$(docker port "$PG_CONTAINER" 5432/tcp | awk -F: '/^0\.0\.0\.0:/ {print $2; exit}')
-          if [ -z "$PG_PORT" ]; then
-            PG_PORT=$(docker port "$PG_CONTAINER" 5432/tcp | head -1 | awk -F: '{print $NF}')
-          fi
-          if [ -z "$PG_PORT" ]; then
-            echo "::error::Could not resolve host port for $PG_CONTAINER"
-            docker port "$PG_CONTAINER" 5432/tcp || true
-            docker logs "$PG_CONTAINER" || true
-            exit 1
-          fi
-          echo "DATABASE_URL=postgres://dev:dev@127.0.0.1:${PG_PORT}/molecule?sslmode=disable" >> "$GITHUB_ENV"
-          for i in $(seq 1 30); do
-            docker exec "$PG_CONTAINER" pg_isready -U dev >/dev/null 2>&1 && exit 0
-            sleep 1
-          done
-          docker logs "$PG_CONTAINER" || true
-          exit 1
-
-      - name: Start Redis
-        run: |
-          set -euo pipefail
-          docker rm -f "$REDIS_CONTAINER" 2>/dev/null || true
-          docker run -d --name "$REDIS_CONTAINER" -p 0:6379 redis:7 >/dev/null
-          REDIS_PORT=$(docker port "$REDIS_CONTAINER" 6379/tcp | awk -F: '/^0\.0\.0\.0:/ {print $2; exit}')
-          if [ -z "$REDIS_PORT" ]; then
-            REDIS_PORT=$(docker port "$REDIS_CONTAINER" 6379/tcp | head -1 | awk -F: '{print $NF}')
-          fi
-          if [ -z "$REDIS_PORT" ]; then
-            echo "::error::Could not resolve host port for $REDIS_CONTAINER"
-            docker port "$REDIS_CONTAINER" 6379/tcp || true
-            docker logs "$REDIS_CONTAINER" || true
-            exit 1
-          fi
-          echo "REDIS_URL=redis://127.0.0.1:${REDIS_PORT}" >> "$GITHUB_ENV"
-          for i in $(seq 1 15); do
-            docker exec "$REDIS_CONTAINER" redis-cli ping 2>/dev/null | grep -q PONG && exit 0
-            sleep 1
-          done
-          docker logs "$REDIS_CONTAINER" || true
-          exit 1
-
-      - name: Pick platform port
-        run: |
-          set -euo pipefail
-          PLATFORM_PORT=$(python3 - <<'PY'
-          import socket
-          with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-              s.bind(("127.0.0.1", 0))
-              print(s.getsockname()[1])
-          PY
-          )
-          echo "PORT=${PLATFORM_PORT}" >> "$GITHUB_ENV"
-          echo "BASE=http://127.0.0.1:${PLATFORM_PORT}" >> "$GITHUB_ENV"
-
-      - name: Build platform
-        working-directory: workspace-server
-        run: go build -o platform-server ./cmd/server
-
-      - name: Populate template manifests for dev-mode E2E
-        run: |
-          set -euo pipefail
-          if command -v jq >/dev/null 2>&1; then
-            bash scripts/clone-manifest.sh manifest.json workspace-configs-templates org-templates plugins
-          else
-            echo "::warning::jq unavailable; dev-mode template assertion may fail if templates are absent"
-          fi
-
-      - name: Start platform
-        run: |
-          set -euo pipefail
-          ./workspace-server/platform-server > workspace-server/platform.log 2>&1 &
-          echo $! > workspace-server/platform.pid
-          for i in $(seq 1 30); do
-            curl -sf "$BASE/health" >/dev/null && exit 0
-            sleep 1
-          done
-          cat workspace-server/platform.log || true
-          exit 1
-
-      - name: Run comprehensive E2E
-        run: bash tests/e2e/test_comprehensive_e2e.sh
-
-      - name: Run workspace abilities E2E
-        run: bash tests/e2e/test_workspace_abilities_e2e.sh
-
-      - name: Run dev-mode E2E
-        run: bash tests/e2e/test_dev_mode.sh
-
-      - name: Start stub A2A agents
-        run: |
-          set -euo pipefail
-          cat > /tmp/molecule-stub-a2a.py <<'PY'
-          import json
-          from http.server import BaseHTTPRequestHandler, HTTPServer
-
-          class Handler(BaseHTTPRequestHandler):
-              def do_POST(self):
-                  length = int(self.headers.get("content-length", "0"))
-                  raw = self.rfile.read(length) if length else b"{}"
-                  try:
-                      req = json.loads(raw)
-                  except Exception:
-                      req = {}
-                  method = req.get("method")
-                  if method not in ("message/send", None):
-                      body = {"jsonrpc": "2.0", "id": req.get("id"), "error": {"code": -32601, "message": "method not found"}}
-                  else:
-                      body = {
-                          "jsonrpc": "2.0",
-                          "id": req.get("id", "stub"),
-                          "result": {
-                              "role": "agent",
-                              "parts": [{"kind": "text", "type": "text", "text": "stub agent response"}],
-                          },
-                      }
-                  data = json.dumps(body, separators=(",", ":")).encode()
-                  self.send_response(200)
-                  self.send_header("content-type", "application/json")
-                  self.send_header("content-length", str(len(data)))
-                  self.end_headers()
-                  self.wfile.write(data)
-              def log_message(self, *_):
-                  return
-
-          HTTPServer(("127.0.0.1", 18080), Handler).serve_forever()
-          PY
-          python3 /tmp/molecule-stub-a2a.py > /tmp/molecule-stub-a2a.log 2>&1 &
-          echo $! > /tmp/molecule-stub-a2a.pid
-
-      - name: Seed external agents for legacy A2A/activity scripts
-        run: |
-          set -euo pipefail
-          create_agent() {
-            local name="$1" role="$2"
-            curl -sS -X POST "$BASE/workspaces" \
-              -H "Content-Type: application/json" \
-              -d "{\"name\":\"${name}\",\"role\":\"${role}\",\"tier\":1,\"runtime\":\"external\",\"external\":true,\"url\":\"http://127.0.0.1:18080\"}" \
-              | python3 -c "import json,sys; print(json.load(sys.stdin)['id'])"
-          }
-          ECHO_ID=$(create_agent "Echo Agent" "Echo")
-          SEO_ID=$(create_agent "SEO Agent" "SEO")
-          curl -sS -X POST "$BASE/registry/register" -H "Content-Type: application/json" \
-            -d "{\"id\":\"$ECHO_ID\",\"url\":\"http://127.0.0.1:18080\",\"agent_card\":{\"name\":\"Echo Agent\",\"skills\":[{\"id\":\"echo\",\"name\":\"Echo\"}]}}" >/dev/null
-          curl -sS -X POST "$BASE/registry/register" -H "Content-Type: application/json" \
-            -d "{\"id\":\"$SEO_ID\",\"url\":\"http://127.0.0.1:18080\",\"agent_card\":{\"name\":\"SEO Agent\",\"skills\":[{\"id\":\"seo\",\"name\":\"SEO\"}]}}" >/dev/null
-
-      - name: Run activity E2E
-        run: bash tests/e2e/test_activity_e2e.sh
-
-      - name: Run A2A E2E
-        run: bash tests/e2e/test_a2a_e2e.sh
-
-      - name: Runtime-dependent legacy E2E preflight
-        run: |
-          set -euo pipefail
-          if [ -f workspace-configs-templates/claude-code-default/.auth-token ] && docker image inspect workspace:latest >/dev/null 2>&1; then
-            bash tests/e2e/test_claude_code_e2e.sh
-            bash tests/e2e/test_chat_upload_e2e.sh
-          else
-            echo "::notice::Skipping test_claude_code_e2e.sh and test_chat_upload_e2e.sh: require workspace:latest plus workspace-configs-templates/claude-code-default/.auth-token"
-          fi
-
-      - name: Dump platform log on failure
-        if: failure()
-        run: cat workspace-server/platform.log || true
-
-      - name: Stop platform and stub agents
-        if: always()
-        run: |
-          if [ -f workspace-server/platform.pid ]; then
-            kill "$(cat workspace-server/platform.pid)" 2>/dev/null || true
-          fi
-          if [ -f /tmp/molecule-stub-a2a.pid ]; then
-            kill "$(cat /tmp/molecule-stub-a2a.pid)" 2>/dev/null || true
-          fi
-
-      - name: Stop service containers
-        if: always()
-        run: |
-          docker rm -f "$PG_CONTAINER" 2>/dev/null || true
-          docker rm -f "$REDIS_CONTAINER" 2>/dev/null || true
@@ -1,403 +0,0 @@
-name: E2E Peer Visibility (literal MCP list_peers)
-
-# WHY A DEDICATED WORKFLOW (not folded into e2e-staging-saas.yml)
-# --------------------------------------------------------------
-# This is the systemic fix for a real trust failure. Hermes and OpenClaw
-# were reported "fleet-verified / cascade-complete" because the *proxy*
-# signals were green (registry registration + heartbeat for Hermes; model
-# round-trip 200 for OpenClaw). A freshly-provisioned workspace asked on
-# canvas "can you see your peers" actually FAILS:
-#   - Hermes: 401 on the molecule MCP `list_peers` call
-#   - OpenClaw: native `sessions_list` fallback, sees no platform peers
-# Tasks #142/#159 were even marked "completed" under this proxy flaw.
-#
-# A dedicated workflow (vs extending e2e-staging-saas.yml) because:
-#   - It must provision MULTIPLE distinct runtimes (hermes, openclaw,
-#     claude-code) in ONE org and assert each sees the others. The
-#     full-saas script is single-runtime-per-run (E2E_RUNTIME) and folding
-#     a multi-runtime matrix into it would conflate concerns and bloat its
-#     already-45-min run.
-#   - It needs its own concurrency group so it doesn't fight full-saas /
-#     canvas for the staging org-creation quota.
-#   - It needs an independent, non-required status-context name so it can
-#     be RED today (the in-flight Hermes-401 / OpenClaw-MCP-wiring fixes
-#     have not landed) WITHOUT wedging unrelated merges — and flipped to
-#     REQUIRED in one branch-protection edit once it goes green
-#     (flip-to-required checklist: molecule-core#1296).
-#
-# THE ASSERTION IS NOT A PROXY. The driving script
-# tests/e2e/test_peer_visibility_mcp_staging.sh issues the byte-for-byte
-# JSON-RPC `tools/call name=list_peers` envelope to `POST
-# /workspaces/:id/mcp` using each workspace's OWN bearer token, through
-# the real WorkspaceAuth + MCPRateLimiter middleware chain — the exact
-# call mcp_molecule_list_peers makes from a canvas agent. It does NOT
-# read a registry row, /health, the heartbeat table, or
-# GET /registry/:id/peers.
-#
-# HONEST GATE — NO continue-on-error. Per feedback_fix_root_not_symptom a
-# fake-green mask would defeat the entire purpose. This workflow goes red
-# on today's broken behavior and green only when the root-cause fixes
-# actually land. It is intentionally NOT in branch_protections — see PR
-# body for the required-vs-not decision + flip tracking issue.
-#
-# Gitea 1.22.6 / act_runner notes honored:
-#   - No cross-repo `uses:` (feedback_gitea_cross_repo_uses_blocked). The
-#     actions/checkout SHA is the one e2e-staging-canvas.yml already uses
-#     successfully (a mirrored SHA — see #1277/PR#1292 root-cause).
-#   - 2026-05-21 retrigger: verify fresh platform-tenant image after the
-#     publish Buildx DOCKER_CONFIG fix restored staging-latest image updates.
-#   - Per-SHA concurrency, not global (feedback_concurrency_group_per_sha).
-#   - Workflow-level GITHUB_SERVER_URL pinned
-#     (feedback_act_runner_github_server_url).
-#   - pr-validate posts a status under the same check name so a
-#     workflow-only PR is not silently statusless and the context is
-#     flip-to-required-ready (mirrors e2e-staging-saas.yml's proven shape;
-#     real EC2-provisioning E2E is push/dispatch/cron only — it is 30+ min
-#     and cannot run per-PR-update).
-#
-# LOCAL BACKEND (added 2026-05-15 — feedback_local_must_mimic_production,
-# feedback_mandatory_local_e2e_before_ship, feedback_local_test_before_
-# staging_e2e)
-# --------------------------------------------------------------------
-# The standing rule is that the local prod-mimic stack runs a MANDATORY
-# local-Postgres E2E BEFORE staging E2E. A staging-only peer-visibility
-# gate caught regressions late + expensively (cold EC2). The
-# `peer-visibility-local` job below runs the SAME byte-identical
-# assertion (tests/e2e/lib/peer_visibility_assert.sh) against the local
-# docker-compose stack — built + booted exactly like e2e-api.yml's
-# proven E2E API Smoke Test job (ephemeral pg/redis ports, go build,
-# background platform-server). It runs on PR + push (local boot is
-# minutes, not the 30+ min cold-EC2 path), so peer-visibility is part of
-# the local gate that fires before the staging E2E.
-#
-# It is its OWN non-required status context `E2E Peer Visibility (local)`.
-# The local backend uses external-mode workspaces by default so it tests
-# the literal platform MCP list_peers path without depending on local
-# template container boot/heartbeat. Container-mode runtime boot remains
-# available via PV_LOCAL_PROVISION_MODE=container for targeted debugging.
-
-on:
-  push:
-    branches: [main]
-    paths:
-      - 'workspace-server/internal/handlers/mcp.go'
-      - 'workspace-server/internal/handlers/mcp_tools.go'
-      - 'workspace-server/internal/middleware/**'
-      - 'workspace-server/internal/handlers/registry.go'
-      - 'workspace-server/internal/handlers/workspace.go'
-      - 'tests/e2e/test_peer_visibility_mcp_staging.sh'
-      - 'tests/e2e/test_peer_visibility_token_mint_staging.sh'
-      - 'tests/e2e/test_peer_visibility_mcp_local.sh'
-      - 'tests/e2e/lib/peer_visibility_assert.sh'
-      - '.gitea/workflows/e2e-peer-visibility.yml'
-  pull_request:
-    branches: [main]
-    paths:
-      - 'workspace-server/internal/handlers/mcp.go'
-      - 'workspace-server/internal/handlers/mcp_tools.go'
-      - 'workspace-server/internal/middleware/**'
-      - 'workspace-server/internal/handlers/registry.go'
-      - 'workspace-server/internal/handlers/workspace.go'
-      - 'tests/e2e/test_peer_visibility_mcp_staging.sh'
-      - 'tests/e2e/test_peer_visibility_token_mint_staging.sh'
-      - 'tests/e2e/test_peer_visibility_mcp_local.sh'
-      - 'tests/e2e/lib/peer_visibility_assert.sh'
-      - '.gitea/workflows/e2e-peer-visibility.yml'
-  workflow_dispatch:
-  schedule:
-    # 07:30 UTC daily — catches AMI / template-hermes / template-openclaw
-    # drift even on quiet days. Offset 30m from e2e-staging-saas (07:00)
-    # so the two don't collide on the staging org-creation quota.
-    - cron: '30 7 * * *'
-
-concurrency:
-  # Per-SHA (feedback_concurrency_group_per_sha). A single global group
-  # would let a queued staging/main push behind a PR run get cancelled,
-  # leaving any gate that reads "completed run at SHA" stuck.
-  group: e2e-peer-visibility-${{ github.event.pull_request.head.sha || github.sha }}
-  cancel-in-progress: false
-
-env:
-  GITHUB_SERVER_URL: https://git.moleculesai.app
-
-jobs:
-  # PR path: post a real status under the required-ready check name so a
-  # workflow-only PR is never silently statusless. The actual EC2 E2E is
-  # push/dispatch/cron only (30+ min). This is NOT a fake-green mask of
-  # the real assertion — it validates the driving script's bash syntax
-  # and inline-python so a broken test script fails at PR time.
-  pr-validate:
-    name: E2E Peer Visibility
-    runs-on: ubuntu-latest
-    if: github.event_name == 'pull_request'
-    timeout-minutes: 5
-    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-      - name: Validate driving scripts + shared assertion lib
-        run: |
-          bash -n tests/e2e/lib/peer_visibility_assert.sh
-          echo "lib/peer_visibility_assert.sh — bash syntax OK"
-          bash -n tests/e2e/test_peer_visibility_mcp_staging.sh
-          echo "test_peer_visibility_mcp_staging.sh — bash syntax OK"
-          bash -n tests/e2e/test_peer_visibility_token_mint_staging.sh
-          echo "test_peer_visibility_token_mint_staging.sh — bash syntax OK"
-          bash -n tests/e2e/test_peer_visibility_mcp_local.sh
-          echo "test_peer_visibility_mcp_local.sh — bash syntax OK"
-          legacy_token_suffix="test""-token"
-          if rg -n "$legacy_token_suffix" tests/e2e/test_*staging*.sh; then
-            echo "::error::staging E2E must use production-safe admin token minting"
-            exit 1
-          fi
-          echo "Staging fresh-provision MCP list_peers E2E runs on push to"
-          echo "main / workflow_dispatch / daily cron (30+ min EC2 boot)."
-          echo "The LOCAL backend runs in the peer-visibility-local job"
-          echo "below on this same PR (local docker-compose stack)."
-
-  # LOCAL gate: same byte-identical assertion against the local prod-mimic
-  # docker-compose stack — the MANDATORY local-E2E that must run BEFORE
-  # the staging E2E (feedback_mandatory_local_e2e_before_ship,
-  # feedback_local_test_before_staging_e2e). Bootstrap mirrors
-  # e2e-api.yml's proven E2E API Smoke Test job (per-run container names +
-  # ephemeral host ports so concurrent host-network act_runner runs don't
-  # collide; go build; background platform-server). Its OWN non-required
-  # status context `E2E Peer Visibility (local)` — non-required-by-design
-  # exactly like the staging job (flip-to-required tracked at
-  # molecule-core#1296). HONEST gate, NO continue-on-error mask
-  # (feedback_fix_root_not_symptom). Runs on PR +
-  # push (local boot is minutes, not the 30+ min cold-EC2 path).
-  # bp-required: pending #1296
-  peer-visibility-local:
-    name: E2E Peer Visibility (local)
-    runs-on: docker-host
-    timeout-minutes: 30
-    env:
-      # Per-run names + ephemeral ports — same collision-avoidance as
-      # e2e-api.yml (host-network act_runner; feedback_act_runner_*).
-      PG_CONTAINER: pg-e2e-pv-${{ github.run_id }}-${{ github.run_attempt }}
-      REDIS_CONTAINER: redis-e2e-pv-${{ github.run_id }}-${{ github.run_attempt }}
-      # LLM keys so hermes/openclaw can actually boot. The local script
-      # SKIPs (not fails) any runtime whose key is absent, so a partially
-      # keyed CI env still exercises whatever it can.
-      CLAUDE_CODE_OAUTH_TOKEN: ${{ secrets.E2E_CLAUDE_CODE_OAUTH_TOKEN }}
-      E2E_MINIMAX_API_KEY: ${{ secrets.MOLECULE_STAGING_MINIMAX_API_KEY }}
-      E2E_ANTHROPIC_API_KEY: ${{ secrets.MOLECULE_STAGING_ANTHROPIC_API_KEY }}
-      E2E_OPENAI_API_KEY: ${{ secrets.MOLECULE_STAGING_OPENAI_API_KEY }}
-      PV_RUNTIMES: "hermes openclaw claude-code"
-      PV_LOCAL_PROVISION_MODE: external
-      ADMIN_TOKEN: local-e2e-admin-token
-      MOLECULE_ADMIN_TOKEN: local-e2e-admin-token
-    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-      - uses: actions/setup-go@40f1582b2485089dde7abd97c1529aa768e1baff # v5
-        with:
-          go-version: 'stable'
-          cache: true
-          cache-dependency-path: workspace-server/go.sum
-      - name: Pre-pull alpine + ensure provisioner network
-        run: |
-          docker pull alpine:latest >/dev/null
-          docker network create molecule-core-net >/dev/null 2>&1 || true
-          echo "alpine:latest pre-pulled; molecule-core-net ensured."
-      - name: Start Postgres (docker, ephemeral port)
-        run: |
-          docker rm -f "$PG_CONTAINER" 2>/dev/null || true
-          docker run -d --name "$PG_CONTAINER" \
-            -e POSTGRES_USER=dev -e POSTGRES_PASSWORD=dev -e POSTGRES_DB=molecule \
-            -p 0:5432 postgres:16 >/dev/null
-          PG_PORT=$(docker port "$PG_CONTAINER" 5432/tcp | awk -F: '/^0\.0\.0\.0:/ {print $2; exit}')
-          [ -n "$PG_PORT" ] || PG_PORT=$(docker port "$PG_CONTAINER" 5432/tcp | head -1 | awk -F: '{print $NF}')
-          if [ -z "$PG_PORT" ]; then
-            echo "::error::Could not resolve host port for $PG_CONTAINER"
-            docker logs "$PG_CONTAINER" || true; exit 1
-          fi
-          echo "DATABASE_URL=postgres://dev:dev@127.0.0.1:${PG_PORT}/molecule?sslmode=disable" >> "$GITHUB_ENV"
-          for i in $(seq 1 30); do
-            docker exec "$PG_CONTAINER" pg_isready -U dev >/dev/null 2>&1 && { echo "Postgres ready after ${i}s"; exit 0; }
-            sleep 1
-          done
-          echo "::error::Postgres did not become ready in 30s"; docker logs "$PG_CONTAINER" || true; exit 1
-      - name: Start Redis (docker, ephemeral port)
-        run: |
-          docker rm -f "$REDIS_CONTAINER" 2>/dev/null || true
-          docker run -d --name "$REDIS_CONTAINER" -p 0:6379 redis:7 >/dev/null
-          REDIS_PORT=$(docker port "$REDIS_CONTAINER" 6379/tcp | awk -F: '/^0\.0\.0\.0:/ {print $2; exit}')
-          [ -n "$REDIS_PORT" ] || REDIS_PORT=$(docker port "$REDIS_CONTAINER" 6379/tcp | head -1 | awk -F: '{print $NF}')
-          if [ -z "$REDIS_PORT" ]; then
-            echo "::error::Could not resolve host port for $REDIS_CONTAINER"
-            docker logs "$REDIS_CONTAINER" || true; exit 1
-          fi
-          echo "REDIS_URL=redis://127.0.0.1:${REDIS_PORT}" >> "$GITHUB_ENV"
-          for i in $(seq 1 15); do
-            docker exec "$REDIS_CONTAINER" redis-cli ping 2>/dev/null | grep -q PONG && { echo "Redis ready after ${i}s"; exit 0; }
-            sleep 1
-          done
-          echo "::error::Redis did not become ready in 15s"; docker logs "$REDIS_CONTAINER" || true; exit 1
-      - name: Build platform
-        working-directory: workspace-server
-        run: go build -o platform-server ./cmd/server
-      - name: Pick platform port
-        run: |
-          PLATFORM_PORT=$(python3 - <<'PY'
-          import socket
-          with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-              s.bind(("127.0.0.1", 0))
-              print(s.getsockname()[1])
-          PY
-          )
-          echo "PORT=${PLATFORM_PORT}" >> "$GITHUB_ENV"
-          echo "BASE=http://127.0.0.1:${PLATFORM_PORT}" >> "$GITHUB_ENV"
-          echo "Platform host port: ${PLATFORM_PORT}"
-      - name: Kill stale platform-server before start
-        run: |
-          killed=0
-          for pid in $(grep -l "platform-serve" /proc/[0-9]*/comm 2>/dev/null); do
-            kpid="${pid%/comm}"; kpid="${kpid##*/}"
-            cmdline=$(cat "/proc/${kpid}/cmdline" 2>/dev/null | tr '\0' ' ')
-            if echo "$cmdline" | grep -q "platform-server"; then
-              echo "Killing stale platform-server pid ${kpid}"
-              kill "$kpid" 2>/dev/null || true; killed=$((killed + 1))
-            fi
-          done
-          [ "$killed" -gt 0 ] && sleep 2 || true
-          echo "stale-kill done ($killed killed)"
-      - name: Start platform (background)
-        working-directory: workspace-server
-        run: |
-          ./platform-server > platform.log 2>&1 &
-          echo $! > platform.pid
-      - name: Wait for /health
-        run: |
-          for i in $(seq 1 30); do
-            curl -sf "$BASE/health" > /dev/null && { echo "Platform up after ${i}s"; exit 0; }
-            sleep 1
-          done
-          echo "::error::Platform did not become healthy in 30s"
-          cat workspace-server/platform.log || true; exit 1
-      - name: Run LOCAL fresh-provision peer-visibility E2E (literal MCP list_peers)
-        # HONEST gate — NO continue-on-error. The local backend uses
-        # external-mode workspaces so this context tests the literal MCP
-        # peer-visibility path without coupling to template container boot.
-        run: bash tests/e2e/test_peer_visibility_mcp_local.sh
-      - name: Dump platform log on failure
-        if: failure()
-        run: cat workspace-server/platform.log || true
-      - name: Stop platform
-        if: always()
-        run: |
-          if [ -f workspace-server/platform.pid ]; then
-            kill "$(cat workspace-server/platform.pid)" 2>/dev/null || true
-          fi
-      - name: Stop service containers
-        if: always()
-        run: |
-          docker rm -f "$PG_CONTAINER" 2>/dev/null || true
-          docker rm -f "$REDIS_CONTAINER" 2>/dev/null || true
-
-  # Real STAGING gate: provisions a throwaway org + sibling-per-runtime,
-  # drives the LITERAL list_peers MCP call per runtime, asserts 200 +
-  # expected peer set, then scoped teardown. push(main)/dispatch/cron only.
-  peer-visibility:
-    name: E2E Peer Visibility
-    runs-on: ubuntu-latest
-    if: github.event_name != 'pull_request'
-    timeout-minutes: 60
-
-    env:
-      MOLECULE_CP_URL: https://staging-api.moleculesai.app
-      MOLECULE_ADMIN_TOKEN: ${{ secrets.CP_STAGING_ADMIN_API_TOKEN }}
-      # LLM provider key so each runtime can authenticate at boot.
-      # Priority MiniMax → direct-Anthropic → OpenAI matches
-      # test_staging_full_saas.sh's secrets-injection chain.
-      E2E_MINIMAX_API_KEY: ${{ secrets.MOLECULE_STAGING_MINIMAX_API_KEY }}
-      E2E_ANTHROPIC_API_KEY: ${{ secrets.MOLECULE_STAGING_ANTHROPIC_API_KEY }}
-      E2E_OPENAI_API_KEY: ${{ secrets.MOLECULE_STAGING_OPENAI_API_KEY }}
-      E2E_RUN_ID: "${{ github.run_id }}-${{ github.run_attempt }}"
-      PV_RUNTIMES: "hermes openclaw claude-code"
-
-    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-
-      - name: Verify admin token present
-        run: |
-          if [ -z "$MOLECULE_ADMIN_TOKEN" ]; then
-            echo "::error::CP_STAGING_ADMIN_API_TOKEN secret not set (Railway staging CP_ADMIN_API_TOKEN)"
-            exit 2
-          fi
-          echo "Admin token present"
-
-      - name: Verify an LLM key present
-        run: |
-          if [ -z "${E2E_MINIMAX_API_KEY:-}" ] && [ -z "${E2E_ANTHROPIC_API_KEY:-}" ] && [ -z "${E2E_OPENAI_API_KEY:-}" ]; then
-            echo "::error::No LLM provider key set — workspaces fail at boot with 'No provider API key found'. Set MOLECULE_STAGING_MINIMAX_API_KEY (or ANTHROPIC / OPENAI)."
-            exit 2
-          fi
-          echo "LLM key present"
-
-      - name: CP staging health preflight
-        run: |
-          code=$(curl -sS -o /dev/null -w "%{http_code}" --max-time 10 "$MOLECULE_CP_URL/health")
-          if [ "$code" != "200" ]; then
-            echo "::error::Staging CP unhealthy (HTTP $code) — infra, not a workspace bug. Failing loud per feedback_fix_root_not_symptom."
-            exit 1
-          fi
-          echo "Staging CP healthy"
-
-      - name: Run fresh-provision peer-visibility E2E (literal MCP list_peers)
-        run: bash tests/e2e/test_peer_visibility_mcp_staging.sh
-
-      # Belt-and-braces scoped teardown: the script installs an EXIT/INT/
-      # TERM trap, but if the runner itself is cancelled the trap may not
-      # fire. This always() step deletes ONLY the e2e-pv-<run_id> org this
-      # run created — never a cluster-wide sweep
-      # (feedback_never_run_cluster_cleanup_tests_on_live_platform). The
-      # admin DELETE is idempotent so double-invoking is safe;
-      # sweep-stale-e2e-orgs is the final net (slug starts with 'e2e-').
-      - name: Teardown safety net (runs on cancel/failure)
-        if: always()
-        env:
-          ADMIN_TOKEN: ${{ secrets.CP_STAGING_ADMIN_API_TOKEN }}
-        run: |
-          set +e
-          orgs=$(curl -sS "$MOLECULE_CP_URL/cp/admin/orgs?limit=500" \
-            -H "Authorization: Bearer $ADMIN_TOKEN" 2>/dev/null \
-            | python3 -c "
-          import json, sys, os, datetime
-          run_id = os.environ.get('GITHUB_RUN_ID', '')
-          try:
-              d = json.load(sys.stdin)
-          except Exception:
-              print(''); sys.exit(0)
-          # ONLY sweep slugs from THIS run. e2e-pv-<YYYYMMDD>-<run_id>-...
-          # Sweep today AND yesterday's UTC date so a midnight-crossing run
-          # still matches its own slug (same bug class as the saas/canvas
-          # safety nets).
-          today = datetime.date.today()
-          yest = today - datetime.timedelta(days=1)
-          dates = (today.strftime('%Y%m%d'), yest.strftime('%Y%m%d'))
-          if run_id:
-              prefixes = tuple(f'e2e-pv-{dt}-{run_id}-' for dt in dates)
-          else:
-              prefixes = tuple(f'e2e-pv-{dt}-' for dt in dates)
-          orgs = d if isinstance(d, list) else d.get('orgs', [])
-          cands = [o['slug'] for o in orgs
-                   if any(o.get('slug','').startswith(p) for p in prefixes)
-                   and o.get('instance_status') not in ('purged',)]
-          print('\n'.join(cands))
-          " 2>/dev/null)
-          for slug in $orgs; do
-            echo "Safety-net teardown: $slug"
-            set +e
-            curl -sS -o /tmp/pv-cleanup.out -w "%{http_code}" \
-              -X DELETE "$MOLECULE_CP_URL/cp/admin/tenants/$slug" \
-              -H "Authorization: Bearer $ADMIN_TOKEN" \
-              -H "Content-Type: application/json" \
-              -d "{\"confirm\":\"$slug\"}" >/tmp/pv-cleanup.code
-            set -e
-            code=$(cat /tmp/pv-cleanup.code 2>/dev/null || echo "000")
-            if [ "$code" = "200" ] || [ "$code" = "204" ]; then
-              echo "[teardown] deleted $slug (HTTP $code)"
-            else
-              echo "::warning::pv teardown for $slug returned HTTP $code — sweep-stale-e2e-orgs will catch it within MAX_AGE_MINUTES. Body: $(head -c 300 /tmp/pv-cleanup.out 2>/dev/null)"
-            fi
-          done
-          exit 0
@@ -16,9 +16,9 @@ name: E2E Staging Canvas (Playwright)
 # e2e-staging-saas.yml (which tests the API shape) by exercising the
 # actual browser + canvas bundle against live staging.
 #
-# Triggers: push to main, PR touching canvas sources + this workflow only
-# after the PR enters `merge-queue`, manual dispatch, and scheduled cron to
-# catch browser/runtime drift even when canvas is quiet.
+# Triggers: push to main/staging or PR touching canvas sources + this workflow,
+# manual dispatch, and weekly cron to catch browser/runtime drift even
+# when canvas is quiet.
 # Added staging to push/pull_request branches so the auto-promote gate
 # check (--event push --branch staging) can see a completed run for this
 # workflow — mirrors what PR #1891 does for e2e-api.yml.
@@ -37,10 +37,9 @@ on:
  pull_request:
    branches: [main]
  schedule:
-    # Nightly at 08:00 UTC — catches Chrome / Playwright / Next.js
+    # Weekly on Sunday 08:00 UTC — catches Chrome / Playwright / Next.js
    # release-note-shaped regressions that don't ride in with a PR.
-    - cron: '0 8 * * *'
-  workflow_dispatch:
+    - cron: '0 8 * * 0'

 concurrency:
  # Per-SHA grouping (changed 2026-04-28 from a single global group). The
@@ -71,7 +70,7 @@ jobs:
  detect-changes:
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    outputs:
      canvas: ${{ steps.decide.outputs.canvas }}
@@ -80,13 +79,10 @@ jobs:
        with:
          fetch-depth: 0
      - id: decide
-        env:
-          GITEA_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          QUEUE_LABEL: merge-queue
        # Inline replacement for dorny/paths-filter — see e2e-api.yml.
-        # Cron and manual triggers always run real work (no diff context).
+        # Cron triggers always run real work (no diff context).
        run: |
-          if [ "${{ github.event_name }}" = "schedule" ] || [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
+          if [ "${{ github.event_name }}" = "schedule" ]; then
            echo "canvas=true" >> "$GITHUB_OUTPUT"
            exit 0
          fi
@@ -106,26 +102,9 @@ jobs:
            exit 0
          fi
          CHANGED=$(git diff --name-only "$BASE" HEAD)
-          if ! echo "$CHANGED" | grep -qE '^(canvas/|\.gitea/workflows/e2e-staging-canvas\.yml$)'; then
-            echo "canvas=false" >> "$GITHUB_OUTPUT"
-            exit 0
-          fi
-          if [ "${{ github.event_name }}" != "pull_request" ]; then
-            echo "canvas=true" >> "$GITHUB_OUTPUT"
-            exit 0
-          fi
-
-          authfile=$(mktemp)
-          chmod 600 "$authfile"
-          printf 'header = "Authorization: token %s"\n' "$GITEA_TOKEN" > "$authfile"
-          labels=$(curl -fsS -K "$authfile" \
-            "${{ github.server_url }}/api/v1/repos/${{ github.repository }}/issues/${{ github.event.pull_request.number }}/labels" \
-            | python3 -c 'import json,sys; print("\n".join(label.get("name","") for label in json.load(sys.stdin)))')
-          rm -f "$authfile"
-          if printf '%s\n' "$labels" | grep -qx "$QUEUE_LABEL"; then
+          if echo "$CHANGED" | grep -qE '^(canvas/|\.gitea/workflows/e2e-staging-canvas\.yml$)'; then
            echo "canvas=true" >> "$GITHUB_OUTPUT"
          else
-            echo "PR is not in merge-queue; skipping heavy E2E Staging Canvas for normal PR path."
            echo "canvas=false" >> "$GITHUB_OUTPUT"
          fi

@@ -140,7 +119,7 @@ jobs:
    name: Canvas tabs E2E
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 40

@@ -190,14 +169,7 @@ jobs:
      - name: Install Playwright browsers
        if: needs.detect-changes.outputs.canvas == 'true'
        timeout-minutes: 10
-        run: |
-          PREBAKED_PLAYWRIGHT=/ms-playwright
-          if [ -d "${PREBAKED_PLAYWRIGHT}" ] && find "${PREBAKED_PLAYWRIGHT}" -maxdepth 3 -type f -name 'chrome' | grep -q .; then
-            echo "Using prebaked Playwright Chromium from ${PREBAKED_PLAYWRIGHT}"
-            echo "PLAYWRIGHT_BROWSERS_PATH=${PREBAKED_PLAYWRIGHT}" >> "$GITHUB_ENV"
-            exit 0
-          fi
-          npx playwright install --with-deps chromium
+        run: npx playwright install --with-deps chromium

      - name: Run staging canvas E2E
        if: needs.detect-changes.outputs.canvas == 'true'
@@ -84,7 +84,7 @@ jobs:
    name: E2E Staging External Runtime
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 25

@@ -49,9 +49,6 @@ on:
      - 'workspace-server/internal/middleware/**'
      - 'workspace-server/internal/provisioner/**'
      - 'tests/e2e/test_staging_full_saas.sh'
-      - 'tests/e2e/lib/completion_assert.sh'
-      - 'tests/e2e/lib/aws_leak_check.sh'
-      - 'tests/e2e/test_aws_leak_check.sh'
      - '.gitea/workflows/e2e-staging-saas.yml'
  pull_request:
    branches: [main]
@@ -62,9 +59,6 @@ on:
      - 'workspace-server/internal/middleware/**'
      - 'workspace-server/internal/provisioner/**'
      - 'tests/e2e/test_staging_full_saas.sh'
-      - 'tests/e2e/lib/completion_assert.sh'
-      - 'tests/e2e/lib/aws_leak_check.sh'
-      - 'tests/e2e/test_aws_leak_check.sh'
      - '.gitea/workflows/e2e-staging-saas.yml'
  workflow_dispatch:
  schedule:
@@ -94,31 +88,31 @@ jobs:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          fetch-depth: 1
-        # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+        # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
        continue-on-error: true

      - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
        with:
          python-version: "3.11"
-        # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+        # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
        continue-on-error: true

      - name: YAML validation (best-effort)
        run: |
          echo "e2e-staging-saas.yml — PR validation: workflow YAML is valid."
          echo "E2E step runs only when provisioning-critical files change."
-        # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+        # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
        continue-on-error: true

-  # Actual E2E: runs on trunk pushes and PRs that touch provisioning-critical
-  # paths. pr-validate remains as the lightweight workflow-shape check for PRs,
-  # but it is not a substitute for live staging proof when this workflow or the
-  # staging harness changes.
+  # Actual E2E: runs on trunk pushes (main + staging). NOT the PR-fire-only
+  # path — pr-validate above posts success for workflow-only PRs.
  e2e-staging-saas:
    name: E2E Staging SaaS
    runs-on: ubuntu-latest
+    # Only runs on trunk pushes. PR paths get pr-validate instead.
+    if: github.event.pull_request.base.ref == ''
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 45
    permissions:
@@ -133,11 +127,6 @@ jobs:
      # (dead in org secret store) to CP_STAGING_ADMIN_API_TOKEN per
      # internal#322 — see this PR for the cross-workflow sweep.
      MOLECULE_ADMIN_TOKEN: ${{ secrets.CP_STAGING_ADMIN_API_TOKEN }}
-      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
-      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
-      AWS_DEFAULT_REGION: us-east-2
-      E2E_AWS_LEAK_CHECK: required
-      E2E_AWS_TERMINATE_LEAKS: '1'
      # MiniMax is the PRIMARY LLM auth path post-2026-05-04. Switched
      # from hermes+OpenAI default after #2578 (the staging OpenAI key
      # account went over quota and stayed dead for 36+ hours, taking
@@ -154,21 +143,16 @@ jobs:
      # block). See #2578 PR comment for the rationale.
      E2E_ANTHROPIC_API_KEY: ${{ secrets.MOLECULE_STAGING_ANTHROPIC_API_KEY }}
      # OpenAI fallback — kept wired so an operator-dispatched run with
-      # E2E_RUNTIME=hermes or =codex via workflow_dispatch can still
+      # E2E_RUNTIME=hermes or =langgraph via workflow_dispatch can still
      # exercise the OpenAI path.
      E2E_OPENAI_API_KEY: ${{ secrets.MOLECULE_STAGING_OPENAI_API_KEY }}
-      # google-adk (operator-dispatched only) auths Gemini with an
-      # AI-Studio key. Org policy disallows API keys in PROD (Vertex+ADC
-      # there); CI uses the keyed AI-Studio path with config model
-      # google_genai:gemini-2.5-pro. Vertex remains the supported prod path.
-      E2E_GOOGLE_API_KEY: ${{ secrets.MOLECULE_STAGING_GOOGLE_API_KEY }}
      E2E_RUNTIME: ${{ github.event.inputs.runtime || 'claude-code' }}
      # Pin the model when running on the default claude-code path —
      # the per-runtime default ("sonnet") routes to direct Anthropic
      # and defeats the cost saving. Operators can override via the
      # workflow_dispatch flow (no input wired here yet — runtime
      # override is enough for ad-hoc).
-      E2E_MODEL_SLUG: ${{ github.event.inputs.runtime == 'hermes' && 'openai/gpt-4o' || github.event.inputs.runtime == 'codex' && 'openai/gpt-4o' || github.event.inputs.runtime == 'google-adk' && 'google_genai:gemini-2.5-pro' || 'MiniMax-M2' }}
+      E2E_MODEL_SLUG: ${{ github.event.inputs.runtime == 'hermes' && 'openai/gpt-4o' || github.event.inputs.runtime == 'langgraph' && 'openai:gpt-4o' || 'MiniMax-M2.7-highspeed' }}
      E2E_RUN_ID: "${{ github.run_id }}-${{ github.run_attempt }}"
      E2E_KEEP_ORG: ${{ github.event.inputs.keep_org && '1' || '0' }}

@@ -181,18 +165,12 @@ jobs:
            echo "::error::CP_STAGING_ADMIN_API_TOKEN secret not set (Railway staging CP_ADMIN_API_TOKEN)"
            exit 2
          fi
-          for var in AWS_ACCESS_KEY_ID AWS_SECRET_ACCESS_KEY; do
-            if [ -z "${!var:-}" ]; then
-              echo "::error::$var not set — EC2 leak verification cannot run"
-              exit 2
-            fi
-          done
          echo "Admin token present ✓"

      - name: Verify LLM key present
        run: |
          # Per-runtime key check — claude-code uses MiniMax; hermes /
-          # codex (operator-dispatched only) use OpenAI. Hard-fail
+          # langgraph (operator-dispatched only) use OpenAI. Hard-fail
          # rather than soft-skip per #2578's lesson — empty key
          # silently falls through to the wrong SECRETS_JSON branch and
          # produces a confusing auth error 5 min later instead of the
@@ -213,14 +191,10 @@ jobs:
                required_secret_value=""
              fi
              ;;
-            codex|hermes)
+            langgraph|hermes)
              required_secret_name="MOLECULE_STAGING_OPENAI_API_KEY"
              required_secret_value="${E2E_OPENAI_API_KEY:-}"
              ;;
-            google-adk)
-              required_secret_name="MOLECULE_STAGING_GOOGLE_API_KEY"
-              required_secret_value="${E2E_GOOGLE_API_KEY:-}"
-              ;;
            *)
              echo "::warning::Unknown E2E_RUNTIME='${E2E_RUNTIME}' — skipping LLM-key check"
              required_secret_name=""
@@ -37,7 +37,7 @@ jobs:
    name: Intentional-failure teardown sanity
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 20

@@ -47,11 +47,6 @@ jobs:
      # (dead in org secret store) to CP_STAGING_ADMIN_API_TOKEN per
      # internal#322 — see this PR for the cross-workflow sweep.
      MOLECULE_ADMIN_TOKEN: ${{ secrets.CP_STAGING_ADMIN_API_TOKEN }}
-      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
-      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
-      AWS_DEFAULT_REGION: us-east-2
-      E2E_AWS_LEAK_CHECK: required
-      E2E_AWS_TERMINATE_LEAKS: '1'
      E2E_MODE: smoke
      E2E_RUNTIME: hermes
      E2E_RUN_ID: "sanity-${{ github.run_id }}"
@@ -66,12 +61,6 @@ jobs:
            echo "::error::CP_STAGING_ADMIN_API_TOKEN not set"
            exit 2
          fi
-          for var in AWS_ACCESS_KEY_ID AWS_SECRET_ACCESS_KEY; do
-            if [ -z "${!var:-}" ]; then
-              echo "::error::$var not set — EC2 leak verification cannot run"
-              exit 2
-            fi
-          done

      # Inverted assertion: the run MUST fail. If it passes, the
      # E2E_INTENTIONAL_FAILURE path is broken.
@@ -7,11 +7,10 @@
 #   PR_NUMBER  — set via ${{ github.event.pull_request.number }} from the trigger
 #   POST_COMMENT — "true" to post/update comment on PR
 #
-# Gating logic (MVP signals 1,2,3,4,6):
+# Gating logic (MVP signals 1,2,3,6):
 #   1. Author-aware agent-tag comment scan
 #   2. REQUEST_CHANGES reviews state machine
 #   3. Staleness detection (SOP-12: review.commit_id != PR.head_sha + >1 working day)
-#   4. Branch divergence / scope-creep guard (base-sha vs target HEAD; mc#365)
 #   6. CI required-checks awareness
 #
 # Exit code: 0=CLEAR, 1=BLOCKED, 2=ERROR
@@ -33,24 +32,6 @@ on:
  # iterating all open PRs when PR_NUMBER is empty.
  workflow_dispatch:

-# Serialize per PR (or per repo for schedule/manual ticks) to prevent
-# the fan-out OOM class documented in
-# `reference_operator_host_python3_oom_storm_2026_05_18`. `edited`
-# events fan out on every PR-body edit; combined with the hourly cron
-# and synchronize bursts this workflow can stack runs of the same
-# workflow_id on the same PR (each ~4GB anon-RSS) and trip the
-# `--memory=4g --memory-swap=8g` per-container cap.
-#
-# NO `cancel-in-progress` (defaults to false). Per
-# `feedback_janitor_supersede_must_group_by_workflow_id`, cancelling
-# in-flight runs of any required-check-shaped workflow risks the
-# dismiss_stale_approvals + empty-commit-rerun dance (Gitea 1.22.6 has
-# no REST rerun). The gate-check is `continue-on-error: true` +
-# idempotent (POST/PATCH gate-check comment by context) so sequential
-# ticks are strictly safe.
-concurrency:
-  group: gate-check-v3-${{ github.event.pull_request.number || github.event.issue.number || github.ref }}
-
 permissions:
  # read: contents — for checkout (base ref, not PR head for security)
  # read: pull-requests — for reading PR info via API
@@ -66,7 +47,7 @@ jobs:
  # bp-exempt: PR advisory bot; merge blocking is enforced by CI status and branch protection.
  gate-check:
    runs-on: ubuntu-latest
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true  # Never block on our own detector failing
    steps:
      - name: Check out BASE ref (never PR-head under pull_request_target)
@@ -102,41 +83,25 @@ jobs:
          REPO: ${{ github.repository }}
        run: |
          set -euo pipefail
-          # Fetch all open PRs and run gate-check on each. This scheduled
-          # refresher is advisory; a transient Gitea list timeout must not turn
-          # main red. PR-specific gate-check runs still use normal failure
-          # semantics.
+          # Fetch all open PRs and run gate-check on each
+          # socket.setdefaulttimeout(15): defence-in-depth for missing SOP_TIER_CHECK_TOKEN.
+          # gate_check.py uses timeout=15 on every urlopen call; this catches the
+          # inline Python polling loop too (issue #603).
          pr_numbers=$(python3 <<'PY'
          import json
          import os
          import socket
-          import sys
-          import time
-          import urllib.error
          import urllib.request

-          socket.setdefaulttimeout(30)
+          socket.setdefaulttimeout(15)
          token = os.environ["GITEA_TOKEN"]
          repo = os.environ["REPO"]
-          url = f"https://git.moleculesai.app/api/v1/repos/{repo}/pulls?state=open&limit=100"
-          last_error = None
-          for attempt in range(1, 4):
-              req = urllib.request.Request(
-                  url,
-                  headers={"Authorization": f"token {token}", "Accept": "application/json"},
-              )
-              try:
-                  with urllib.request.urlopen(req, timeout=30) as r:
-                      prs = json.loads(r.read())
-                  break
-              except (TimeoutError, OSError, urllib.error.URLError, urllib.error.HTTPError) as exc:
-                  last_error = exc
-                  print(f"warning: PR list fetch attempt {attempt}/3 failed: {exc}", file=sys.stderr)
-                  if attempt < 3:
-                      time.sleep(2 * attempt)
-          else:
-              print(f"warning: skipped scheduled gate-check refresh; failed to list open PRs after 3 attempts: {last_error}", file=sys.stderr)
-              raise SystemExit(0)
+          req = urllib.request.Request(
+              f"https://git.moleculesai.app/api/v1/repos/{repo}/pulls?state=open&limit=100",
+              headers={"Authorization": f"token {token}", "Accept": "application/json"},
+          )
+          with urllib.request.urlopen(req) as r:
+              prs = json.loads(r.read())
          for pr in prs:
              print(pr["number"])
          PY
@@ -13,12 +13,8 @@ name: gitea-merge-queue
 #   - add `merge-queue-hold` to pause a queued PR without removing it

 on:
-  # Schedule moved to operator-config:
-  #   /etc/cron.d/molecule-core-merge-queue ->
-  #   /usr/local/bin/molecule-core-cron-bot.sh merge-queue
-  #
-  # The queue bot still processes one PR per tick, but no longer occupies
-  # one of the shared Actions runners just to poll.
+  schedule:
+    - cron: '*/5 * * * *'
  workflow_dispatch:

 permissions:
@@ -56,9 +52,5 @@ jobs:
          # explicitly instead of the combined state avoids false-pause when
          # non-blocking jobs (continue-on-error: true) have failed — those
          # failures pollute combined state but do not gate merges.
-          # NOTE: the event-suffixed context name is intentional — branch protection
-          # MUST require `CI / all-required (pull_request)` (with suffix), NOT the
-          # bare `CI / all-required`. Gitea treats absent contexts as pending, not
-          # skipped; requiring the bare name silently blocks all merges (issue #1473).
          PUSH_REQUIRED_CONTEXTS: CI / all-required (push)
        run: python3 .gitea/scripts/gitea-merge-queue.py
@@ -77,37 +77,47 @@ env:
 jobs:
  detect-changes:
    name: detect-changes
-    # mc#1529 §1: pin to `docker-host` so the integration job runs on the
-    # operator-host runners (molecule-runner-*), which carry the
-    # `molecule-core-net` bridge network this workflow depends on. PC2
-    # runners (hongming-pc-runner-*) also advertise ubuntu-latest but
-    # don't have that network — the previous `runs-on: ubuntu-latest`
-    # rolled the dice and hard-failed the bridge-inspect step ~30% of
-    # the time. detect-changes itself doesn't need the bridge, but keeping
-    # both jobs on the same label avoids workspace-volume cross-host
-    # surprises and keeps the routing rule discoverable in one place.
-    runs-on: docker-host
-    # mc#1982 Phase 3 (RFC §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    runs-on: ubuntu-latest
+    # mc#774 Phase 3 (RFC §1): surface broken workflows without blocking.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    outputs:
      handlers: ${{ steps.filter.outputs.handlers }}
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
-          # A full-history checkout can exceed the runner's quiet/startup
-          # window before the path filter emits logs. Fetch the common push
-          # case cheaply; the script below fetches the exact BASE SHA if it is
-          # not present in the shallow checkout.
-          fetch-depth: 2
+          fetch-depth: 0
      - id: filter
+        # Inline replacement for dorny/paths-filter — see e2e-api.yml.
        run: |
-          python3 .gitea/scripts/detect-changes.py \
-            --profile handlers-postgres \
-            --event-name "${{ github.event_name }}" \
-            --pr-base-sha "${{ github.event.pull_request.base.sha }}" \
-            --base-ref "${{ github.event.pull_request.base.ref }}" \
-            --push-before "${GITHUB_EVENT_BEFORE:-}"
+          # Gitea Actions evaluates github.event.before to empty string in shell
+          # scripts. Use GITHUB_EVENT_BEFORE shell env var instead (Gitea
+          # correctly populates it for push events). PR case uses template var.
+          BASE=""
+          if [ "${{ github.event_name }}" = "pull_request" ] && [ -n "${{ github.event.pull_request.base.sha }}" ]; then
+            BASE="${{ github.event.pull_request.base.sha }}"
+          elif [ -n "$GITHUB_EVENT_BEFORE" ]; then
+            BASE="$GITHUB_EVENT_BEFORE"
+          fi
+          if [ -z "$BASE" ] || echo "$BASE" | grep -qE '^0+$'; then
+            echo "handlers=true" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          # timeout 30 guards against the case where BASE points to a ref that
+          # git can resolve but cat-file hangs (rare on corrupted objects).
+          if ! timeout 30 git cat-file -e "$BASE" 2>/dev/null; then
+            git fetch --depth=1 origin "$BASE" 2>/dev/null || true
+          fi
+          if ! timeout 30 git cat-file -e "$BASE" 2>/dev/null; then
+            echo "handlers=true" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          CHANGED=$(git diff --name-only "$BASE" HEAD)
+          if echo "$CHANGED" | grep -qE '^(workspace-server/internal/handlers/|workspace-server/internal/wsauth/|workspace-server/migrations/|\.gitea/workflows/handlers-postgres-integration\.yml$)'; then
+            echo "handlers=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "handlers=false" >> "$GITHUB_OUTPUT"
+          fi

  # Single-job-with-per-step-if pattern: always runs to satisfy the
  # required-check name on branch protection; real work gates on the
@@ -115,11 +125,9 @@ jobs:
  integration:
    name: Handlers Postgres Integration
    needs: detect-changes
-    # mc#1529 §1: must run on operator-host (where `molecule-core-net`
-    # exists). See detect-changes for the full routing rationale.
-    runs-on: docker-host
-    # mc#1982 Phase 3 (RFC §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    runs-on: ubuntu-latest
+    # mc#774 Phase 3 (RFC §1): surface broken workflows without blocking.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    env:
      # Unique name per run so concurrent jobs don't collide on the
@@ -62,15 +62,9 @@ env:
 jobs:
  # bp-exempt: change detector only; downstream Harness Replays is the meaningful gate.
  detect-changes:
-    # mc#1529 follow-on: pin to `docker-host` so this lane lands on
-    # Linux operator-host runners (the only ones with a working
-    # docker.sock + `molecule-core-net`). The bare `ubuntu-latest`
-    # label is also matched by hongming-pc-runner-* (Windows act_runner
-    # v1.0.3), where the `docker compose ...` exec below fails. Mirror
-    # of mc#1543; see internal#512 for class defect.
-    runs-on: docker-host
+    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    outputs:
      run: ${{ steps.decide.outputs.run }}
@@ -168,11 +162,9 @@ jobs:
  harness-replays:
    needs: detect-changes
    name: Harness Replays
-    # mc#1529 follow-on: `docker compose ... ps/logs` against tenant-alpha/
-    # beta containers. Must run on operator-host Linux (docker-host).
-    runs-on: docker-host
+    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 30
    steps:
@@ -1,6 +1,6 @@
 name: lint-bp-context-emit-match

-# Tier 2f scheduled lint (per mc#1982) — detects drift between
+# Tier 2f scheduled lint (per mc#774) — detects drift between
 # `branch_protections/<branch>.status_check_contexts` and the set of
 # contexts emitted by `.gitea/workflows/*.yml`.
 #
@@ -60,7 +60,7 @@ name: lint-bp-context-emit-match
 #
 # Cross-links
 # -----------
-# - mc#1982 (the RFC that specs this lint)
+# - mc#774 (the RFC that specs this lint)
 # - internal#349 (cross-repo BP sweep)
 # - feedback_phantom_required_check_after_gitea_migration
 # - feedback_tier_label_ids_are_per_repo
@@ -91,10 +91,10 @@ jobs:
    name: lint-bp-context-emit-match
    runs-on: ubuntu-latest
    timeout-minutes: 5
-    # Phase 4 (RFC #219 §1): 22 days green since 2026-05-11 port,
-    # well past the 7-clean-run threshold. Scheduled failure is now
-    # a hard CI signal.
-    continue-on-error: false
+    # Phase 3 (RFC #219 §1): surface drift without blocking. After 7
+    # clean scheduled runs on main, flip to false so a scheduled
+    # failure is a hard CI signal.
+    continue-on-error: true  # mc#774 Phase 3 — flip to false after 7 clean main runs
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065  # v5.6.0
@@ -1,6 +1,6 @@
 name: lint-continue-on-error-tracking

-# Tier 2e hard-gate lint (per mc#1982) — every
+# Tier 2e hard-gate lint (per mc#774) — every
 # `continue-on-error: true` in `.gitea/workflows/*.yml` must carry a
 # `# mc#NNNN` or `# internal#NNNN` tracker comment within 2 lines,
 # the referenced issue must be OPEN, and ≤14 days old.
@@ -8,7 +8,7 @@ name: lint-continue-on-error-tracking
 # Why this exists
 # ---------------
 # `continue-on-error: true` on `platform-build` had been hiding
-# mc#1982-class regressions for ~3 weeks before #656 surfaced them on
+# mc#774-class regressions for ~3 weeks before #656 surfaced them on
 # 2026-05-12. A 14-day cap on tracker age forces a review cycle and
 # surfaces mask-drift within at most 14 days of the original defect.
 # Each `continue-on-error: true` gets a paper trail — close or renew.
@@ -45,12 +45,12 @@ name: lint-continue-on-error-tracking
 # close-and-flip, or document the deliberate keep-mask in a fresh
 # 14-day-renewable tracker. After main is clean for 3 days,
 # follow-up PR flips this workflow's continue-on-error to false.
-# Tracking: mc#1982.
+# Tracking: mc#774.
 #
 # Cross-links
 # -----------
-# - mc#1982 (the RFC that specs this lint)
-# - mc#1982 (the empirical masked-3-weeks case)
+# - mc#774 (the RFC that specs this lint)
+# - mc#774 (the empirical masked-3-weeks case)
 # - feedback_chained_defects_in_never_tested_workflows
 # - feedback_behavior_based_ast_gates
 # - feedback_strict_root_only_after_class_a
@@ -93,13 +93,13 @@ jobs:
  lint:
    name: lint-continue-on-error-tracking
    runs-on: ubuntu-latest
-    timeout-minutes: 20
+    timeout-minutes: 10
    # Phase 3 (RFC #219 §1): surface masked defects without blocking
    # PRs. Pre-existing continue-on-error: true directives on main
    # all violate this lint at first — intentional. Flip to false
-    # follow-up after main is clean for 3 days. mc#1982.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
-    continue-on-error: true  # mc#1982 Phase 3 mask — 14d forced-renewal cadence
+    # follow-up after main is clean for 3 days. mc#774.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    continue-on-error: true  # mc#774 Phase 3 mask — 14d forced-renewal cadence
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065  # v5.6.0
@@ -48,9 +48,11 @@ jobs:
  scan:
    name: Scan workflows for curl status-capture pollution
    runs-on: ubuntu-latest
-    # Phase 4 (RFC #219 §1): 22 days green since 2026-05-11 port.
-    # mc#1982 mask removed — no surfaced defects in this lane.
-    continue-on-error: false
+    # Phase 3 (RFC #219 §1): surface broken workflows without blocking
+    # the PR. Follow-up PR flips this off after surfaced defects are
+    # triaged.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    continue-on-error: true
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
      - name: Find curl ... -w '%{http_code}' ... || echo "000" subshells
@@ -1,306 +0,0 @@
-name: Lint forbidden tenant-env keys
-
-# RFC#523 Layer 3 (task #146): scan workspace_secrets-writer Go code
-# under workspace-server/ for new code that hardcodes a forbidden
-# operator-scope env var NAME (GITEA_TOKEN, CP_ADMIN_API_TOKEN,
-# RAILWAY_TOKEN, INFISICAL_OPERATOR_TOKEN, MOLECULE_OPERATOR_*, …).
-#
-# Catches the class "a new writer accidentally widens the propagation
-# set" — e.g. a future env-mutator plugin that sets envVars["GITEA_TOKEN"]
-# directly. Today the L1 runtime guard would abort the provision, but
-# this lint surfaces the offending code at PR review time instead of
-# at first provision attempt.
-#
-# Companion layers:
-#   - L1: workspace-server/internal/handlers/workspace_provision_forbidden_env.go
-#         (fail-closed abort at provision time)
-#   - L2: workspace/entrypoint.sh top-of-file env-grep + exit 1
-#
-# Open-source-template-friendly: the deny pattern is generic. A fork
-# can copy this workflow and replace OPERATOR_KEY_PATTERN with its
-# own operator-scope key names.
-#
-# Path-filter discipline:
-#   This workflow runs on every PR (no paths: filter — see
-#   feedback_path_filtered_workflow_cant_be_required). The scan itself
-#   targets workspace_secrets-writer paths via grep -r; it's fast
-#   (sub-second) so unconditional run is fine.
-#
-# ── 2026-06-01 CI-scheduler-fanout consolidation (fix/ci-scheduler-fanout) ──
-# The RFC#523 sibling lint formerly in its own file
-# `lint-no-tenant-gitea-token.yml` (the broader "no repo-host token into
-# any tenant-writer surface" scan) is now a SECOND job in THIS workflow
-# (`scan-tenant-token-write`). Both are sub-second Go-source greps that
-# fired as two separate workflow runs on every PR — pure scheduler
-# fan-out. Folding the sibling in here drops one workflow run + one
-# checkout per PR while keeping BOTH scans firing unconditionally on
-# every PR (the no-paths discipline above is preserved — neither job is
-# paths-filtered). The moved job keeps its exact `name:` so its emitted
-# status context is unchanged in substance; its `# bp-exempt:` directive
-# moves with it (Tier 2g). The old `Lint no tenant GITEA or GITHUB token
-# write / …` context is retired (a disappearing context needs no
-# directive; only NEW emitters do).
-
-on:
-  pull_request:
-    types: [opened, synchronize, reopened]
-  push:
-    branches: [main, staging]
-
-env:
-  GITHUB_SERVER_URL: https://git.moleculesai.app
-
-jobs:
-  scan:
-    name: Scan workspace_secrets writers for forbidden env keys
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-        with:
-          fetch-depth: 1
-
-      - name: Scan for forbidden operator-scope env key NAMES in writer paths
-        run: |
-          set -euo pipefail
-
-          # Forbidden EXACT-MATCH env var names. Kept in lockstep with
-          # workspace-server/internal/handlers/workspace_provision_forbidden_env.go
-          # forbiddenTenantEnvKeys. The Go-side test
-          # TestIsForbiddenTenantEnvKey_ExactMatches is the source of
-          # truth — if Go-side adds a key, also add it here (and
-          # vice-versa). Drift between the two is the failure mode this
-          # entire 3-layer guardrail is designed to catch.
-          FORBIDDEN_KEYS=(
-            "GITEA_TOKEN" "GITEA_PAT"
-            "GITHUB_TOKEN" "GITHUB_PAT" "GH_TOKEN"
-            "GITLAB_TOKEN" "GL_TOKEN"
-            "BITBUCKET_TOKEN"
-            "CP_ADMIN_API_TOKEN" "CP_ADMIN_TOKEN"
-            "INFISICAL_OPERATOR_TOKEN" "INFISICAL_BOOTSTRAP_TOKEN"
-            "RAILWAY_TOKEN" "RAILWAY_PERSONAL_API_TOKEN"
-            "HETZNER_TOKEN" "HETZNER_API_TOKEN"
-          )
-
-          # Forbidden PREFIX patterns — operator-scope families.
-          FORBIDDEN_PREFIXES=(
-            "MOLECULE_OPERATOR_"
-          )
-
-          # Writer paths: Go source under workspace-server/ that
-          # writes to the env-vars map or to workspace_secrets DB rows.
-          # Tests, the forbidden-env source itself, and the silent-
-          # strip denylist are exempt (they LIST the keys by design).
-          SCAN_ROOT="workspace-server/internal"
-          # Exempt paths fall in two classes:
-          #   1. The deny-set definitions + the silent-strip denylist:
-          #      they LIST the forbidden names by design.
-          #   2. Pre-RFC#523 persona-merge / config-read paths that
-          #      already handle these names correctly (the silent-
-          #      strip downstream + the new L1 fail-closed cover the
-          #      runtime risk; these reads are unchanged).
-          # New code MUST NOT be added to this list without reviewer
-          # signoff and a one-line justification in this diff.
-          EXEMPT_PATHS=(
-            # Class 1 — deny-set definitions
-            "workspace-server/internal/handlers/workspace_provision_forbidden_env.go"
-            "workspace-server/internal/handlers/workspace_provision_forbidden_env_test.go"
-            "workspace-server/internal/provisioner/provisioner.go"
-            "workspace-server/internal/provisioner/provisioner_test.go"
-            # Class 2 — pre-existing persona-fallback / org-helper paths
-            # that set the GITEA_TOKEN fallback lane (stripped downstream
-            # by provisioner.buildContainerEnv per forensic #145). The
-            # new L1 fail-closed runs BEFORE these writers, so any
-            # operator-scope leak via global/workspace_secrets is
-            # already caught. See applyAgentGitHTTPCreds doc-comment.
-            "workspace-server/internal/handlers/agent_git_identity.go"
-            "workspace-server/internal/handlers/org_helpers.go"
-            "workspace-server/internal/handlers/org.go"
-            # Class 2 — CP→platform admin auth (NOT a tenant env write;
-            # this is the control-plane HTTP auth header source).
-            "workspace-server/internal/provisioner/cp_provisioner.go"
-          )
-
-          # Build a single grep -F pattern: every forbidden key wrapped
-          # in quotes (Go string-literal form, which is how env-map
-          # writes appear). e.g. envVars["GITEA_TOKEN"] = ... or
-          # `"GITEA_TOKEN":` in a literal-map declaration.
-          #
-          # We deliberately match the quoted form so a comment that
-          # happens to spell the name without quotes (e.g. "see
-          # GITEA_TOKEN below") doesn't trip the lint.
-          PATTERN=""
-          for k in "${FORBIDDEN_KEYS[@]}"; do
-            PATTERN="${PATTERN}\"${k}\"\n"
-          done
-          for p in "${FORBIDDEN_PREFIXES[@]}"; do
-            # Prefix match needs a regex; switch to grep -E below for
-            # this slice. Kept conceptually here so the deny set lives
-            # in one place; scan is run twice (literal + prefix).
-            true
-          done
-
-          # Build exempt-paths grep filter — `grep -v -f` style.
-          EXEMPT_FILTER=$(mktemp)
-          trap 'rm -f "$EXEMPT_FILTER"' EXIT
-          for p in "${EXEMPT_PATHS[@]}"; do
-            echo "$p" >> "$EXEMPT_FILTER"
-          done
-
-          # --- Exact-match scan ---
-          HITS=""
-          for k in "${FORBIDDEN_KEYS[@]}"; do
-            # Only .go files; skip _test.go for the writer-path scan
-            # since tests legitimately reference the names. The
-            # writer-path lint targets PRODUCTION code only.
-            found=$(grep -rn --include='*.go' --exclude='*_test.go' "\"${k}\"" "$SCAN_ROOT" 2>/dev/null \
-                    | grep -v -F -f "$EXEMPT_FILTER" || true)
-            if [ -n "$found" ]; then
-              HITS="${HITS}${found}\n"
-            fi
-          done
-
-          # --- Prefix scan ---
-          for prefix in "${FORBIDDEN_PREFIXES[@]}"; do
-            found=$(grep -rnE --include='*.go' --exclude='*_test.go' "\"${prefix}[A-Z0-9_]+\"" "$SCAN_ROOT" 2>/dev/null \
-                    | grep -v -F -f "$EXEMPT_FILTER" || true)
-            if [ -n "$found" ]; then
-              HITS="${HITS}${found}\n"
-            fi
-          done
-
-          if [ -n "$HITS" ]; then
-            echo "::error::RFC#523 Layer 3: forbidden operator-scope env var name(s) hardcoded in tenant-workspace writer paths:"
-            printf "$HITS"
-            echo ""
-            echo "These env-var NAMES are on the operator-scope deny list (see"
-            echo "workspace-server/internal/handlers/workspace_provision_forbidden_env.go)."
-            echo "If your code legitimately needs to inject one of these for a"
-            echo "non-tenant code path, add the file to EXEMPT_PATHS in this"
-            echo "workflow with a one-line justification — reviewer signoff required."
-            exit 1
-          fi
-
-          echo "OK No forbidden operator-scope env key names hardcoded in writer paths."
-
-  # bp-exempt: advisory RFC#523 lint; PR review gate is review-driven, not BP-driven.
-  # (Carried with the workflow-name rename in PR mc#1593 so the renamed
-  # context emission satisfies lint_required_context_exists_in_bp Tier 2g.)
-  scan-tenant-token-write:
-    name: Scan for repo-host token write into tenant workspace surface
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-        with:
-          fetch-depth: 1
-
-      - name: Find Go files referencing a tenant-writer surface AND a repo-host token
-        run: |
-          set -euo pipefail
-
-          # Repo-host token NAMES — the threat-model subset. Operator-fleet
-          # tokens (CP_ADMIN_API_TOKEN, RAILWAY_TOKEN, INFISICAL_*) are
-          # caught by lint-forbidden-env-keys.yml's broader deny set; this
-          # lint focuses on the git-host class so a single co-occurrence
-          # match has a low false-positive rate.
-          FORBIDDEN_KEYS=(
-            "GITEA_TOKEN"
-            "GITEA_PAT"
-            "GITHUB_TOKEN"
-            "GITHUB_PAT"
-            "GH_TOKEN"
-          )
-
-          # Tenant-writer surface markers. A file matches the surface set
-          # if it references ANY of these strings. This is the "is this
-          # code path writing into a tenant workspace?" heuristic.
-          # Curated to catch the actual code shapes used in this repo
-          # (verified by grep against current main 2026-05-19):
-          #   - "workspace_secrets" / "global_secrets"  → DB table writes
-          #   - "seedAllowList"                          → CP-side seed table
-          #   - "/settings/secrets"                      → tenant HTTP API write
-          #   - "envVars["                               → in-memory env map write
-          #   - "containerEnv"                           → docker-run env-set
-          #   - "userData"                               → EC2 user-data script
-          #   - "provisionPayload" / "provisionContext"  → provision-request shape
-          SURFACE_PATTERN='workspace_secrets|global_secrets|seedAllowList|/settings/secrets|envVars\[|containerEnv|userData|provisionPayload|provisionContext'
-
-          # Files that legitimately reference these names AND a surface
-          # marker, but do so for guard / strip / test / doc-comment
-          # reasons. New entries require reviewer signoff and a one-line
-          # justification in the diff.
-          EXEMPT_FILES=(
-            # RFC#523 L1 deny-set source-of-truth + tests
-            "workspace-server/internal/handlers/workspace_provision_forbidden_env.go"
-            "workspace-server/internal/handlers/workspace_provision_forbidden_env_test.go"
-            # Forensic-#145 silent-strip denylist (defense-in-depth, by design lists the names)
-            "workspace-server/internal/provisioner/provisioner.go"
-            "workspace-server/internal/provisioner/provisioner_test.go"
-            # Pre-RFC#523 persona-fallback / org-helper paths. The L1
-            # fail-closed runs BEFORE these writers; downstream silent-strip
-            # also covers them. See applyAgentGitHTTPCreds doc-comment.
-            "workspace-server/internal/handlers/agent_git_identity.go"
-            "workspace-server/internal/handlers/org_helpers.go"
-            "workspace-server/internal/handlers/org.go"
-            # CP→platform admin auth (NOT a tenant env write).
-            "workspace-server/internal/provisioner/cp_provisioner.go"
-          )
-
-          # Build an extended-regex alternation of forbidden keys.
-          KEY_ALT="$(IFS='|'; echo "${FORBIDDEN_KEYS[*]}")"
-
-          # Find candidate files: Go non-test sources that contain a
-          # tenant-writer surface marker.
-          mapfile -t CANDIDATES < <(
-            grep -rlE --include='*.go' --exclude='*_test.go' \
-              "${SURFACE_PATTERN}" . 2>/dev/null \
-            | sed 's|^\./||' \
-            | sort -u
-          )
-
-          if [ "${#CANDIDATES[@]}" -eq 0 ]; then
-            echo "OK No tenant-writer-surface files found in tree (unexpected, but not a lint failure)."
-            exit 0
-          fi
-
-          HITS=""
-          for f in "${CANDIDATES[@]}"; do
-            # Skip exempt files.
-            skip=0
-            for ex in "${EXEMPT_FILES[@]}"; do
-              if [ "$f" = "$ex" ]; then skip=1; break; fi
-            done
-            [ "$skip" = "1" ] && continue
-
-            # File contains a surface marker; now grep for a forbidden
-            # key NAME. We require a QUOTED-literal match to avoid
-            # firing on a comment like "// also handle GITEA_TOKEN".
-            #
-            # The literal form catches:
-            #   - os.Getenv("GITEA_TOKEN")
-            #   - envVars["GITEA_TOKEN"] = ...
-            #   - {envKey: "GITEA_TOKEN", tenantKey: "GITEA_TOKEN"}
-            # but not:
-            #   - // see GITEA_TOKEN below   (no quotes)
-            found=$(grep -nE "\"(${KEY_ALT})\"" "$f" 2>/dev/null || true)
-            if [ -n "$found" ]; then
-              HITS="${HITS}--- ${f} ---\n${found}\n"
-            fi
-          done
-
-          if [ -n "$HITS" ]; then
-            echo "::error::Task #146 lint: repo-host token name(s) quoted in a tenant-writer-surface file:"
-            printf "$HITS"
-            echo ""
-            echo "These files reference a tenant-writer surface (workspace_secrets,"
-            echo "seedAllowList, /settings/secrets, containerEnv, userData, etc.)"
-            echo "AND quote a repo-host token name (GITEA_TOKEN/GITHUB_TOKEN/…)."
-            echo "Per RFC#523 threat model, tenant workspaces MUST NOT receive"
-            echo "operator-scope repo-host tokens. If your code legitimately needs"
-            echo "to reference one of these names in a tenant-writer file (e.g."
-            echo "a deny-set definition or silent-strip list), add the file to"
-            echo "EXEMPT_FILES with a one-line justification — reviewer signoff"
-            echo "required."
-            exit 1
-          fi
-
-          echo "OK No tenant-writer-surface file co-mentions a repo-host token literal."
@@ -1,6 +1,6 @@
 name: lint-mask-pr-atomicity

-# Tier 2d hard-gate lint (per mc#1982) — blocks PRs that touch
+# Tier 2d hard-gate lint (per mc#774) — blocks PRs that touch
 # `.gitea/workflows/ci.yml` and modify ONLY ONE of {continue-on-error,
 # all-required.sentinel.needs} without a `Paired: #NNN` reference in
 # the PR body or in a commit message.
@@ -37,13 +37,13 @@ name: lint-mask-pr-atomicity
 # This workflow lands at `continue-on-error: true` (Phase 3 — surface
 # regressions without blocking PRs while the rule beds in).
 # Follow-up PR flips to `false` once we have ≥3 days of clean runs on
-# `main` and no false-positives. Tracking issue: mc#1982.
+# `main` and no false-positives. Tracking issue: mc#774.
 #
 # Cross-links
 # -----------
-# - mc#1982 (the RFC that specs this lint)
+# - mc#774 (the RFC that specs this lint)
 # - PR#665 / PR#668 (the empirical split-pair)
-# - mc#1982 (the main-red incident the split caused)
+# - mc#774 (the main-red incident the split caused)
 # - feedback_strict_root_only_after_class_a
 # - feedback_behavior_based_ast_gates
 #
@@ -92,8 +92,8 @@ jobs:
    # Phase 3 (RFC #219 §1): surface broken shapes without blocking
    # PRs. Follow-up PR flips this to `false` once recent runs on main
    # are confirmed clean (eat-our-own-dogfood discipline mirrors
-    # PR#673's same-shape comment). Tracking: mc#1982.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # PR#673's same-shape comment). Tracking: mc#774.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    steps:
      - name: Check out PR head with full history (need base SHA blobs)
@@ -4,7 +4,7 @@ name: Lint pre-flip continue-on-error
 # on any job in `.gitea/workflows/*.yml` WITHOUT proof that the affected
 # job's recent runs on the target branch (PR base) are actually green.
 #
-# Empirical class: PR #656 / mc#1982. PR #656 (RFC internal#219 Phase 4)
+# Empirical class: PR #656 / mc#774. PR #656 (RFC internal#219 Phase 4)
 # flipped 5 platform-build-class jobs `continue-on-error: true → false`
 # on the basis of a "verified green on main via combined-status check".
 # But that "green" was the LIE the prior `continue-on-error: true`
@@ -13,7 +13,7 @@ name: Lint pre-flip continue-on-error
 # job-level status. The precondition the PR claimed to verify was
 # structurally fooled by the bug being flipped.
 #
-# mc#1982 captured the surfaced defects (2 mutually-masked regressions):
+# mc#774 captured the surfaced defects (2 mutually-masked regressions):
 #   - Class 1: sqlmock helper drift since 2f36bb9a (24 days old)
 #   - Class 2: OFFSEC-001 contract collision since 7d1a189f (1 day old)
 #
@@ -55,7 +55,7 @@ name: Lint pre-flip continue-on-error
 #   - YAML parse error in one of the workflow files: warn-only,
 #     don't block — the YAML lint workflows catch this separately.
 #
-# Cross-links: PR#656, mc#1982, PR#665 (interim re-mask),
+# Cross-links: PR#656, mc#774, PR#665 (interim re-mask),
 # Quirk #10 (internal#342 + dup #287), hongming-pc2 charter
 # §SOP-N rule (e), feedback_strict_root_only_after_class_a,
 # feedback_no_shared_persona_token_use.
@@ -99,8 +99,8 @@ jobs:
    timeout-minutes: 8
    # Phase 3 (RFC internal#219 §1): surface broken flips without blocking
    # the PR yet. Follow-up flips this to `false` once the workflow itself
-    # has clean recent runs on main. mc#1982 interim — remove when CoE→false.
-    continue-on-error: true  # mc#1982
+    # has clean recent runs on main. mc#774 interim — remove when CoE→false.
+    continue-on-error: true  # mc#774
    steps:
      - name: Check out PR head (full history for base-SHA access)
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
@@ -1,6 +1,6 @@
 name: lint-required-context-exists-in-bp

-# Tier 2g hard-gate lint (per mc#1982) — diff-based PR-time
+# Tier 2g hard-gate lint (per mc#774) — diff-based PR-time
 # check. When a PR adds a NEW commit-status emission (workflow YAML
 # `name:` + job `name:`-or-key + on:-event), the workflow file must
 # carry one of three directives adjacent to the new job:
@@ -16,7 +16,7 @@ name: lint-required-context-exists-in-bp
 # PR#656 added `CI / all-required (pull_request)` as a sentinel
 # context that workflows emit, but BP did NOT list it. When
 # platform-build failed, all-required failed, but BP let the PR
-# merge anyway → cascade to mc#1982. With this lint, PR#656 would
+# merge anyway → cascade to mc#774. With this lint, PR#656 would
 # have been blocked until either the BP PATCH ran alongside OR
 # the author added a `bp-required: pending` directive.
 #
@@ -27,7 +27,7 @@ name: lint-required-context-exists-in-bp
 # share the workflow-context enumeration helpers
 # (`_event_map`, `workflow_contexts`, `_job_display`) but the
 # semantics are intentionally distinct so they're separate scripts.
-# Co-design is documented in mc#1982.
+# Co-design is documented in mc#774.
 #
 # Directive comment lives in the workflow file (NOT PR body)
 # ----------------------------------------------------------
@@ -42,13 +42,13 @@ name: lint-required-context-exists-in-bp
 # Lands at `continue-on-error: true` (Phase 3 — surface the
 # pattern without blocking PRs while the directive convention
 # beds in). After 7 days of clean runs on `main` with no false
-# positives, follow-up flips to `false`. Tracking: mc#1982.
+# positives, follow-up flips to `false`. Tracking: mc#774.
 #
 # Cross-links
 # -----------
-# - mc#1982 (the RFC that specs this lint)
+# - mc#774 (the RFC that specs this lint)
 # - PR#656 (the empirical case)
-# - mc#1982 (the surfaced cascade)
+# - mc#774 (the surfaced cascade)
 # - feedback_phantom_required_check_after_gitea_migration (Tier 2f cousin)
 # - feedback_behavior_based_ast_gates
 #
@@ -81,10 +81,10 @@ jobs:
    name: lint-required-context-exists-in-bp
    runs-on: ubuntu-latest
    timeout-minutes: 5
-    # Phase 4 (RFC #219 §1): 22 days green since 2026-05-11 port,
-    # well past the 7-clean-day threshold. PR-time failure is now
-    # a hard CI signal.
-    continue-on-error: false
+    # Phase 3 (RFC #219 §1): surface the pattern without blocking PRs
+    # while the directive convention beds in. Follow-up flip to false
+    # after 7 clean days on main. mc#774.
+    continue-on-error: true  # mc#774 Phase 3 — flip to false after 7 clean main runs
    steps:
      - name: Check out PR head with full history (need base SHA blobs)
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
@@ -1,164 +0,0 @@
-name: lint-required-workflows-docker-host-pinned
-
-# Fail-closed lint that catches workflows touching docker.sock without
-# pinning `runs-on:` to a Linux-only label.
-#
-# Class defect (internal#512 + mc#1529 + today's oc#81/82/83 + autogen#8):
-# the `ubuntu-latest` label is advertised by BOTH the Linux operator-host
-# runners (molecule-runner-*) AND the Windows act_runner v1.0.3 on
-# hongming-pc-runner-*. Job placement is non-deterministic. When a docker-
-# bound job lands on a Windows runner, `docker run`/`docker login`/
-# `docker compose` fail with platform-specific errors ("protocol not
-# available", "cannot exec", etc.) — placement-dependent, not transient.
-#
-# This lint enforces the convention: any workflow whose YAML body
-# contains a docker exec (`docker run|build|buildx|compose|pull|push|
-# exec|tag|login|cp|inspect|ps` OR `docker/build-push-action|docker/
-# login-action|docker/setup-buildx`) MUST pin every job's `runs-on:` to
-# one of:
-#   - docker-host  (general docker.sock work — molecule-runner-*)
-#   - publish      (image build/push — molecule-runner-publish-*)
-#
-# Comments and heredoc/markdown bodies that merely MENTION docker are
-# excluded by the detection rule (see scan.py below).
-#
-# Per `feedback_never_skip_ci`: this is fail-closed (exit 1 on miss).
-
-on:
-  pull_request:
-    paths:
-      - '.gitea/workflows/**'
-  push:
-    branches: [main, staging]
-    paths:
-      - '.gitea/workflows/**'
-
-permissions:
-  contents: read
-
-env:
-  GITHUB_SERVER_URL: https://git.moleculesai.app
-
-jobs:
-  lint-docker-host-pin:
-    name: Lint docker-host pin on docker-touching workflows
-    runs-on: docker-host
-    timeout-minutes: 5
-    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-
-      - name: Scan workflows for docker-bound jobs missing docker-host/publish pin
-        run: |
-          set -euo pipefail
-          python3 - <<'PY'
-          import os
-          import re
-          import sys
-
-          # Docker-step detection: real exec, not just word-mention in comments.
-          # We strip comment-only lines, then look for the docker subcommand
-          # tokens at word-boundary, OR uses: docker/* actions.
-          DOCKER_EXEC = re.compile(
-              r'(?<!\w)docker\s+(run|build|buildx|compose|pull|push|exec|tag|login|cp|inspect|ps)\b'
-          )
-          DOCKER_ACTION = re.compile(
-              r'uses:\s*docker/(build-push-action|login-action|setup-buildx-action|setup-qemu-action)'
-          )
-          # Detect a job header line like `  myjob:` (2-space indent) AND its runs-on.
-          JOB_HEADER = re.compile(r'^( {2})([a-zA-Z0-9_-]+):\s*$')
-          RUNS_ON    = re.compile(r'^( {4})runs-on:\s*(.+?)\s*$')
-
-          ALLOWED_LABELS = {'docker-host', 'publish'}
-
-          fails = []
-          warnings = []
-
-          # Gitea is SSOT for molecule-core CI per task #347 / memory
-          # reference_molecule_core_actions_gitea_only. The legacy
-          # .github/workflows/ tree was deleted in SSOT-Instance-4 (#331).
-          roots = []
-          for root in ('.gitea/workflows',):
-              if os.path.isdir(root):
-                  roots.append(root)
-
-          for root in roots:
-              for fn in sorted(os.listdir(root)):
-                  if not (fn.endswith('.yml') or fn.endswith('.yaml')):
-                      continue
-                  path = os.path.join(root, fn)
-                  with open(path) as f:
-                      raw_lines = f.readlines()
-
-                  # Parse job headers + their runs-on. Simple line scan; relies on
-                  # 2-space job indent + 4-space runs-on indent under `jobs:`.
-                  jobs = []
-                  current = None
-                  in_jobs = False
-                  for i, line in enumerate(raw_lines, 1):
-                      if re.match(r'^jobs:\s*$', line):
-                          in_jobs = True
-                          continue
-                      if not in_jobs:
-                          continue
-                      mh = JOB_HEADER.match(line)
-                      if mh:
-                          if current:
-                              current['end'] = i - 1
-                              jobs.append(current)
-                          current = {'name': mh.group(2), 'line': i, 'end': len(raw_lines), 'runs_on': None}
-                          continue
-                      mr = RUNS_ON.match(line)
-                      if mr and current and current['runs_on'] is None:
-                          current['runs_on'] = mr.group(2).strip()
-                  if current:
-                      jobs.append(current)
-
-                  for j in jobs:
-                      # Strip pure-comment lines for docker-exec detection so
-                      # documentation comments don't trigger the lint. Scan the
-                      # current job body only: a workflow may contain one
-                      # docker-bound job and several harmless metadata jobs.
-                      job_lines = raw_lines[j['line'] - 1:j['end']]
-                      scan_text = ''.join(
-                          l for l in job_lines
-                          if not re.match(r'^\s*#', l)
-                      )
-                      has_docker = bool(DOCKER_EXEC.search(scan_text)) or bool(DOCKER_ACTION.search(scan_text))
-                      if not has_docker:
-                          continue
-                      ro = j['runs_on']
-                      if ro is None:
-                          # Reusable workflow caller (`uses:` instead of `runs-on:`) —
-                          # skip; rule enforced in the called workflow.
-                          continue
-                      # Strip surrounding [ ] and quotes.
-                      ro_norm = ro.strip('[]').strip().strip('"\'')
-                      # Multi-label "[a, b]" — split.
-                      labels = [t.strip().strip('"\'') for t in ro_norm.split(',') if t.strip()]
-                      if any(lbl in ALLOWED_LABELS for lbl in labels):
-                          continue
-                      # Allow caller-supplied label expressions; spell the
-                      # marker indirectly so Gitea's expression parser does
-                      # not try to parse this Python heredoc.
-                      expression_marker = '$' + '{{'
-                      if any(expression_marker in lbl for lbl in labels):
-                          continue
-                      fails.append(
-                          f"{path}:{j['line']}: job `{j['name']}` uses docker but runs-on={ro!r} "
-                          f"(must be one of {sorted(ALLOWED_LABELS)})"
-                      )
-
-          if fails:
-              print("FAIL: docker-bound jobs missing docker-host/publish pin:")
-              for f in fails:
-                  print(f"  - {f}")
-              print()
-              print("Why this rule exists (internal#512 + mc#1529):")
-              print("  Bare `ubuntu-latest` is advertised by BOTH Linux operator-host")
-              print("  runners AND Windows hongming-pc-runner-* (act_runner v1.0.3).")
-              print("  Docker-bound jobs that land on Windows fail non-deterministically.")
-              print("  Pin to `docker-host` (general) or `publish` (image build/push).")
-              sys.exit(1)
-
-          print("OK: all docker-bound jobs are pinned to docker-host or publish.")
-          PY
@@ -1,117 +0,0 @@
-name: Lint shellcheck (arm64 pilot)
-
-# Mac-CI dual-track pilot (#233). ADDITIVE / NOT REQUIRED.
-#
-# Validates the arm64 self-hosted lane (no docker.sock, no privileged
-# ops) before any required gate moves onto it.
-#
-# Runner label mapping (2026-05-22 fix): the actual Mac mini runner
-# registered in this Gitea ships labels
-#   ["self-hosted","macos-self-hosted-arm64","arm64-darwin"]
-# — no plain `arm64`. The earlier `runs-on: [self-hosted, arm64]`
-# could not match any registered runner so every fire of this workflow
-# was assigned task_id=0 / runner_id=NULL → Gitea cancelled it. The
-# rows showed up as Cancelled in the action status feed (not Failed)
-# but the lane never actually ran. Workflow now selects on
-# `arm64-darwin` which is the canonical Mac-arm64 label per the
-# Mac mini's registration (per internal#494 capability-honest labels).
-#
-# If we later want to add a Linux-arm64 runner to the same lane, add
-# both labels to that runner's registration AND broaden the selector
-# here — don't rename `arm64-darwin` (it's Mac-specific by design and
-# `feedback_pc2_runner_labels_must_stay_narrow` rule applies).
-#
-# Pairs with internal#543 (RFC: Mac arm64 multi-arch runner-base) and
-# internal#494 (multi-arch runner-base capability-honest labels).
-# No paths: filter on purpose (feedback_path_filtered_workflow_cant_be_required).
-
-on:
-  pull_request:
-    branches:
-      - main
-      - staging
-  push:
-    branches:
-      - main
-
-permissions:
-  contents: read
-
-jobs:
-  shellcheck-arm64:
-    name: shellcheck-arm64 (pilot)
-    runs-on: [self-hosted, arm64-darwin]
-    # NOT a required check; safe to sit pending until Mac runner is up.
-    # If the Mac runner has trouble pulling actions/checkout we fall
-    # back to a plain git clone (see step 'fallback clone').
-    timeout-minutes: 10
-    env:
-      GITHUB_SERVER_URL: https://git.moleculesai.app
-    steps:
-      - name: Identify runner
-        run: |
-          set -eu
-          echo "arch=$(uname -m)"
-          echo "kernel=$(uname -sr)"
-          echo "shell=$BASH_VERSION"
-          # Sanity: must actually be arm64. If amd64 sneaks in here,
-          # fail fast — that means the label routing is wrong.
-          case "$(uname -m)" in
-            aarch64|arm64) echo "arm64 confirmed" ;;
-            *) echo "ERROR: expected arm64, got $(uname -m)"; exit 1 ;;
-          esac
-
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 1
-
-      - name: Install shellcheck (arm64)
-        continue-on-error: true
-        run: |
-          set -eu
-          if command -v shellcheck >/dev/null 2>&1; then
-            echo "shellcheck already present: $(shellcheck --version | head -1)"
-          else
-            # Prefer apt if the runner base ships it; else download arm64 binary.
-            if command -v apt-get >/dev/null 2>&1; then
-              sudo apt-get update -qq
-              sudo apt-get install -y --no-install-recommends shellcheck
-            else
-              SC_VER=v0.10.0
-              curl -fsSL "https://github.com/koalaman/shellcheck/releases/download/${SC_VER}/shellcheck-${SC_VER}.linux.aarch64.tar.xz" \
-                | tar -xJf - --strip-components=1
-              sudo mv shellcheck /usr/local/bin/
-            fi
-          fi
-          shellcheck --version | head -2
-
-      - name: Run shellcheck on .gitea/scripts/*.sh
-        continue-on-error: true
-        run: |
-          set -eu
-          # Only the scripts we control under .gitea/scripts. Pilot
-          # scope is intentionally narrow — broaden in a follow-up
-          # once the lane is proven.
-          if ! command -v shellcheck >/dev/null 2>&1; then
-            echo "WARN: shellcheck binary not found — skipping (pilot mode)"
-            exit 0
-          fi
-          # NOTE: macOS ships Bash 3.2 (Apple license), no `mapfile`
-          # (Bash 4+ builtin). Mac mini runner empirically failed at
-          # `mapfile: command not found` (run 79275 / task 145654).
-          # Use the portable `while read` pattern instead — works on
-          # both Bash 3.2 (macOS) and Bash 4+ (Linux).
-          TARGETS=()
-          while IFS= read -r f; do
-            TARGETS+=("$f")
-          done < <(find .gitea/scripts -maxdepth 2 -type f -name '*.sh' | sort)
-          if [ "${#TARGETS[@]}" -eq 0 ]; then
-            echo "No .sh files found under .gitea/scripts — nothing to check"
-            exit 0
-          fi
-          echo "Checking ${#TARGETS[@]} file(s):"
-          printf '  %s\n' "${TARGETS[@]}"
-          # SC1091 = couldn't follow non-constant source; expected for
-          # CI-time analysis without the full runtime layout.
-          shellcheck --severity=error --exclude=SC1091 "${TARGETS[@]}"
@@ -55,7 +55,7 @@ jobs:
    # Phase 3 (RFC #219 §1): surface broken shapes without blocking PRs.
    # Follow-up PR flips this off after the 4 existing-on-main rule-2
    # (workflow_run) violations are migrated to a supported trigger.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
@@ -42,48 +42,27 @@ permissions:
  packages: write

 env:
-  # SSOT-Instance-10 (#333): ECR registry triplet (account.dkr.ecr.region.amazonaws.com)
-  # sourced from org/repo var `ECR_REGISTRY` with the current prod-account literal as
-  # bootstrap fallback. When the org var is set, the fallback becomes dead code and
-  # switching accounts/regions is a one-line change at the org level (instead of
-  # touching every workflow). Pattern mirrors `vars.CP_URL || 'literal'` already in
-  # use below in this repo's staging-verify.yml.
-  IMAGE_NAME: ${{ vars.ECR_REGISTRY || '153263036946.dkr.ecr.us-east-2.amazonaws.com' }}/molecule-ai/canvas
+  IMAGE_NAME: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/canvas
  GITHUB_SERVER_URL: https://git.moleculesai.app

 jobs:
  # bp-exempt: post-merge image publication side effect; CI / all-required gates source changes.
  build-and-push:
    name: Build & push canvas image
-    # Dedicated publish/release lane (internal#462 / #394 / #399). Ship
-    # path (on: push:main, canvas/**) — reserved capacity so a merged
-    # canvas fix's image build never FIFO-queues behind PR required-CI.
-    # The `publish` label resolves ONLY to the molecule-runner-publish-*
-    # sub-pool (config.publish.yaml). HARD DEPENDENCY: this MUST land
-    # AFTER the publish-lane runners are registered/advertising `publish`
-    # — the earlier #599 `docker` label attempt queued indefinitely with
-    # zero eligible runners precisely because the label was targeted
-    # before any runner advertised it (see #576). The lane is registered
-    # in this rollout (internal#462) so the precondition holds.
-    runs-on: publish
+    # REVERTED (infra/revert-docker-runner-label): `runs-on: ubuntu-latest` restored.
+    # The `docker` label is not registered on any act_runner. `runs-on: [ubuntu-latest, docker]`
+    # causes jobs to queue indefinitely with zero eligible runners — strictly worse than the
+    # pre-#599 coin-flip (50% success rate). Once the `docker` label is registered on
+    # ≥2 runners, re-apply the fix from #599 (infra/docker-runner-label).
+    # See issue #576 + infra-lead pulse ~00:30Z.
+    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    steps:
      - name: Checkout
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2

-      # Keep Docker auth/buildx state inside the job temp dir. Publish
-      # runners can inherit a HOME/DOCKER_CONFIG path that is host-owned
-      # and not writable from the job container; docker login otherwise
-      # fails before the image build starts.
-      - name: Prepare writable Docker config
-        run: |
-          set -euo pipefail
-          export DOCKER_CONFIG="$RUNNER_TEMP/docker-config"
-          mkdir -p "$DOCKER_CONFIG/buildx/certs"
-          echo "DOCKER_CONFIG=$DOCKER_CONFIG" >> "$GITHUB_ENV"
-
      - name: Log in to ECR
        env:
          IMAGE_NAME: ${{ env.IMAGE_NAME }}
@@ -0,0 +1,152 @@
+name: publish-runtime-autobump
+
+# Auto-bump-on-workspace-edit half of the publish pipeline.
+#
+# Why this file exists (issue #351):
+#   Gitea Actions does not correctly disambiguate `paths:` from `tags:`
+#   when both are bundled under a single `on.push` key. The result is
+#   that tag pushes get filtered out and `publish-runtime.yml` never
+#   fires — `action_run` rows: 0. This was unnoticed pre-2026-05-11
+#   because PYPI_TOKEN was absent (publishes would have failed anyway).
+#
+#   Split design:
+#     - publish-runtime.yml         : on.push.tags only        (the publisher)
+#     - publish-runtime-autobump.yml: on.push.branches+paths   (this file — the version-bumper)
+#
+#   This file computes the next version from PyPI's latest, pushes a
+#   `runtime-v$VERSION` tag, and exits. The tag push then triggers
+#   publish-runtime.yml via its tags-only trigger.
+#
+# Concurrency: shares the `publish-runtime` group with publish-runtime.yml
+# so concurrent workspace pushes serialize at the bump step. Without
+# this, two pushes minutes apart could both read PyPI latest=0.1.129
+# and try to tag 0.1.130 simultaneously, only one of which would land.
+
+on:
+  # Run on PR pushes to post a success status so Gitea can merge the PR.
+  # All steps use continue-on-error: true so operational failures
+  # (PyPI unreachable, DISPATCH_TOKEN missing) do not block merge.
+  pull_request:
+    paths:
+      - "workspace/**"
+  # Bump-and-tag on main/staging push (the actual operational trigger).
+  push:
+    branches:
+      - main
+      - staging
+    paths:
+      - "workspace/**"
+  # Manual dispatch — useful when Gitea Actions API (/actions/*) is
+  # unreachable (e.g. act_runner 404 on Gitea 1.22.6) and we cannot
+  # re-trigger via curl.
+  workflow_dispatch:
+
+permissions:
+  contents: write  # required to push tags back
+
+concurrency:
+  group: publish-runtime
+  cancel-in-progress: false
+
+jobs:
+  # PR-validation path: always succeeds so Gitea can merge workflow-only PRs.
+  # Operational failures (PyPI unreachable, missing DISPATCH_TOKEN) are
+  # surfaced via continue-on-error: true rather than blocking the merge.
+  # The actual bump work happens on the main/staging push after merge.
+  # bp-exempt: advisory validation for runtime publication; not a branch-protection gate.
+  pr-validate:
+    runs-on: ubuntu-latest
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    continue-on-error: true  # do not block PR merge on operational failures
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          fetch-depth: 1
+
+      - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: "3.11"
+
+      - name: Validate PyPI connectivity (best-effort)
+        run: |
+          set -eu
+          echo "=== Checking PyPI accessibility ==="
+          LATEST=$(curl -fsS --retry 3 --max-time 10 \
+            https://pypi.org/pypi/molecule-ai-workspace-runtime/json \
+            | python -c "import sys,json; print(json.load(sys.stdin)['info']['version'])" \
+            || echo "PyPI unreachable (non-blocking for PR validation)")
+          echo "Latest: ${LATEST:-unknown}"
+
+  # Actual bump-and-tag: runs on main/staging pushes, posts real success/failure.
+  # No continue-on-error — operational failures here trip the main-red
+  # watchdog, which is the desired signal for infrastructure degradation.
+  # bp-exempt: post-merge tag publication side effect; CI / all-required gates source changes.
+  bump-and-tag:
+    runs-on: ubuntu-latest
+    # Only fire on push events (main/staging after PR merge). Pull_request
+    # events are handled by pr-validate above; we do NOT bump on every
+    # push-synchronize because that would race with the PR head.
+    #
+    # NOTE: the prior condition `github.event.pull_request.base.ref == ''`
+    # was broken — on a PR-merge push in Gitea Actions, the pull_request
+    # context is still attached (base.ref='main'), so the condition always
+    # evaluated to false and bump-and-tag was permanently skipped.
+    if: github.event_name == 'push'
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          fetch-depth: 1
+
+      - name: Fetch tags for collision check
+        run: git fetch origin --tags --depth=1
+
+      - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: "3.11"
+
+      - name: Compute next version from PyPI latest
+        id: bump
+        run: |
+          set -eu
+          LATEST=$(curl -fsS --retry 3 https://pypi.org/pypi/molecule-ai-workspace-runtime/json \
+            | python -c "import sys,json; print(json.load(sys.stdin)['info']['version'])")
+          MAJOR=$(echo "$LATEST" | cut -d. -f1)
+          MINOR=$(echo "$LATEST" | cut -d. -f2)
+          PATCH=$(echo "$LATEST" | cut -d. -f3)
+          VERSION="${MAJOR}.${MINOR}.$((PATCH+1))"
+          echo "PyPI latest=$LATEST -> next=$VERSION"
+          if ! echo "$VERSION" | grep -qE '^[0-9]+\.[0-9]+\.[0-9]+$'; then
+            echo "::error::computed version $VERSION does not match PEP 440 X.Y.Z"
+            exit 1
+          fi
+          if git tag --list | grep -qx "runtime-v$VERSION"; then
+            echo "::error::tag runtime-v$VERSION already exists in this repo. Manual intervention required (PyPI and Gitea tag history are out of sync)."
+            exit 1
+          fi
+          echo "version=$VERSION" >> "$GITHUB_OUTPUT"
+
+      - name: Push runtime-v$VERSION tag
+        env:
+          DISPATCH_TOKEN: ${{ secrets.DISPATCH_TOKEN }}
+          VERSION: ${{ steps.bump.outputs.version }}
+          GITEA_URL: https://git.moleculesai.app
+        run: |
+          set -eu
+          if [ -z "$DISPATCH_TOKEN" ]; then
+            echo "::error::DISPATCH_TOKEN secret is not set — needed to push the tag back to molecule-core."
+            exit 1
+          fi
+          git config user.name  "publish-runtime autobump"
+          git config user.email "publish-runtime@moleculesai.app"
+          git tag -a "runtime-v$VERSION" \
+            -m "Auto-bump on workspace/** edit on $GITHUB_REF" \
+            -m "Triggered by: $GITHUB_REF @ $GITHUB_SHA" \
+            -m "publish-runtime.yml will pick up this tag and upload to PyPI"
+          # Push via DISPATCH_TOKEN (a Gitea PAT). Using the bot identity
+          # ensures the resulting tag-push event is dispatched to
+          # publish-runtime.yml; act_runner's default GITHUB_TOKEN cannot
+          # trigger downstream workflows.
+          git remote set-url origin "${GITEA_URL#https://}"
+          git remote set-url origin "https://x-access-token:${DISPATCH_TOKEN}@${GITEA_URL#https://}/molecule-ai/molecule-core.git"
+          git push origin "runtime-v$VERSION"
+          echo "✓ pushed runtime-v$VERSION — publish-runtime.yml should fire next"
@@ -0,0 +1,339 @@
+name: publish-runtime
+
+# Gitea Actions port of .github/workflows/publish-runtime.yml.
+#
+# Ported 2026-05-10 (issue #206). Key differences from the GitHub version:
+#   - Gitea Actions reads .gitea/workflows/, not .github/workflows/
+#   - Dropped `environment: pypi-publish` — Gitea Actions does not support
+#     named environments or OIDC trusted publishers
+#   - Replaced `pypa/gh-action-pypi-publish@release/v1` (OIDC) with
+#     `twine upload` using PYPI_TOKEN secret — same mechanism as a local
+#     `python -m twine upload` with a PyPI token
+#   - Replaced `github.ref_name` (GitHub-only) with `${GITHUB_REF#refs/tags/}`
+#     — Gitea Actions exposes github.ref (the full ref) but not ref_name
+#   - Dropped `merge_group` trigger (Gitea has no merge queue)
+#
+# 2026-05-10 (issue #348): originally restored `staging`/`main` branch +
+# `workspace/**` path-filter trigger in PR #349.
+#
+# 2026-05-11 (issue #351): REVERTED the branches+paths trigger from THIS
+# file. Bundling `paths` with `tags` under a single `on.push` key caused
+# Gitea Actions to never dispatch the workflow for tag-push events (0
+# runs in `action_run` for workflow_id='publish-runtime.yml' since the
+# port, including the runtime-v1.0.0 tag — which is why PyPI is still at
+# 0.1.129 despite a v1.0.0 Gitea tag existing).
+#
+# The auto-bump-on-workspace-edit trigger now lives in
+# `.gitea/workflows/publish-runtime-autobump.yml`. That file computes the
+# next version from PyPI's latest and pushes a `runtime-v$VERSION` tag,
+# which THIS file then picks up via the tags-only trigger below.
+#
+# This decoupling means Gitea's path-vs-tag evaluator never has to
+# disambiguate — each file has a single unambiguous trigger shape.
+#
+# PyPI publishing: requires PYPI_TOKEN repository secret (or org-level secret).
+# Set via: repo Settings → Actions → Variables and Secrets → New Secret.
+# The token should be a PyPI API token scoped to molecule-ai-workspace-runtime.
+#
+# The DISPATCH_TOKEN cascade (git push to template repos) is unchanged —
+# it uses the Gitea API directly and was already Gitea-compatible.
+
+on:
+  push:
+    tags:
+      - "runtime-v*"
+  workflow_dispatch:
+  # 2026-05-11 (root cause of #351 / 0 runs ever):
+  # Gitea 1.22.6's workflow parser rejects `workflow_dispatch.inputs.version`
+  # with "unknown on type" — it mis-treats the inputs sub-keys as top-level
+  # `on:` event types. Log line:
+  #   actions/workflows.go:DetectWorkflows() [W] ignore invalid workflow
+  #   "publish-runtime.yml": unknown on type: map["version": {...}]
+  # That `[W] ignore invalid workflow` is silent UX — the workflow never
+  # registers, so it never fires for ANY event (push.tags included).
+  # Removing the inputs block restores parsing. Manual dispatch from the
+  # Gitea UI now triggers the PyPI auto-bump fallback in `Derive version`
+  # below (no `inputs.version` to read).
+
+permissions:
+  contents: read
+
+# Serialize publishes so two concurrent tag pushes don't both compute
+# "latest+1" and race on PyPI upload. The second one waits.
+concurrency:
+  group: publish-runtime
+  cancel-in-progress: false
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    outputs:
+      version: ${{ steps.version.outputs.version }}
+      wheel_sha256: ${{ steps.wheel_hash.outputs.wheel_sha256 }}
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: "3.11"
+          cache: pip
+
+      - name: Derive version (tag or PyPI auto-bump)
+        id: version
+        run: |
+          if echo "$GITHUB_REF" | grep -q "^refs/tags/runtime-v"; then
+            # Tag is `runtime-vX.Y.Z` — strip the prefix.
+            VERSION="${GITHUB_REF#refs/tags/runtime-v}"
+          else
+            # workflow_dispatch path (no inputs supported on Gitea 1.22.6) or
+            # any other non-tag trigger: derive from PyPI latest + patch bump.
+            LATEST=$(curl -fsS --retry 3 https://pypi.org/pypi/molecule-ai-workspace-runtime/json \
+              | python -c "import sys,json; print(json.load(sys.stdin)['info']['version'])")
+            MAJOR=$(echo "$LATEST" | cut -d. -f1)
+            MINOR=$(echo "$LATEST" | cut -d. -f2)
+            PATCH=$(echo "$LATEST" | cut -d. -f3)
+            VERSION="${MAJOR}.${MINOR}.$((PATCH+1))"
+            echo "Auto-bumped from PyPI latest $LATEST -> $VERSION"
+          fi
+          if ! echo "$VERSION" | grep -qE '^[0-9]+\.[0-9]+\.[0-9]+(\.dev[0-9]+|rc[0-9]+|a[0-9]+|b[0-9]+|\.post[0-9]+)?$'; then
+            echo "::error::version $VERSION does not match PEP 440"
+            exit 1
+          fi
+          echo "version=$VERSION" >> "$GITHUB_OUTPUT"
+          echo "Publishing molecule-ai-workspace-runtime $VERSION"
+
+      - name: Install build tooling
+        run: pip install build twine
+
+      - name: Build package from workspace/
+        run: |
+          python scripts/build_runtime_package.py \
+            --version "${{ steps.version.outputs.version }}" \
+            --out "${{ runner.temp }}/runtime-build"
+
+      - name: Build wheel + sdist
+        working-directory: ${{ runner.temp }}/runtime-build
+        run: python -m build
+
+      - name: Capture wheel SHA256 for cascade content-verification
+        id: wheel_hash
+        working-directory: ${{ runner.temp }}/runtime-build
+        run: |
+          set -eu
+          WHEEL=$(ls dist/*.whl 2>/dev/null | head -1)
+          if [ -z "$WHEEL" ]; then
+            echo "::error::No .whl in dist/ — \`python -m build\` must have failed silently"
+            exit 1
+          fi
+          HASH=$(sha256sum "$WHEEL" | awk '{print $1}')
+          echo "wheel_sha256=${HASH}" >> "$GITHUB_OUTPUT"
+          echo "Local wheel SHA256 (pre-upload): ${HASH}"
+          echo "Wheel filename: $(basename "$WHEEL")"
+
+      - name: Verify package contents (sanity)
+        working-directory: ${{ runner.temp }}/runtime-build
+        run: |
+          python -m twine check dist/*
+          python -m venv /tmp/smoke
+          /tmp/smoke/bin/pip install --quiet dist/*.whl
+          /tmp/smoke/bin/python "$GITHUB_WORKSPACE/scripts/wheel_smoke.py"
+
+      - name: Publish to PyPI
+        # working-directory matches the preceding Build/Verify steps. Without
+        # this, twine runs from the default workspace checkout dir where
+        # `dist/` doesn't exist and fails with:
+        #   ERROR InvalidDistribution: Cannot find file (or expand pattern): 'dist/*'
+        # Caught on the first-ever successful dispatch of this workflow
+        # (run 5097, 2026-05-11 02:08Z) — every other step in the publish
+        # job already had this working-directory; Publish was missing it.
+        working-directory: ${{ runner.temp }}/runtime-build
+        env:
+          # PYPI_TOKEN: repository secret scoped to molecule-ai-workspace-runtime.
+          # Set via: Settings → Actions → Variables and Secrets → New Secret.
+          # Format: pypi-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
+          PYPI_TOKEN: ${{ secrets.PYPI_TOKEN }}
+        run: |
+          if [ -z "$PYPI_TOKEN" ]; then
+            echo "::error::PYPI_TOKEN secret is not set — set it at Settings → Actions → Variables and Secrets → New Secret."
+            echo "::error::Required format: pypi-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"
+            exit 1
+          fi
+          python -m twine upload \
+            --repository pypi \
+            --username __token__ \
+            --password "$PYPI_TOKEN" \
+            dist/*
+
+  cascade:
+    needs: publish
+    runs-on: ubuntu-latest
+    steps:
+      - name: Wait for PyPI to propagate the new version
+        env:
+          RUNTIME_VERSION: ${{ needs.publish.outputs.version }}
+          EXPECTED_SHA256: ${{ needs.publish.outputs.wheel_sha256 }}
+        run: |
+          set -eu
+          if [ -z "$EXPECTED_SHA256" ]; then
+            echo "::error::publish job did not expose wheel_sha256 — cannot verify wheel content. Refusing to fan out cascade."
+            exit 1
+          fi
+          python -m venv /tmp/propagation-probe
+          PROBE=/tmp/propagation-probe/bin
+          $PROBE/pip install --upgrade --quiet pip
+          for i in $(seq 1 30); do
+            if $PROBE/pip install \
+                  --quiet \
+                  --no-cache-dir \
+                  --force-reinstall \
+                  --no-deps \
+                  "molecule-ai-workspace-runtime==${RUNTIME_VERSION}" \
+                  >/dev/null 2>&1; then
+              INSTALLED=$($PROBE/pip show molecule-ai-workspace-runtime 2>/dev/null \
+                          | awk -F': ' '/^Version:/{print $2}')
+              if [ "$INSTALLED" = "$RUNTIME_VERSION" ]; then
+                echo "✓ PyPI resolved $RUNTIME_VERSION (install check)"
+                break
+              fi
+            fi
+            if [ $i -eq 30 ]; then
+              echo "::error::pip install --no-cache-dir molecule-ai-workspace-runtime==${RUNTIME_VERSION} never resolved within ~5 min."
+              echo "::error::Refusing to fan out cascade against a potentially stale PyPI index."
+              exit 1
+            fi
+            echo "  [$i/30] waiting for PyPI to propagate ${RUNTIME_VERSION}..."
+            sleep 4
+          done
+
+          # Stage (b): download wheel + SHA256 compare against what we built.
+          # Catches Fastly stale-content serving old bytes under a new version URL.
+          #
+          # Caught run 5196 (first-ever successful publish, 2026-05-11): the
+          # previous one-liner `HASH=$(pip download ... && sha256sum ...)`
+          # captured pip's stdout (`Collecting molecule-ai-workspace-runtime
+          # ==X.Y.Z`) into HASH, then the SHA comparison failed against the
+          # leaked `Collecting...` string. `2>/dev/null` silences stderr but
+          # NOT stdout; pip writes its progress to stdout by default.
+          # Fix: split into two steps, silence pip's stdout explicitly, capture
+          # only sha256sum's output into HASH.
+          python -m pip download \
+            --no-deps \
+            --no-cache-dir \
+            --dest /tmp/wheel-probe \
+            --quiet \
+            "molecule-ai-workspace-runtime==${RUNTIME_VERSION}" \
+            >/dev/null 2>&1
+          HASH=$(sha256sum /tmp/wheel-probe/*.whl | awk '{print $1}')
+          if [ "$HASH" != "$EXPECTED_SHA256" ]; then
+            echo "::error::PyPI propagated $RUNTIME_VERSION but wheel content SHA256 mismatch."
+            echo "::error::Expected: $EXPECTED_SHA256"
+            echo "::error::Got:      $HASH"
+            echo "::error::Fastly may be serving stale content. Refusing to fan out cascade."
+            exit 1
+          fi
+          echo "✓ PyPI CDN verified (SHA256 match)"
+
+      - name: Fan out via push to .runtime-version
+        env:
+          # Gitea PAT with write:repository scope on the 8 cascade-active
+          # template repos. Used for git push to each template repo's main
+          # branch, which trips their `on: push: branches: [main]` trigger
+          # on publish-image.yml.
+          DISPATCH_TOKEN: ${{ secrets.DISPATCH_TOKEN }}
+          RUNTIME_VERSION: ${{ needs.publish.outputs.version }}
+        run: |
+          set +e   # don't abort on a single repo failure — collect them all
+
+          if [ -z "$DISPATCH_TOKEN" ]; then
+            if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
+              echo "::warning::DISPATCH_TOKEN secret not set — skipping cascade."
+              echo "::warning::set it at Settings → Actions → Variables and Secrets → New Secret."
+              exit 0
+            fi
+            echo "::error::DISPATCH_TOKEN secret missing — cascade cannot fan out."
+            echo "::error::PyPI was published, but the 8 template repos will NOT pick up the new version."
+            exit 1
+          fi
+          VERSION="$RUNTIME_VERSION"
+          if [ -z "$VERSION" ]; then
+            echo "::error::publish job did not expose a version output"
+            exit 1
+          fi
+
+          GITEA_URL="${GITEA_URL:-https://git.moleculesai.app}"
+          TEMPLATES="claude-code hermes openclaw codex langgraph crewai autogen deepagents gemini-cli"
+          FAILED=""
+          SKIPPED=""
+
+          git config --global user.name  "publish-runtime cascade"
+          git config --global user.email "publish-runtime@moleculesai.app"
+
+          WORKDIR="$(mktemp -d)"
+          for tpl in $TEMPLATES; do
+            REPO="molecule-ai/molecule-ai-workspace-template-$tpl"
+            CLONE="$WORKDIR/$tpl"
+
+            HTTP=$(curl -sS -o /dev/null -w "%{http_code}" \
+              -H "Authorization: token $DISPATCH_TOKEN" \
+              "$GITEA_URL/api/v1/repos/$REPO/contents/.github/workflows/publish-image.yml")
+            if [ "$HTTP" = "404" ]; then
+              echo "↷ $tpl has no publish-image.yml — soft-skip"
+              SKIPPED="$SKIPPED $tpl"
+              continue
+            fi
+
+            attempt=0
+            success=false
+            while [ $attempt -lt 3 ]; do
+              attempt=$((attempt + 1))
+              rm -rf "$CLONE"
+              if ! git clone --depth=1 \
+                  "https://x-access-token:${DISPATCH_TOKEN}@${GITEA_URL#https://}/$REPO.git" \
+                  "$CLONE" >/tmp/clone.log 2>&1; then
+                echo "::warning::clone $tpl attempt $attempt failed: $(tail -n3 /tmp/clone.log)"
+                sleep 2
+                continue
+              fi
+
+              cd "$CLONE"
+              echo "$VERSION" > .runtime-version
+
+              if git diff --quiet -- .runtime-version; then
+                echo "✓ $tpl already at $VERSION — no commit needed"
+                success=true
+                cd - >/dev/null
+                break
+              fi
+
+              git add .runtime-version
+              git commit -m "chore: pin runtime to $VERSION (publish-runtime cascade)" \
+                -m "Co-Authored-By: publish-runtime cascade <publish-runtime@moleculesai.app>" \
+                >/dev/null
+
+              if git push origin HEAD:main >/tmp/push.log 2>&1; then
+                echo "✓ $tpl pushed $VERSION on attempt $attempt"
+                success=true
+                cd - >/dev/null
+                break
+              fi
+
+              echo "::warning::push $tpl attempt $attempt failed, pull-rebasing"
+              git pull --rebase origin main >/tmp/rebase.log 2>&1 || true
+              cd - >/dev/null
+            done
+
+            if [ "$success" != "true" ]; then
+              FAILED="$FAILED $tpl"
+            fi
+          done
+          rm -rf "$WORKDIR"
+
+          if [ -n "$FAILED" ]; then
+            echo "::error::Cascade incomplete after 3 retries each. Failed:$FAILED"
+            exit 1
+          fi
+          if [ -n "$SKIPPED" ]; then
+            echo "Cascade complete: pinned $VERSION. Soft-skipped (no publish-image.yml):$SKIPPED"
+          else
+            echo "Cascade complete: $VERSION pinned across all manifest workspace_templates."
+          fi
@@ -25,12 +25,8 @@ name: publish-workspace-server-image
 #   staging-<sha>. Set repo variable or secret PROD_AUTO_DEPLOY_DISABLED=true
 #   to stop production rollout while keeping image publishing enabled.
 #
-# Primary ECR target: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/*
-# Optional staging tenant mirror target:
-#   004947743811.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/platform-tenant
+# ECR target: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/*
 # Required secrets: AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AUTO_SYNC_TOKEN
-# Staging ECR grants the primary SSOT-managed publisher principal repository
-# policy access, so no persistent staging AWS access keys are required.
 #
 # mc#711: Docker daemon not accessible on ubuntu-latest runner (molecule-canonical-1
 # shows client-only in `docker info` — daemon not running). DinD mount is present but
@@ -47,40 +43,18 @@ on:
 # `cancel-in-progress: false`; that is not acceptable for a workflow with a
 # production deploy job. Per-SHA image tags are immutable, and staging-latest is
 # best-effort last-writer-wins metadata.
-#
-# 2026-05-20 retrigger: run #86994 on mc#1589 merge sha 0f0f1ba2 failed at
-# setup-buildx-action with EACCES on PC2 WSL publish runner — the runner's
-# DOCKER_CONFIG=/home/hongming/.docker-ecr/ dir didn't have a buildx/certs
-# subdir writable by the container's UID 1001. Hot-patched the dir perms;
-# this chore push retriggers the workflow. Proper fix (per-runner
-# DOCKER_CONFIG owned by 1001, internal#597 --env HOME=/home/runner pattern)
-# is tracked as a CI-hygiene follow-up — not in scope here.

 permissions:
  contents: read
  packages: write

 env:
-  # SSOT-Instance-10 (#333): ECR registry triplet (account.dkr.ecr.region.amazonaws.com)
-  # sourced from org/repo var `ECR_REGISTRY` with the current prod-account literal as
-  # bootstrap fallback. When the org var is set, the fallback becomes dead code and
-  # switching accounts/regions is a one-line change at the org level (instead of
-  # touching every workflow). Pattern mirrors `vars.CP_URL || 'literal'` already in
-  # use below in this repo's staging-verify.yml.
-  IMAGE_NAME: ${{ vars.ECR_REGISTRY || '153263036946.dkr.ecr.us-east-2.amazonaws.com' }}/molecule-ai/platform
-  TENANT_IMAGE_NAME: ${{ vars.ECR_REGISTRY || '153263036946.dkr.ecr.us-east-2.amazonaws.com' }}/molecule-ai/platform-tenant
-  STAGING_TENANT_IMAGE_NAME: ${{ vars.STAGING_ECR_REGISTRY || '004947743811.dkr.ecr.us-east-2.amazonaws.com' }}/molecule-ai/platform-tenant
+  IMAGE_NAME: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/platform
+  TENANT_IMAGE_NAME: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/platform-tenant

 jobs:
  build-and-push:
-    # Dedicated publish/release lane (internal#462 / #394 / #399). This
-    # is a post-merge ship job (on: push:main) — it must NOT FIFO-compete
-    # with PR required-CI on the shared pool (PR#1350's prod image build
-    # was delayed ~25min this way). The `publish` label resolves ONLY to
-    # the reserved molecule-runner-publish-* sub-pool (config.publish.yaml,
-    # OUTSIDE the managed 1..20 range) so a merged fix's image build
-    # starts immediately while PR-CI keeps the general pool.
-    runs-on: publish
+    runs-on: ubuntu-latest
    steps:
      - name: Checkout
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
@@ -140,18 +114,6 @@ jobs:
        run: |
          echo "sha=${GITHUB_SHA::7}" >> "$GITHUB_OUTPUT"

-      # Keep Buildx state inside the job temp dir. The publish runner's
-      # inherited DOCKER_CONFIG can point at a host-owned ECR config path
-      # (/home/hongming/.docker-ecr), which caused setup-buildx-action to
-      # fail before image build with EACCES creating buildx/certs.
-      - name: Prepare writable Docker config
-        run: |
-          set -euo pipefail
-          export DOCKER_CONFIG="$RUNNER_TEMP/docker-config"
-          mkdir -p "$DOCKER_CONFIG/buildx/certs"
-          echo "DOCKER_CONFIG=$DOCKER_CONFIG" >> "$GITHUB_ENV"
-          docker buildx version
-
      # Build + push platform image (inline ECR auth — mirrors the operator-host
      # approach; credentials come from GITHUB_SECRET_AWS_ACCESS_KEY_ID /
      # GITHUB_SECRET_AWS_SECRET_ACCESS_KEY in Gitea Actions).
@@ -187,14 +149,9 @@ jobs:
            --push .

      # Build + push tenant image (Go platform + Next.js canvas in one image).
-      # Push the same build to the staging account too so fresh staging/E2E
-      # tenants can pull without cross-account ECR reads. The staging ECR repo
-      # policy trusts the primary SSOT-managed publisher principal; do not add
-      # separate persistent staging AWS access keys here.
      - name: Build & push tenant image to ECR (staging-<sha> + staging-latest)
        env:
          TENANT_IMAGE_NAME: ${{ env.TENANT_IMAGE_NAME }}
-          STAGING_TENANT_IMAGE_NAME: ${{ env.STAGING_TENANT_IMAGE_NAME }}
          TAG_SHA: staging-${{ steps.tags.outputs.sha }}
          TAG_LATEST: staging-latest
          GIT_SHA: ${{ github.sha }}
@@ -205,19 +162,8 @@ jobs:
        run: |
          set -euo pipefail
          ECR_REGISTRY="${TENANT_IMAGE_NAME%%/*}"
-          STAGING_ECR_REGISTRY="${STAGING_TENANT_IMAGE_NAME%%/*}"
          aws ecr get-login-password --region us-east-2 | \
            docker login --username AWS --password-stdin "${ECR_REGISTRY}"
-          aws ecr get-login-password --region us-east-2 | \
-            docker login --username AWS --password-stdin "${STAGING_ECR_REGISTRY}"
-
-          build_tags=(
-            --tag "${TENANT_IMAGE_NAME}:${TAG_SHA}"
-            --tag "${TENANT_IMAGE_NAME}:${TAG_LATEST}"
-            --tag "${STAGING_TENANT_IMAGE_NAME}:${TAG_SHA}"
-            --tag "${STAGING_TENANT_IMAGE_NAME}:${TAG_LATEST}"
-          )
-
          docker buildx build \
            --file ./workspace-server/Dockerfile.tenant \
            --build-arg NEXT_PUBLIC_PLATFORM_URL= \
@@ -226,7 +172,8 @@ jobs:
            --label "org.opencontainers.image.revision=${GIT_SHA}" \
            --label "org.opencontainers.image.created=$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
            --label "molecule.workflow.run_id=${GITHUB_RUN_ID}" \
-            "${build_tags[@]}" \
+            --tag "${TENANT_IMAGE_NAME}:${TAG_SHA}" \
+            --tag "${TENANT_IMAGE_NAME}:${TAG_LATEST}" \
            --push .

  # bp-exempt: production deploy side-effect; merge is gated by CI / all-required and this job waits for push CI before acting.
@@ -234,18 +181,13 @@ jobs:
    name: Production auto-deploy
    needs: build-and-push
    if: ${{ github.event_name == 'push' && github.ref == 'refs/heads/main' }}
-    # Side-effect deploy only; image publish success is the durable artifact. mc#1982
-    continue-on-error: true
-    # Publish/release lane (internal#462) — production deploy of a merged
-    # fix; reserved capacity, never queued behind PR-CI.
-    runs-on: publish
-    timeout-minutes: 90
+    runs-on: ubuntu-latest
+    timeout-minutes: 75
    env:
      CP_URL: ${{ vars.PROD_CP_URL || 'https://api.moleculesai.app' }}
      CP_ADMIN_API_TOKEN: ${{ secrets.CP_ADMIN_API_TOKEN }}
      GITEA_HOST: git.moleculesai.app
      GITEA_TOKEN: ${{ secrets.PROD_AUTO_DEPLOY_CONTROL_TOKEN || secrets.AUTO_SYNC_TOKEN }}
-      CI_STATUS_TIMEOUT_SECONDS: "3600"
      PROD_AUTO_DEPLOY_DISABLED: ${{ vars.PROD_AUTO_DEPLOY_DISABLED || secrets.PROD_AUTO_DEPLOY_DISABLED || '' }}
      PROD_AUTO_DEPLOY_CANARY_SLUG: ${{ vars.PROD_AUTO_DEPLOY_CANARY_SLUG || 'hongming' }}
      PROD_AUTO_DEPLOY_SOAK_SECONDS: ${{ vars.PROD_AUTO_DEPLOY_SOAK_SECONDS || '60' }}
@@ -304,19 +246,26 @@ jobs:
          python3 .gitea/scripts/prod-auto-deploy.py assert-enabled
          PLAN="$RUNNER_TEMP/prod-auto-deploy-plan.json"
          TARGET_TAG="$(jq -r '.target_tag' "$PLAN")"
+          BODY="$(jq -c '.body' "$PLAN")"
+
+          echo "POST $CP_URL/cp/admin/tenants/redeploy-fleet"
+          echo "  target_tag: $TARGET_TAG"
+          echo "  body: $BODY"

          HTTP_RESPONSE="$RUNNER_TEMP/prod-redeploy-response.json"
+          HTTP_CODE_FILE="$RUNNER_TEMP/prod-redeploy-http-code.txt"
          set +e
-          python3 .gitea/scripts/prod-auto-deploy.py rollout \
-            --plan "$PLAN" \
-            --response "$HTTP_RESPONSE"
-          ROLLOUT_EXIT=$?
+          curl -sS -o "$HTTP_RESPONSE" -w '%{http_code}' \
+            -m 1200 \
+            -H "Authorization: Bearer $CP_ADMIN_API_TOKEN" \
+            -H "Content-Type: application/json" \
+            -X POST "$CP_URL/cp/admin/tenants/redeploy-fleet" \
+            -d "$BODY" > "$HTTP_CODE_FILE"
          set -e

-          if [ ! -s "$HTTP_RESPONSE" ]; then
-            jq -nc --arg error "rollout command exited $ROLLOUT_EXIT before writing a response" \
-              '{ok:false, results:[], error:$error}' > "$HTTP_RESPONSE"
-          fi
+          HTTP_CODE="$(cat "$HTTP_CODE_FILE" 2>/dev/null || echo "000")"
+          [ -z "$HTTP_CODE" ] && HTTP_CODE="000"
+          echo "HTTP $HTTP_CODE"
          jq '{ok, result_count: (.results // [] | length)}' "$HTTP_RESPONSE" || true

          {
@@ -324,36 +273,23 @@ jobs:
            echo ""
            echo "**Commit:** \`${GITHUB_SHA:0:7}\`"
            echo "**Target tag:** \`$TARGET_TAG\`"
+            echo "**HTTP:** $HTTP_CODE"
            echo ""
            echo "### Per-tenant result"
            echo ""
-            echo "| Slug | Phase | SSM Status | Exit | Healthz | On target | Error present |"
-            echo "|------|-------|------------|------|---------|-----------|---------------|"
-            jq -r '.results[]? | "| \(.slug) | \(.phase) | \(.ssm_status // "-") | \(.ssm_exit_code) | \(.healthz_ok) | \(.verified_on_target) | \((.error // "") != "") |"' "$HTTP_RESPONSE" || true
-            # internal#724: stragglers are tenants enumerated but not proven
-            # on the target build. Surface them loudly — a non-empty list
-            # means the rollout did NOT fully land.
-            STRAGGLERS="$(jq -r '(.stragglers // []) | join(", ")' "$HTTP_RESPONSE")"
-            if [ -n "$STRAGGLERS" ]; then
-              echo ""
-              echo "### ⚠ Stragglers (NOT on target tag \`$TARGET_TAG\`)"
-              echo ""
-              echo "\`$STRAGGLERS\`"
-            fi
+            echo "| Slug | Phase | SSM Status | Exit | Healthz | Error present |"
+            echo "|------|-------|------------|------|---------|---------------|"
+            jq -r '.results[]? | "| \(.slug) | \(.phase) | \(.ssm_status // "-") | \(.ssm_exit_code) | \(.healthz_ok) | \((.error // "") != "") |"' "$HTTP_RESPONSE" || true
          } >> "$GITHUB_STEP_SUMMARY"

-          OK="$(jq -r '.ok' "$HTTP_RESPONSE")"
-          if [ "$OK" != "true" ]; then
-            STRAGGLERS="$(jq -r '(.stragglers // []) | join(", ")' "$HTTP_RESPONSE")"
-            if [ -n "$STRAGGLERS" ]; then
-              echo "::error::incomplete rollout — tenants not on target tag $TARGET_TAG: $STRAGGLERS"
-            fi
-            echo "::error::redeploy-fleet reported ok=false; production rollout halted."
+          if [ "$HTTP_CODE" != "200" ]; then
+            echo "::error::redeploy-fleet returned HTTP $HTTP_CODE"
            exit 1
          fi
-          if [ "$ROLLOUT_EXIT" -ne 0 ]; then
-            echo "::error::redeploy-fleet rollout failed with exit code $ROLLOUT_EXIT."
-            exit "$ROLLOUT_EXIT"
+          OK="$(jq -r '.ok' "$HTTP_RESPONSE")"
+          if [ "$OK" != "true" ]; then
+            echo "::error::redeploy-fleet reported ok=false; production rollout halted."
+            exit 1
          fi

      - name: Verify reachable tenants report this SHA
@@ -9,22 +9,10 @@
 #   Triggers on:
 #     - `pull_request_target`: opened, synchronize, reopened
 #         → initial status posts when PR opens / re-pushes
-#     - `pull_request_review` types: [submitted]
-#         → re-evaluate when a team member submits an APPROVE review so
-#           the gate flips immediately (no wait for the next push or
-#           slash-command). Verified live: sop-tier-check.yml uses this
-#           same event and provably fires (produces
-#           `sop-tier-check / tier-check (pull_request_review)` contexts).
-#           The job-level `if:` guard checks
-#           `github.event.review.state == 'APPROVED' || 'approved'` so
-#           only APPROVE reviews run the evaluator; COMMENT and
-#           REQUEST_CHANGES are skipped at the job level.
-#           Branch-protection requires the `(pull_request_target)`
-#           context variant, so the review-event path EXPLICITLY POSTS
-#           the required context via the API. Trust boundary preserved
-#           (BASE ref, no PR-head).
-#     - comment refires are handled by `sop-checklist.yml` review-refire job
-#         → `/qa-recheck` slash-command re-evaluates this gate.
+#     - comment refires are handled by `review-refire-comments.yml`
+#         → a single issue_comment dispatcher prevents every SOP/review
+#           comment from enqueueing separate qa/security/tier jobs on
+#           Gitea 1.22.6 before job-level `if:` can skip them.
 #   Workflow name = `qa-review` ; job name = `approved`.
 #   The job's own pass/fail conclusion publishes the status context
 #   `qa-review / approved (<event>)` — NO `POST /statuses` call → NO
@@ -97,26 +85,20 @@ name: qa-review
 on:
  pull_request_target:
    types: [opened, synchronize, reopened]
-  pull_request_review:
-    types: [submitted]

 permissions:
  contents: read
  pull-requests: read
-  statuses: write

 jobs:
  # bp-exempt: PR review bot signal; required merge state is enforced by CI / all-required.
  approved:
    # Gate the job:
    #   - On pull_request_target events: always run.
-    #   - On pull_request_review_approved events: run so the gate flips
-    #     immediately when a team member submits an APPROVE review.
-    # Comment-triggered refires live in sop-checklist.yml review-refire job.
+    # Comment-triggered refires live in review-refire-comments.yml. Keeping
+    # this workflow PR-only avoids comment-triggered queue storms.
    if: |
-      github.event_name == 'pull_request_target' ||
-      (github.event_name == 'pull_request_review' &&
-       (github.event.review.state == 'APPROVED' || github.event.review.state == 'approved'))
+      github.event_name == 'pull_request_target'
    runs-on: ubuntu-latest
    steps:
      - name: Privilege check (A1.1 — INFORMATIONAL log only, NOT a gate)
@@ -160,7 +142,6 @@ jobs:
          ref: ${{ github.event.repository.default_branch }}

      - name: Evaluate qa-review
-        id: eval
        env:
          GITEA_TOKEN: ${{ secrets.SOP_TIER_CHECK_TOKEN || secrets.GITHUB_TOKEN }}
          GITEA_HOST: git.moleculesai.app
@@ -175,66 +156,3 @@ jobs:
          REVIEW_CHECK_DEBUG: '0'
          REVIEW_CHECK_STRICT: '0'
        run: bash .gitea/scripts/review-check.sh
-
-      - name: Post required status context on pull_request_review
-        # Gitea Actions auto-publishes (pull_request_review) context
-        # for this event, but branch-protection requires (pull_request_target).
-        # We explicitly POST the BP-required context so the gate flips.
-        # Trust boundary: same BASE-ref script result, no PR-head code.
-        #
-        # TOKEN FIX (RC 8326): uses STATUS_POST_TOKEN (CTO-granted,
-        # msg d52cc72a). Dedicated narrow-scoped write:repository token
-        # for the explicit status POST. Evaluator step stays on
-        # SOP_TIER_CHECK_TOKEN (read-only) per deliberate security
-        # separation: eval computes, POST writes, never the same cred.
-        if: github.event_name == 'pull_request_review' && always()
-        env:
-          GITEA_TOKEN: ${{ secrets.STATUS_POST_TOKEN }}
-          GITEA_HOST: git.moleculesai.app
-          REPO: ${{ github.repository }}
-          PR_NUMBER: ${{ github.event.pull_request.number || github.event.issue.number }}
-          EVAL_OUTCOME: ${{ steps.eval.outcome }}
-        run: |
-          set -euo pipefail
-          authfile=$(mktemp)
-          chmod 600 "$authfile"
-          printf 'header = "Authorization: token %s"\n' "$GITEA_TOKEN" > "$authfile"
-
-          prfile=$(mktemp)
-          code=$(curl -sS -o "$prfile" -w '%{http_code}' -K "$authfile" \
-            "https://${GITEA_HOST}/api/v1/repos/${REPO}/pulls/${PR_NUMBER}")
-          if [ "$code" != "200" ]; then
-            echo "::error::GET /pulls/${PR_NUMBER} returned HTTP ${code}"
-            rm -f "$prfile" "$authfile"
-            exit 1
-          fi
-          head_sha=$(jq -r '.head.sha // ""' "$prfile")
-          rm -f "$prfile"
-
-          if [ "$EVAL_OUTCOME" = "success" ]; then
-            status_state="success"
-            description="Approved via pull_request_review trigger"
-          else
-            status_state="failure"
-            description="Review check failed via pull_request_review trigger"
-          fi
-
-          body=$(jq -nc \
-            --arg state "$status_state" \
-            --arg context "qa-review / approved (pull_request_target)" \
-            --arg description "$description" \
-            '{state:$state, context:$context, description:$description}')
-
-          post_code=$(curl -sS -o /dev/null -w '%{http_code}' -X POST \
-            -K "$authfile" -H "Content-Type: application/json" \
-            -d "$body" \
-            "https://${GITEA_HOST}/api/v1/repos/${REPO}/statuses/${head_sha}")
-
-          rm -f "$authfile"
-
-          if [ "$post_code" != "200" ] && [ "$post_code" != "201" ]; then
-            echo "::error::POST /statuses/${head_sha} returned HTTP ${post_code}"
-            exit 1
-          fi
-
-          echo "::notice::posted ${status_state} for context=\"qa-review / approved (pull_request_target)\" on sha=${head_sha}"
@@ -51,7 +51,7 @@ jobs:
    name: Audit Railway env vars for drift-prone pins
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 10

@@ -68,12 +68,9 @@ jobs:
  # bp-exempt: production redeploy is a side-effect workflow, not a merge gate.
  redeploy:
    if: ${{ github.event_name == 'workflow_dispatch' }}
-    # Dedicated publish/release lane (internal#462 / #394 / #399).
-    # Production tenant redeploy — a deploy action, reserved capacity so
-    # it never queues behind PR-CI. `publish` -> molecule-runner-publish-*.
-    runs-on: publish
+    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 25
    env:
@@ -151,11 +148,6 @@ jobs:
            exit 1
          fi

-          # confirm:true ack required by CP /cp/admin/tenants/redeploy-fleet
-          # contract (cp#228 / task #308) for fleet-wide intent. Empty body
-          # / {confirm:false} / {only_slugs:[]} → 400. This caller redeploys
-          # the entire prod fleet (canary + fan-out), no slug scoping, so
-          # confirm:true is correct.
          BODY=$(jq -nc \
            --arg tag "$TARGET_TAG" \
            --arg canary "$CANARY_SLUG" \
@@ -167,8 +159,7 @@ jobs:
              canary_slug: $canary,
              soak_seconds: $soak,
              batch_size: $batch,
-              dry_run: $dry,
-              confirm: true
+              dry_run: $dry
            }')

          echo "POST $CP_URL/cp/admin/tenants/redeploy-fleet"
@@ -75,12 +75,9 @@ env:
 jobs:
  # bp-exempt: post-merge staging redeploy side effect; CI / all-required gates source changes.
  redeploy:
-    # Dedicated publish/release lane (internal#462 / #394 / #399).
-    # Post-merge staging redeploy — a deploy action, reserved capacity.
-    # `publish` -> molecule-runner-publish-* sub-pool.
-    runs-on: publish
+    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 25
    steps:
@@ -123,11 +120,6 @@ jobs:
            exit 1
          fi

-          # confirm:true ack required by CP /cp/admin/tenants/redeploy-fleet
-          # contract (cp#228 / task #308) for fleet-wide intent. Empty body
-          # / {confirm:false} / {only_slugs:[]} → 400. Staging IS the
-          # canary, no slug scoping; this rolls the entire staging fleet,
-          # so confirm:true is correct.
          BODY=$(jq -nc \
            --arg tag "$TARGET_TAG" \
            --arg canary "$CANARY_SLUG" \
@@ -139,8 +131,7 @@ jobs:
              canary_slug: $canary,
              soak_seconds: $soak,
              batch_size: $batch,
-              dry_run: $dry,
-              confirm: true
+              dry_run: $dry
            }')

          echo "POST $CP_URL/cp/admin/tenants/redeploy-fleet"
@@ -54,7 +54,7 @@ jobs:
        # runners with internet access to package mirrors). Falls back to GitHub
        # binary download. GitHub releases may be blocked on some runner networks
        # (infra#241 follow-up).
-        # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+        # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
        continue-on-error: true
        run: |
          if apt-get update -qq && apt-get install -y -qq jq; then
@@ -1,16 +1,11 @@
-# DEPRECATED — superseded by `.gitea/workflows/sop-checklist.yml`.
+# Consolidated comment dispatcher for manual review/tier refires.
 #
-# The review-refire logic (qa/security/tier slash-command dispatch) has been
-# merged into sop-checklist.yml as the `review-refire` job. This workflow
-# is kept as a no-op stub to avoid a gap during the transition window where
-# this file may be deleted while sop-checklist.yml has not yet been merged.
-#
-# After sop-checklist.yml lands, this file will be deleted (issue #1280).
-#
-# Historical behavior (superseded):
 # Gitea 1.22 queues one run per workflow subscribed to `issue_comment` before
-# evaluating job-level `if:`. Previously this workflow was the single
-# non-SOP comment subscriber for qa/security/tier refire slash commands.
+# evaluating job-level `if:`. SOP-heavy PRs therefore created queue storms when
+# qa-review, security-review, sop-checklist, and sop-tier-refire all
+# listened to comments. This workflow is the single non-SOP comment subscriber:
+# ordinary comments no-op quickly; slash commands post the required status
+# contexts to the PR head SHA.

 name: review-refire-comments

@@ -23,17 +18,92 @@ permissions:
  pull-requests: read
  statuses: write

-concurrency:
-  group: ${{ github.repository }}-${{ github.workflow }}-${{ github.event.issue.number || github.ref }}
-  cancel-in-progress: true
-
 jobs:
-  # No-op stub — all refire logic moved to sop-checklist.yml review-refire job.
-  # Kept to avoid transition gap; will be deleted after sop-checklist.yml merges.
  dispatch:
    runs-on: ubuntu-latest
    steps:
-      - name: Deprecated — refire logic moved to sop-checklist.yml
+      - name: Classify comment
+        id: classify
+        env:
+          COMMENT_BODY: ${{ github.event.comment.body }}
+          IS_PR: ${{ github.event.issue.pull_request != null }}
        run: |
-          echo "::warning::review-refire-comments.yml is deprecated. Refire logic is now in sop-checklist.yml review-refire job. This workflow is a no-op stub pending deletion (issue #1280)."
-          exit 0
+          set -euo pipefail
+          {
+            echo "run_qa=false"
+            echo "run_security=false"
+            echo "run_tier=false"
+          } >> "$GITHUB_OUTPUT"
+          if [ "$IS_PR" != "true" ]; then
+            echo "::notice::not a PR comment; no-op"
+            exit 0
+          fi
+          first_line=$(printf '%s\n' "$COMMENT_BODY" | sed -n '1p')
+          case "$first_line" in
+            /qa-recheck*)
+              echo "run_qa=true" >> "$GITHUB_OUTPUT"
+              ;;
+            /security-recheck*)
+              echo "run_security=true" >> "$GITHUB_OUTPUT"
+              ;;
+            /refire-tier-check*)
+              echo "run_tier=true" >> "$GITHUB_OUTPUT"
+              ;;
+            *)
+              echo "::notice::no supported review refire slash command; no-op"
+              ;;
+          esac
+
+      - name: Check out BASE ref for trusted scripts
+        if: |
+          steps.classify.outputs.run_qa == 'true' ||
+          steps.classify.outputs.run_security == 'true' ||
+          steps.classify.outputs.run_tier == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        with:
+          ref: ${{ github.event.repository.default_branch }}
+
+      - name: Refire qa-review status
+        if: steps.classify.outputs.run_qa == 'true'
+        env:
+          GITEA_TOKEN: ${{ secrets.RFC_324_TEAM_READ_TOKEN || secrets.GITHUB_TOKEN }}
+          GITEA_HOST: git.moleculesai.app
+          REPO: ${{ github.repository }}
+          PR_NUMBER: ${{ github.event.issue.number }}
+          DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
+          TEAM: qa
+          TEAM_ID: '20'
+          REVIEW_CHECK_DEBUG: '0'
+          REVIEW_CHECK_STRICT: '0'
+          COMMENT_AUTHOR: ${{ github.event.comment.user.login }}
+        run: |
+          set -euo pipefail
+          .gitea/scripts/review-refire-status.sh
+
+      - name: Refire security-review status
+        if: steps.classify.outputs.run_security == 'true'
+        env:
+          GITEA_TOKEN: ${{ secrets.RFC_324_TEAM_READ_TOKEN || secrets.GITHUB_TOKEN }}
+          GITEA_HOST: git.moleculesai.app
+          REPO: ${{ github.repository }}
+          PR_NUMBER: ${{ github.event.issue.number }}
+          DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
+          TEAM: security
+          TEAM_ID: '21'
+          REVIEW_CHECK_DEBUG: '0'
+          REVIEW_CHECK_STRICT: '0'
+          COMMENT_AUTHOR: ${{ github.event.comment.user.login }}
+        run: |
+          set -euo pipefail
+          .gitea/scripts/review-refire-status.sh
+
+      - name: Refire sop-tier-check status
+        if: steps.classify.outputs.run_tier == 'true'
+        env:
+          GITEA_TOKEN: ${{ secrets.SOP_TIER_CHECK_TOKEN || secrets.GITHUB_TOKEN }}
+          GITEA_HOST: git.moleculesai.app
+          REPO: ${{ github.repository }}
+          PR_NUMBER: ${{ github.event.issue.number }}
+          COMMENT_AUTHOR: ${{ github.event.comment.user.login }}
+          SOP_DEBUG: '0'
+        run: bash .gitea/scripts/sop-tier-refire.sh
@@ -0,0 +1,101 @@
+name: Runtime Pin Compatibility
+
+# Ported from .github/workflows/runtime-pin-compat.yml on 2026-05-11 per
+# RFC internal#219 §1 sweep.
+#
+# Differences from the GitHub version:
+#   - Dropped `merge_group:` (no Gitea merge queue) and
+#     `workflow_dispatch:` (no inputs, but the trigger itself is
+#     parser-rejected when inputs are absent in some Gitea 1.22.x
+#     builds; safest to drop entirely — manual runs go via cron-trigger
+#     bump or push-with-paths-filter).
+#   - on.paths references .gitea/workflows/runtime-pin-compat.yml (this
+#     file) instead of the .github/ one.
+#   - Workflow-level env.GITHUB_SERVER_URL set.
+#   - `continue-on-error: true` on the job (RFC §1 contract).
+#
+# CI gate that prevents the 5-hour staging outage from 2026-04-24 from
+# recurring (controlplane#253). The original failure mode:
+#   1. molecule-ai-workspace-runtime 0.1.13 declared `a2a-sdk<1.0` in its
+#      requires_dist metadata (incorrect — it actually imports
+#      a2a.server.routes which only exists in a2a-sdk 1.0+)
+#   2. `pip install molecule-ai-workspace-runtime` resolved cleanly
+#   3. `from molecule_runtime.main import main_sync` raised ImportError
+#   4. Every tenant workspace crashed; the canary tenant caught it but
+#      only after 5 hours of degraded staging
+#
+# This workflow installs the CURRENTLY PUBLISHED runtime from PyPI on
+# top of `workspace/requirements.txt` and smoke-imports. Catches:
+#   - Upstream PyPI yanks
+#   - Bad re-releases of molecule-ai-workspace-runtime
+#   - Already-shipped wheels that stop importing because a transitive
+#     dep moved underneath
+
+on:
+  push:
+    branches: [main, staging]
+    paths:
+      # Narrow filter: pypi-latest is sensitive only to changes that
+      # affect what we're INSTALLING (requirements.txt) or WHAT THE
+      # CHECK ITSELF DOES (this workflow file). Edits to workspace/
+      # source code don't change what's on PyPI right now, so they
+      # don't change this gate's verdict.
+      - 'workspace/requirements.txt'
+      - '.gitea/workflows/runtime-pin-compat.yml'
+  pull_request:
+    branches: [main, staging]
+    paths:
+      - 'workspace/requirements.txt'
+      - '.gitea/workflows/runtime-pin-compat.yml'
+  # Daily catch for upstream PyPI publishes that break the pin combo
+  # without any change in our repo (e.g. someone re-yanks an a2a-sdk
+  # release or molecule-ai-workspace-runtime publishes a bad bump).
+  schedule:
+    - cron: '0 13 * * *'  # 06:00 PT
+
+env:
+  GITHUB_SERVER_URL: https://git.moleculesai.app
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  pypi-latest-install:
+    name: PyPI-latest install + import smoke
+    runs-on: ubuntu-latest
+    # Phase 3 (RFC #219 §1): surface broken workflows without blocking
+    # the PR. Follow-up PR flips this off after surfaced defects are
+    # triaged.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    continue-on-error: true
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: '3.11'
+          cache: pip
+          cache-dependency-path: workspace/requirements.txt
+      - name: Install runtime + workspace requirements
+        # Install order is load-bearing: install the runtime FIRST so pip
+        # honors whatever a2a-sdk constraint the runtime metadata declares
+        # (this is the surface that broke in 2026-04-24 — runtime declared
+        # `a2a-sdk<1.0` but actually needed >=1.0). The follow-up install
+        # of workspace/requirements.txt then upgrades a2a-sdk to the
+        # constraint our runtime image actually pins. The import smoke
+        # below verifies the upgraded combination is consistent.
+        run: |
+          python -m venv /tmp/venv
+          /tmp/venv/bin/pip install --upgrade pip
+          /tmp/venv/bin/pip install molecule-ai-workspace-runtime
+          /tmp/venv/bin/pip install -r workspace/requirements.txt
+          /tmp/venv/bin/pip show molecule-ai-workspace-runtime a2a-sdk \
+            | grep -E '^(Name|Version):'
+      - name: Smoke import — fail if metadata declares deps that don't satisfy real imports
+        # WORKSPACE_ID is validated at import time by platform_auth.py — EC2
+        # user-data sets it from the cloud-init template; set a placeholder
+        # here so the import smoke doesn't trip on the env-var guard.
+        env:
+          WORKSPACE_ID: 00000000-0000-0000-0000-000000000001
+        run: |
+          /tmp/venv/bin/python -c "from molecule_runtime.main import main_sync; print('runtime imports OK')"
@@ -0,0 +1,150 @@
+name: Runtime PR-Built Compatibility
+
+# Ported from .github/workflows/runtime-prbuild-compat.yml on 2026-05-11
+# per RFC internal#219 §1 sweep.
+#
+# Differences from the GitHub version:
+#   - Dropped `merge_group:` (no Gitea merge queue) and `workflow_dispatch:`
+#     (Gitea 1.22.6 parser-rejects workflow_dispatch with inputs and is
+#     finicky without them).
+#   - `dorny/paths-filter@v4` replaced with inline `git diff` (per PR#372
+#     pattern for ci.yml port).
+#   - on.paths references .gitea/workflows/runtime-prbuild-compat.yml.
+#   - Workflow-level env.GITHUB_SERVER_URL set.
+#   - `continue-on-error: true` on every job (RFC §1 contract).
+#
+# Companion to `runtime-pin-compat.yml`. That workflow tests what's
+# CURRENTLY PUBLISHED on PyPI; this workflow tests what WOULD BE
+# PUBLISHED if THIS PR merges.
+#
+# Why two workflows: the chicken-and-egg #128 fix added a "PR-built
+# wheel" job to the original runtime-pin-compat.yml, but both jobs
+# shared a `paths:` filter that was the union of their needs
+# (`workspace/**`). That meant the PyPI-latest job ran on every doc
+# edit even though the upstream PyPI artifact can't change with our
+# workspace/ source. Splitting the two means each gets a narrow
+# `paths:` filter that matches the inputs it actually depends on.
+#
+# Catches the failure mode where a PR adds an import requiring a newer
+# SDK than `workspace/requirements.txt` pins:
+#   1. Pip resolves the existing PyPI wheel + the old SDK pin -> smoke
+#      passes (it imports the OLD main.py from the wheel, not the PR's
+#      new main.py).
+#   2. Merge -> publish-runtime.yml ships a wheel WITH the new import.
+#   3. Tenant images redeploy -> all crash on first boot with ImportError.
+
+on:
+  push:
+    branches: [main, staging]
+  pull_request:
+    branches: [main, staging]
+
+env:
+  GITHUB_SERVER_URL: https://git.moleculesai.app
+
+concurrency:
+  # event_name + sha keeps PR sync and the subsequent staging push on the
+  # same SHA from cancelling each other (per feedback_concurrency_group_per_sha).
+  group: ${{ github.workflow }}-${{ github.event_name }}-${{ github.event.pull_request.head.sha || github.sha }}
+  cancel-in-progress: true
+
+jobs:
+  detect-changes:
+    runs-on: ubuntu-latest
+    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    continue-on-error: true
+    outputs:
+      wheel: ${{ steps.decide.outputs.wheel }}
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          fetch-depth: 0
+      - id: decide
+        run: |
+          # Inline replacement for dorny/paths-filter — same pattern
+          # PR#372's ci.yml port used. Diffs against the PR base or the
+          # previous push SHA, then matches against the wheel-relevant
+          # path set.
+          #
+          # NOTE: Gitea Actions does not expose github.event.before as a
+          # shell environment variable. The ${{ github.event.before }} template
+          # expression works inside YAML run: blocks but is evaluated to an
+          # empty string for push events, making the ${VAR:-fallback} always
+          # use the fallback. Use GITHUB_EVENT_BEFORE instead — it IS set in
+          # the runner's shell environment for push events.
+          BASE=""
+          if [ "${{ github.event_name }}" = "pull_request" ]; then
+            BASE="${{ github.event.pull_request.base.sha }}"
+          elif [ -n "$GITHUB_EVENT_BEFORE" ]; then
+            BASE="$GITHUB_EVENT_BEFORE"
+          fi
+          if [ -z "$BASE" ] || echo "$BASE" | grep -qE '^0+$'; then
+            # New branch or no previous SHA: treat as wheel-relevant.
+            echo "wheel=true" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          if ! timeout 30 git cat-file -e "$BASE" 2>/dev/null; then
+            git fetch --depth=1 origin "$BASE" 2>/dev/null || true
+          fi
+          if ! timeout 30 git cat-file -e "$BASE" 2>/dev/null; then
+            echo "wheel=true" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          CHANGED=$(git diff --name-only "$BASE" HEAD)
+          if echo "$CHANGED" | grep -qE '^(workspace/|scripts/build_runtime_package\.py$|scripts/wheel_smoke\.py$|\.gitea/workflows/runtime-prbuild-compat\.yml$)'; then
+            echo "wheel=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "wheel=false" >> "$GITHUB_OUTPUT"
+          fi
+
+  # ONE job (no job-level `if:`) that always runs and reports under the
+  # required-check name `PR-built wheel + import smoke`. Real work is
+  # gated per-step on `needs.detect-changes.outputs.wheel`.
+  local-build-install:
+    needs: detect-changes
+    name: PR-built wheel + import smoke
+    runs-on: ubuntu-latest
+    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    continue-on-error: true
+    steps:
+      - name: No-op pass (paths filter excluded this commit)
+        if: needs.detect-changes.outputs.wheel != 'true'
+        run: |
+          echo "No workspace/ / scripts/{build_runtime_package,wheel_smoke}.py / workflow changes — wheel gate satisfied without rebuilding."
+          echo "::notice::PR-built wheel + import smoke no-op pass (paths filter excluded this commit)."
+      - if: needs.detect-changes.outputs.wheel == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - if: needs.detect-changes.outputs.wheel == 'true'
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: '3.11'
+          cache: pip
+          cache-dependency-path: workspace/requirements.txt
+      - name: Install build tooling
+        if: needs.detect-changes.outputs.wheel == 'true'
+        run: pip install build
+      - name: Build wheel from PR source (mirrors publish-runtime.yml)
+        if: needs.detect-changes.outputs.wheel == 'true'
+        # Use a fixed test version so the wheel filename is predictable.
+        # Doesn't reach PyPI — this build is local-only for the smoke.
+        run: |
+          python scripts/build_runtime_package.py \
+            --version "0.0.0.dev0+pin-compat" \
+            --out /tmp/runtime-build
+          cd /tmp/runtime-build && python -m build
+      - name: Install built wheel + workspace requirements
+        if: needs.detect-changes.outputs.wheel == 'true'
+        run: |
+          python -m venv /tmp/venv-built
+          /tmp/venv-built/bin/pip install --upgrade pip
+          /tmp/venv-built/bin/pip install /tmp/runtime-build/dist/*.whl
+          /tmp/venv-built/bin/pip install -r workspace/requirements.txt
+          /tmp/venv-built/bin/pip show molecule-ai-workspace-runtime a2a-sdk \
+            | grep -E '^(Name|Version):'
+      - name: Smoke import the PR-built wheel
+        if: needs.detect-changes.outputs.wheel == 'true'
+        # Same script publish-runtime.yml runs against the to-be-PyPI wheel.
+        run: |
+          /tmp/venv-built/bin/python "$GITHUB_WORKSPACE/scripts/wheel_smoke.py"
@@ -57,7 +57,7 @@ jobs:
    name: Detect SECRET_PATTERNS drift
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    timeout-minutes: 5
    steps:
@@ -30,11 +30,6 @@ jobs:
  scan:
    name: Scan diff for credential-shaped strings
    runs-on: ubuntu-latest
-    # Hard CI gate — must complete or the PR is unmergable. 10-minute ceiling
-    # is generous for a diff-scan against a single SHA. If this times out, the
-    # runner is frozen and holding a slot — the step timeout triggers clean
-    # failure, releasing the runner for the next job.
-    timeout-minutes: 10
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
@@ -138,14 +133,6 @@ jobs:
            [ -z "$f" ] && continue
            [ "$f" = "$SELF_GITHUB" ] && continue
            [ "$f" = "$SELF_GITEA" ] && continue
-            # Test-fixture exclude (internal#425): the secrets-detector's OWN
-            # unit-test corpus deliberately embeds credential-SHAPED example
-            # strings to exercise the detector. Verified 2026-05-18 synthetic
-            # (fabricated ghp_* fixtures, not real). Without this the scanner
-            # self-trips on its own fixtures and fail-closes every deploy.
-            # Same rationale as the SELF_* excludes above; gate NOT weakened
-            # (all other paths still fully scanned).
-            [ "$f" = "workspace-server/internal/secrets/patterns_test.go" ] && continue
            if [ -n "$DIFF_RANGE" ]; then
              ADDED=$(git diff --no-color --unified=0 "$BASE" "$HEAD" -- "$f" 2>/dev/null | grep -E '^\+[^+]' || true)
            else
@@ -6,44 +6,24 @@
 #
 # See `qa-review.yml` header for the full A1-α / A1.1 / A4 / A5 design
 # rationale; everything below is identical in shape.
-#
-# A1-α addendum (internal#760): review-event trigger added so the security
-# gate flips immediately when a team member submits an APPROVE review.
-# Uses `pull_request_review` types: [submitted] — verified live via
-# sop-tier-check.yml which provably fires this event (produces
-# `sop-tier-check / tier-check (pull_request_review)` contexts).
-# The job-level `if:` guard checks
-# `github.event.review.state == 'APPROVED' || 'approved'` so only APPROVE
-# reviews run the evaluator; COMMENT and REQUEST_CHANGES are skipped at
-# the job level. Branch-protection requires the `(pull_request_target)`
-# context variant, so the review-event path EXPLICITLY POSTS the required
-# context via the API. Trust boundary preserved (BASE ref, no PR-head).

 name: security-review

 on:
  pull_request_target:
    types: [opened, synchronize, reopened]
-  pull_request_review:
-    types: [submitted]

 permissions:
  contents: read
  pull-requests: read
-  statuses: write

 jobs:
  # bp-exempt: PR security review bot signal; required merge state is enforced by CI / all-required.
  approved:
-    # Gate the job:
-    #   - On pull_request_target events: always run.
-    #   - On pull_request_review_approved events: run so the gate flips
-    #     immediately when a team member submits an APPROVE review.
-    # Comment-triggered refires live in sop-checklist.yml review-refire job.
+    # Comment-triggered refires live in review-refire-comments.yml. Keeping
+    # this workflow PR-only avoids comment-triggered queue storms.
    if: |
-      github.event_name == 'pull_request_target' ||
-      (github.event_name == 'pull_request_review' &&
-       (github.event.review.state == 'APPROVED' || github.event.review.state == 'approved'))
+      github.event_name == 'pull_request_target'
    runs-on: ubuntu-latest
    steps:
      - name: Privilege check (A1.1 — INFORMATIONAL log only, NOT a gate)
@@ -76,7 +56,6 @@ jobs:
          ref: ${{ github.event.repository.default_branch }}

      - name: Evaluate security-review
-        id: eval
        env:
          GITEA_TOKEN: ${{ secrets.SOP_TIER_CHECK_TOKEN || secrets.GITHUB_TOKEN }}
          GITEA_HOST: git.moleculesai.app
@@ -88,66 +67,3 @@ jobs:
          REVIEW_CHECK_DEBUG: '0'
          REVIEW_CHECK_STRICT: '0'
        run: bash .gitea/scripts/review-check.sh
-
-      - name: Post required status context on pull_request_review
-        # Gitea Actions auto-publishes (pull_request_review) context
-        # for this event, but branch-protection requires (pull_request_target).
-        # We explicitly POST the BP-required context so the gate flips.
-        # Trust boundary: same BASE-ref script result, no PR-head code.
-        #
-        # TOKEN FIX (RC 8326): uses STATUS_POST_TOKEN (CTO-granted,
-        # msg d52cc72a). Dedicated narrow-scoped write:repository token
-        # for the explicit status POST. Evaluator step stays on
-        # SOP_TIER_CHECK_TOKEN (read-only) per deliberate security
-        # separation: eval computes, POST writes, never the same cred.
-        if: github.event_name == 'pull_request_review' && always()
-        env:
-          GITEA_TOKEN: ${{ secrets.STATUS_POST_TOKEN }}
-          GITEA_HOST: git.moleculesai.app
-          REPO: ${{ github.repository }}
-          PR_NUMBER: ${{ github.event.pull_request.number || github.event.issue.number }}
-          EVAL_OUTCOME: ${{ steps.eval.outcome }}
-        run: |
-          set -euo pipefail
-          authfile=$(mktemp)
-          chmod 600 "$authfile"
-          printf 'header = "Authorization: token %s"\n' "$GITEA_TOKEN" > "$authfile"
-
-          prfile=$(mktemp)
-          code=$(curl -sS -o "$prfile" -w '%{http_code}' -K "$authfile" \
-            "https://${GITEA_HOST}/api/v1/repos/${REPO}/pulls/${PR_NUMBER}")
-          if [ "$code" != "200" ]; then
-            echo "::error::GET /pulls/${PR_NUMBER} returned HTTP ${code}"
-            rm -f "$prfile" "$authfile"
-            exit 1
-          fi
-          head_sha=$(jq -r '.head.sha // ""' "$prfile")
-          rm -f "$prfile"
-
-          if [ "$EVAL_OUTCOME" = "success" ]; then
-            status_state="success"
-            description="Approved via pull_request_review trigger"
-          else
-            status_state="failure"
-            description="Review check failed via pull_request_review trigger"
-          fi
-
-          body=$(jq -nc \
-            --arg state "$status_state" \
-            --arg context "security-review / approved (pull_request_target)" \
-            --arg description "$description" \
-            '{state:$state, context:$context, description:$description}')
-
-          post_code=$(curl -sS -o /dev/null -w '%{http_code}' -X POST \
-            -K "$authfile" -H "Content-Type: application/json" \
-            -d "$body" \
-            "https://${GITEA_HOST}/api/v1/repos/${REPO}/statuses/${head_sha}")
-
-          rm -f "$authfile"
-
-          if [ "$post_code" != "200" ] && [ "$post_code" != "201" ]; then
-            echo "::error::POST /statuses/${head_sha} returned HTTP ${post_code}"
-            exit 1
-          fi
-
-          echo "::notice::posted ${status_state} for context=\"security-review / approved (pull_request_target)\" on sha=${head_sha}"
@@ -2,20 +2,24 @@
 #
 # RFC#351 Step 2 of 6 (implementation MVP).
 #
-# === CONSOLIDATION (issue #1280) ===
+# === DESIGN ===
 #
-# This workflow is the SINGLE `issue_comment` subscriber — the logic from
-# `review-refire-comments.yml` has been merged in. Before this change:
-#   - sop-checklist.yml (pre-2026-05-16) → issue_comment:[created,edited,deleted] → runner slot used, job no-oped
-#   - review-refire-comments.yml           → issue_comment:[created]            → runner slot used, job no-oped
-#   → every non-refire comment occupied 2 runner slots for ~800 s each
-#      (~650 no-op runs/day, ~1,300 runner-slot-occupancy-hours/day).
+# Goal: each PR must answer 7 SOP-checklist questions in its body,
+# and each item must have at least one /sop-ack <slug> comment from
+# a non-author peer in the required team. BP requires the
+# `sop-checklist / all-items-acked (pull_request)` status to merge.
 #
-# Fix (PR #1345 / issue #1280):
-#   - ONE workflow, ONE issue_comment:[created] subscription (no edited/deleted)
-#   - all-items-acked job: pull_request_target OR sop slash-command comments
-#   - review-refire job: qa/security/tier refire slash commands
-#   → ~50% reduction in comment-triggered runner occupancy vs pre-fix.
+# Triggers:
+#   - `pull_request_target`: opened, edited, synchronize, reopened
+#       → fires when PR opens, body is edited (refire — RFC#351 §4),
+#         or new code is pushed (head.sha changes → stale status would
+#         be auto-discarded by BP via dismiss_stale_reviews, but the
+#         status itself is per-SHA so we re-post on the new head).
+#   - `issue_comment`: created, edited, deleted
+#       → fires on any new comment so /sop-ack / /sop-revoke take
+#         effect immediately (Gitea 1.22.6 doesn't refire on
+#         pull_request_review per feedback_pull_request_review_no_refire,
+#         so issue_comment is the canonical refire channel).
 #
 # Trust boundary (mirrors RFC#324 §A4 + sop-tier-check security note):
 #   `pull_request_target` (not `pull_request`) — workflow def is loaded
@@ -47,7 +51,7 @@
 #   /sop-ack <slug-or-numeric-alias> [optional note]
 #       — register a peer-ack for one checklist item.
 #       — slug accepts kebab-case, snake_case, or natural-spaces
-#         (all normalized to canonical kebab-case).
+#         (all normalize to canonical kebab-case).
 #       — numeric 1..7 maps via config.items[*].numeric_alias.
 #       — most-recent (user, slug) directive wins.
 #
@@ -57,13 +61,6 @@
 #       — most-recent (user, slug) directive wins, so a later /sop-ack
 #         re-restores the ack.
 #
-#   /sop-n/a <gate> [reason]
-#       — declare a gate (qa-review, security-review) N/A.
-#       — see sop-checklist-config.yaml n/a_gates section.
-#
-#   /qa-recheck /security-recheck /refire-tier-check
-#       — refire the corresponding status check on the PR head.
-#
 # The eval is read-only + idempotent (read PR + comments + team
 # membership, compute, post status). Re-running on any event is safe —
 # the new status overwrites the previous one for the same context.
@@ -73,7 +70,7 @@ name: sop-checklist
 # Cancel any in-progress runs for the same PR to prevent
 # stale runs from overwriting newer status contexts.
 concurrency:
-  group: ${{ github.repository }}-${{ github.workflow }}-${{ github.event.pull_request.number || github.event.issue.number || github.ref }}
+  group: ${{ github.repository }}-${{ github.event.pull_request.number }}
  cancel-in-progress: true

 # bp-required: yes  ← emits sop-checklist / all-items-acked (pull_request)
@@ -82,22 +79,22 @@ on:
  pull_request_target:
    types: [opened, edited, synchronize, reopened, labeled, unlabeled]
  issue_comment:
-    types: [created]   # NOT [created, edited, deleted] — Gitea 1.22.6 holds a runner slot
-    # at job-parsing time, before job-level if: guards run. edited/deleted events
-    # occupied ~1,300 runner-slot-hours/day on this workflow alone during the
-    # 2026-05-16 freeze. Per PR #1345 fix.
+    types: [created, edited, deleted]

 permissions:
  contents: read
  pull-requests: read
+  # NOTE: `statuses: write` is the GitHub-Actions name for POST /statuses.
+  # Gitea 1.22.6 may not gate on this permission key (it just checks the
+  # token), but listing it explicitly documents intent for the next
+  # platform-version upgrade.
  statuses: write
-  secrets: read

 jobs:
-  # sop-checklist gate: runs on PR lifecycle events OR sop slash commands.
-  # All other comment types (no-op text comments) no longer assign a runner
-  # because this job's if: guard short-circuits before runner assignment.
  all-items-acked:
+    # Run on pull_request_target events always. On issue_comment events,
+    # only when the comment is on a PR (issue_comment fires for issues
+    # too) and the body contains one of the slash-commands.
    if: |
      github.event_name == 'pull_request_target' ||
      (github.event_name == 'issue_comment' &&
@@ -131,95 +128,3 @@ jobs:
            --pr "$PR_NUMBER" \
            --config .gitea/sop-checklist-config.yaml \
            --gitea-host git.moleculesai.app
-
-  # bp-exempt: informational refire handler, not a merge gate. Emits
-  # qa-review/security-review status updates on /qa-recheck et al slash commands.
-  review-refire:
-    if: |
-      github.event_name == 'issue_comment' &&
-      github.event.issue.pull_request != null
-    runs-on: ubuntu-latest
-    steps:
-      - name: Classify comment
-        id: classify
-        env:
-          COMMENT_BODY: ${{ github.event.comment.body }}
-        run: |
-          set -euo pipefail
-          {
-            echo "run_qa=false"
-            echo "run_security=false"
-            echo "run_tier=false"
-          } >> "$GITHUB_OUTPUT"
-          first_line=$(printf '%s\n' "$COMMENT_BODY" | sed -n '1p')
-          case "$first_line" in
-            /qa-recheck*)
-              echo "run_qa=true" >> "$GITHUB_OUTPUT"
-              ;;
-            /security-recheck*)
-              echo "run_security=true" >> "$GITHUB_OUTPUT"
-              ;;
-            /refire-tier-check*)
-              echo "run_tier=true" >> "$GITHUB_OUTPUT"
-              ;;
-            *)
-              echo "::notice::no supported review refire slash command; no-op"
-              ;;
-          esac
-
-      - name: Check out BASE ref for trusted scripts
-        if: |
-          steps.classify.outputs.run_qa == 'true' ||
-          steps.classify.outputs.run_security == 'true' ||
-          steps.classify.outputs.run_tier == 'true'
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
-        with:
-          ref: ${{ github.event.repository.default_branch }}
-
-      - name: Refire qa-review status
-        if: steps.classify.outputs.run_qa == 'true'
-        env:
-          # Evaluator (review-check.sh + GET /pulls) stays on read-scoped token.
-          GITEA_TOKEN: ${{ secrets.SOP_TIER_CHECK_TOKEN || secrets.GITHUB_TOKEN }}
-          # Explicit POST /statuses uses narrow-scoped write:repository token.
-          STATUS_POST_TOKEN: ${{ secrets.STATUS_POST_TOKEN }}
-          GITEA_HOST: git.moleculesai.app
-          REPO: ${{ github.repository }}
-          PR_NUMBER: ${{ github.event.issue.number }}
-          DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
-          TEAM: qa
-          TEAM_ID: '20'
-          REVIEW_CHECK_DEBUG: '0'
-          REVIEW_CHECK_STRICT: '0'
-        run: |
-          set -euo pipefail
-          .gitea/scripts/review-refire-status.sh
-
-      - name: Refire security-review status
-        if: steps.classify.outputs.run_security == 'true'
-        env:
-          # Evaluator (review-check.sh + GET /pulls) stays on read-scoped token.
-          GITEA_TOKEN: ${{ secrets.SOP_TIER_CHECK_TOKEN || secrets.GITHUB_TOKEN }}
-          # Explicit POST /statuses uses narrow-scoped write:repository token.
-          STATUS_POST_TOKEN: ${{ secrets.STATUS_POST_TOKEN }}
-          GITEA_HOST: git.moleculesai.app
-          REPO: ${{ github.repository }}
-          PR_NUMBER: ${{ github.event.issue.number }}
-          DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
-          TEAM: security
-          TEAM_ID: '21'
-          REVIEW_CHECK_DEBUG: '0'
-          REVIEW_CHECK_STRICT: '0'
-        run: |
-          set -euo pipefail
-          .gitea/scripts/review-refire-status.sh
-
-      - name: Refire sop-tier-check status
-        if: steps.classify.outputs.run_tier == 'true'
-        env:
-          GITEA_TOKEN: ${{ secrets.SOP_TIER_CHECK_TOKEN || secrets.GITHUB_TOKEN }}
-          GITEA_HOST: git.moleculesai.app
-          REPO: ${{ github.repository }}
-          PR_NUMBER: ${{ github.event.issue.number }}
-          SOP_DEBUG: '0'
-        run: bash .gitea/scripts/sop-tier-refire.sh
@@ -36,7 +36,7 @@
 # window closed. continue-on-error: true has been removed from the
 # tier-check job; AND-composition is now fully enforced. If you need
 # to temporarily re-introduce a mask, file a tracker and follow the
-# mc#1982 protocol (Tier 2e lint requires a current tracker within
+# mc#774 protocol (Tier 2e lint requires a current tracker within
 # 2 lines of any continue-on-error: true).

 name: sop-tier-check
@@ -61,17 +61,12 @@ on:
  pull_request_review:
    types: [submitted, dismissed, edited]

-concurrency:
-  group: ${{ github.repository }}-${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
-  cancel-in-progress: true
-
 jobs:
  tier-check:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      pull-requests: read
-      secrets: read
    steps:
      - name: Check out base branch (for the script)
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
@@ -92,7 +87,7 @@ jobs:
        # runners). The sop-tier-check script has its own fallback as a
        # third line of defense. continue-on-error: true ensures this step
        # failing does not block the job.
-        # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+        # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
        continue-on-error: true
        run: |
          # apt-get is the primary method — Ubuntu package mirrors are reliably
@@ -113,7 +108,7 @@ jobs:
        # continue-on-error: true at step level — job-level is ignored by Gitea
        # Actions (quirk #10, internal runbooks). Belt-and-suspenders with
        # SOP_FAIL_OPEN=1 + || true below.
-        # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+        # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
        continue-on-error: true
        env:
          GITEA_TOKEN: ${{ secrets.SOP_TIER_CHECK_TOKEN || secrets.GITHUB_TOKEN }}
@@ -81,11 +81,6 @@ jobs:
      # (dead in org secret store) to CP_STAGING_ADMIN_API_TOKEN per
      # internal#322 — see this PR for the cross-workflow sweep.
      MOLECULE_ADMIN_TOKEN: ${{ secrets.CP_STAGING_ADMIN_API_TOKEN }}
-      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
-      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
-      AWS_DEFAULT_REGION: us-east-2
-      E2E_AWS_LEAK_CHECK: required
-      E2E_AWS_TERMINATE_LEAKS: '1'
      # MiniMax is the smoke's PRIMARY LLM auth path post-2026-05-04.
      # Switched from hermes+OpenAI after #2578 (the staging OpenAI key
      # account went over quota and stayed dead for 36+ hours, taking
@@ -112,9 +107,9 @@ jobs:
      E2E_RUNTIME: claude-code
      # Pin the smoke to a specific MiniMax model rather than relying
      # on the per-runtime default (which could resolve to "sonnet" →
-      # direct Anthropic and defeat the cost saving). MiniMax-M2 is the
-      # stable staging MiniMax path used by the full-SaaS smoke.
-      E2E_MODEL_SLUG: MiniMax-M2
+      # direct Anthropic and defeat the cost saving). M2.7-highspeed
+      # is "Token Plan only" but cheap-per-token and fast.
+      E2E_MODEL_SLUG: MiniMax-M2.7-highspeed
      E2E_RUN_ID: "smoke-${{ github.run_id }}"
      # Debug-only: when an operator dispatches with keep_on_failure=true,
      # the smoke script's E2E_KEEP_ORG=1 path skips teardown so the
@@ -134,12 +129,6 @@ jobs:
            echo "::error::CP_STAGING_ADMIN_API_TOKEN not set"
            exit 2
          fi
-          for var in AWS_ACCESS_KEY_ID AWS_SECRET_ACCESS_KEY; do
-            if [ -z "${!var:-}" ]; then
-              echo "::error::$var not set — EC2 leak verification cannot run"
-              exit 2
-            fi
-          done

      - name: Verify LLM key present
        run: |
@@ -75,12 +75,8 @@ permissions:
 env:
  # ECR registry (post-2026-05-06 SSOT for tenant images).
  # publish-workspace-server-image.yml pushes here.
-  # SSOT-Instance-10 (#333): triplet sourced from org/repo var `ECR_REGISTRY` with
-  # the current prod-account literal as bootstrap fallback. When the org var is set,
-  # the fallback becomes dead code and switching accounts/regions is a one-line
-  # change at the org level. Pattern mirrors `vars.CP_URL || 'literal'` below.
-  IMAGE_NAME: ${{ vars.ECR_REGISTRY || '153263036946.dkr.ecr.us-east-2.amazonaws.com' }}/molecule-ai/platform
-  TENANT_IMAGE_NAME: ${{ vars.ECR_REGISTRY || '153263036946.dkr.ecr.us-east-2.amazonaws.com' }}/molecule-ai/platform-tenant
+  IMAGE_NAME: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/platform
+  TENANT_IMAGE_NAME: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/platform-tenant
  # CP endpoint for redeploy-fleet (used in promote step below).
  CP_URL: ${{ vars.CP_URL || 'https://staging-api.moleculesai.app' }}
  GITHUB_SERVER_URL: https://git.moleculesai.app
@@ -90,7 +86,7 @@ jobs:
  staging-smoke:
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    outputs:
      sha: ${{ steps.compute.outputs.sha }}
@@ -212,7 +208,7 @@ jobs:
    if: ${{ needs.staging-smoke.result == 'success' && needs.staging-smoke.outputs.smoke_ran == 'true' }}
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    env:
      SHA: ${{ needs.staging-smoke.outputs.sha }}
@@ -239,11 +235,6 @@ jobs:
          set -euo pipefail

          TARGET_TAG="staging-${SHA}"
-          # confirm:true ack required by CP /cp/admin/tenants/redeploy-fleet
-          # contract (cp#228 / task #308) for fleet-wide intent. Empty body
-          # / {confirm:false} / {only_slugs:[]} → 400. This caller promotes
-          # the verified staging image across the entire prod fleet (canary
-          # + fan-out), no slug scoping, so confirm:true is correct.
          BODY=$(jq -nc \
            --arg tag "$TARGET_TAG" \
            --argjson soak "${SOAK_SECONDS:-120}" \
@@ -253,8 +244,7 @@ jobs:
              target_tag: $tag,
              soak_seconds: $soak,
              batch_size: $batch,
-              dry_run: $dry,
-              confirm: true
+              dry_run: $dry
            }')

          if [ -n "${CANARY_SLUG:-}" ]; then
@@ -53,12 +53,19 @@ name: status-reaper
 # `inputs:` block here. Gitea 1.22.6 rejects the whole workflow as
 # "unknown on type" when `workflow_dispatch.inputs.X` is present.
 on:
-  # Schedule moved to operator-config:
-  #   /etc/cron.d/molecule-core-status-reaper ->
-  #   /usr/local/bin/molecule-core-cron-bot.sh status-reaper
-  #
-  # This keeps the 5-minute compensation cadence but stops a maintenance
-  # bot from consuming Gitea Actions runner slots during PR merge waves.
+  # SCHEDULE RE-ENABLED 2026-05-12 rev3 — interim disable (mc#645) reverted now that
+  # rev3 widens DEFAULT_SWEEP_LIMIT 10 → 30 (covers retroactive-failure timing window).
+  # Sibling watchdog re-enabled in the same PR with timeout-minutes raised 5 → 15.
+  schedule:
+    # Every 5 minutes. Off-zero alignment with sibling cron workflows:
+    # ci-required-drift (`:17`), main-red-watchdog (`:05`),
+    # railway-pin-audit (`:23`). 5-min cadence gives a tight enough
+    # close on schedule-triggered false-reds that main-red-watchdog
+    # (hourly :05) almost never files an issue on the false case.
+    # rev3 keeps `*/5` unchanged per hongming-pc2 03:25Z review:
+    # "trades window-width-cheap for cadence-loady" — N=30 widens
+    # the lookback cheaply without doubling runner load via `*/2`.
+    - cron: '*/5 * * * *'
  workflow_dispatch:

 # Compensating-status POST needs write on repo statuses; no other
@@ -40,12 +40,14 @@ name: Sweep stale AWS Secrets Manager secrets
 # the mostly-orphan tunnels) refuses to nuke past the threshold.

 on:
-  schedule:
-    # Hourly at :30, offset from sweep-cf-orphans (:15) and
-    # sweep-cf-tunnels (:45). This janitor is intentionally schedule-only
-    # for deletes; manual dispatch is forced to dry-run below because Gitea
-    # 1.22.6 rejects workflow_dispatch.inputs.
-    - cron: '30 * * * *'
+  # Disabled as an hourly schedule until the dedicated
+  # AWS_SECRETS_JANITOR_* key exists in the key-management SSOT and is
+  # mirrored into Gitea. Falling back to the molecule-cp app principal is
+  # intentionally not allowed: it lacks account-wide ListSecrets, and
+  # granting that to an application credential would weaken least privilege.
+  #
+  # Keep the manual trigger so operators can validate the workflow immediately
+  # after provisioning the janitor key, then restore the hourly :30 schedule.
  workflow_dispatch:
 # Don't let two sweeps race the same AWS account.
 concurrency:
@@ -62,24 +64,22 @@ jobs:
  sweep:
    name: Sweep AWS Secrets Manager
    runs-on: ubuntu-latest
-    # This is a cost/leak janitor. A scheduled failure must be red so
-    # operators know tenant bootstrap secrets may be leaking.
+    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    continue-on-error: true
    # 30 min cap, mirroring the other janitors. AWS DeleteSecret is
    # fast (~0.3s/call) so even a 100+ backlog drains in seconds
    # under the 8-way xargs parallelism, but the cap is set generously
    # to leave headroom for any actual API hang.
    timeout-minutes: 30
    env:
-      # Keep this literal. Gitea/act_runner 1.22.6 can mis-render
-      # secret-backed expressions with `||`, which produced an invalid
-      # Secrets Manager endpoint in the scheduled janitor.
-      AWS_REGION: us-east-2
+      AWS_REGION: ${{ secrets.AWS_REGION || 'us-east-1' }}
      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_SECRETS_JANITOR_ACCESS_KEY_ID }}
      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRETS_JANITOR_SECRET_ACCESS_KEY }}
      CP_ADMIN_API_TOKEN: ${{ secrets.CP_ADMIN_API_TOKEN }}
      CP_STAGING_ADMIN_API_TOKEN: ${{ secrets.CP_STAGING_ADMIN_API_TOKEN }}
-      MAX_DELETE_PCT: 50
-      GRACE_HOURS: 24
+      MAX_DELETE_PCT: ${{ github.event.inputs.max_delete_pct || '50' }}
+      GRACE_HOURS: ${{ github.event.inputs.grace_hours || '24' }}

    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
@@ -114,25 +114,17 @@ jobs:

      - name: Run sweep
        if: steps.verify.outputs.skip != 'true'
-        # Schedule-vs-dispatch dry-run asymmetry:
-        #   - schedule: execute (the whole point of an hourly janitor).
-        #   - workflow_dispatch: dry-run. Gitea 1.22.6 rejects
-        #     workflow_dispatch.inputs, so there is no safe manual
-        #     "flip it to execute" toggle in this workflow.
-        # The script's MAX_DELETE_PCT gate (default 50%) remains the
-        # second line of defense regardless of trigger.
+        # Schedule-vs-dispatch dry-run asymmetry mirrors sweep-cf-tunnels:
+        #   - Scheduled: input empty → "false" → --execute (the whole
+        #     point of an hourly janitor).
+        #   - Manual workflow_dispatch: input default true → dry-run;
+        #     operator must flip it to actually delete.
        run: |
          set -euo pipefail
-          if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
+          if [ "${{ github.event.inputs.dry_run || 'false' }}" = "true" ]; then
            echo "Running in dry-run mode — no deletions"
            bash scripts/ops/sweep-aws-secrets.sh
          else
            echo "Running with --execute — will delete identified orphans"
            bash scripts/ops/sweep-aws-secrets.sh --execute
          fi
-
-      - name: Notify on sweep failure
-        if: failure()
-        run: |
-          echo "::error::sweep-aws-secrets FAILED — AWS tenant bootstrap secrets may be leaking. Check missing Gitea secrets, staging/prod CP admin tokens, AWS janitor IAM permissions, or the script safety gate."
-          exit 1
@@ -71,7 +71,7 @@ jobs:
    name: Sweep CF orphans
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    # 3 min surfaces hangs (CF API stall, AWS describe-instances stuck)
    # within one cron interval instead of burning a full tick. Realistic
@@ -55,7 +55,7 @@ jobs:
    name: Sweep CF tunnels
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    # 30 min cap. Was 5 min on the theory that the only thing that
    # could take >5min is a CF-API hang — but on 2026-05-02 a backlog
@@ -1,99 +0,0 @@
-name: sync-providers-yaml
-
-# Cross-repo canonical↔synced-copy drift gate (internal#718 P2-A, CTO
-# 2026-05-27 "Distribution = SDK via codegen + verify-CI", multi-repo branch:
-# "codegen-checked-into-each-repo + verify-CI").
-#
-# The canonical provider-registry SSOT is molecule-controlplane
-# internal/providers/providers.yaml. molecule-core has NO Go module dependency
-# on controlplane, so instead of importing it we carry a SYNCED COPY at
-# workspace-server/internal/providers/providers.yaml and gate it.
-#
-# This workflow fetches the canonical providers.yaml from controlplane (via the
-# Gitea raw endpoint, read-only) and byte-compares it against core's synced
-# copy. RED if they differ — meaning the canonical moved and core's copy must be
-# re-synced (copy verbatim + `go generate ./...` + bump
-# canonicalProvidersYAMLSHA256 in sync_canonical_test.go).
-#
-# Pairs with:
-#   * sync_canonical_test.go — hermetic sha pin (catches a hand-edit of core's
-#     copy even with no network); runs in the normal `go test ./...`.
-#   * verify-providers-gen.yml — artifact ↔ synced-copy drift.
-#
-# ENFORCEMENT GATING: standalone workflow, NOT a job in ci.yml and NOT in
-# branch protection (same soak-then-promote posture as verify-providers-gen).
-# It is intentionally absent from ci.yml's job set so the ci-required-drift
-# sentinel does not fire on it.
-#
-# AUTH: uses AUTO_SYNC_TOKEN (the existing cross-repo read token used to sync
-# template/provider content from sibling repos). If the secret is absent the
-# job emits a clear ::warning:: and exits 0 — the hermetic sha pin in
-# sync_canonical_test.go is the always-on backstop, so a missing cross-repo
-# token degrades to "hand-edit still caught, live canonical drift not caught"
-# rather than a hard red that blocks unrelated PRs.
-
-on:
-  pull_request:
-    types: [opened, synchronize, reopened]
-    paths:
-      - 'workspace-server/internal/providers/providers.yaml'
-      - '.gitea/workflows/sync-providers-yaml.yml'
-  push:
-    branches: [main, staging]
-    paths:
-      - 'workspace-server/internal/providers/providers.yaml'
-      - '.gitea/workflows/sync-providers-yaml.yml'
-  schedule:
-    # Daily at :23 — catch a canonical change in controlplane that landed
-    # without a paired core re-sync PR (off-zero to spread cron load).
-    - cron: '23 4 * * *'
-  workflow_dispatch:
-
-env:
-  GITHUB_SERVER_URL: https://git.moleculesai.app
-
-permissions:
-  contents: read
-
-concurrency:
-  group: sync-providers-yaml-${{ github.ref }}
-  cancel-in-progress: true
-
-jobs:
-  compare:
-    name: Compare synced providers.yaml against controlplane canonical
-    runs-on: ubuntu-latest
-    timeout-minutes: 6
-    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-
-      - name: Fetch canonical providers.yaml from controlplane and byte-compare
-        env:
-          AUTO_SYNC_TOKEN: ${{ secrets.AUTO_SYNC_TOKEN }}
-          API_ROOT: ${{ github.server_url }}/api/v1
-        run: |
-          set -euo pipefail
-          if [ -z "${AUTO_SYNC_TOKEN:-}" ]; then
-            echo "::warning::AUTO_SYNC_TOKEN secret missing — skipping the live cross-repo compare."
-            echo "The hermetic sha pin (sync_canonical_test.go) still gates hand-edits of core's copy."
-            echo "Provision AUTO_SYNC_TOKEN (read scope on molecule-controlplane) to enable live canonical-drift detection."
-            exit 0
-          fi
-          CANON_URL="${API_ROOT}/repos/molecule-ai/molecule-controlplane/raw/internal/providers/providers.yaml?ref=main"
-          # Use the /raw endpoint: it returns the file bytes directly. (The
-          # /contents endpoint ignores Accept: application/vnd.gitea.raw on
-          # Gitea 1.22.6 and returns the JSON+base64 envelope, which made this
-          # diff a permanent false RED.)
-          curl -fsS \
-            -H "Authorization: token ${AUTO_SYNC_TOKEN}" \
-            "${CANON_URL}" -o /tmp/canonical-providers.yaml
-          LOCAL=workspace-server/internal/providers/providers.yaml
-          if diff -u /tmp/canonical-providers.yaml "$LOCAL"; then
-            echo "OK — core's synced providers.yaml is byte-identical to the controlplane canonical."
-          else
-            echo "::error::core's synced providers.yaml DRIFTED from the controlplane canonical (SSOT)."
-            echo "Re-sync: copy controlplane internal/providers/providers.yaml verbatim over"
-            echo "  $LOCAL, run 'go generate ./...' in workspace-server/, and bump"
-            echo "  canonicalProvidersYAMLSHA256 in internal/providers/sync_canonical_test.go."
-            exit 1
-          fi
@@ -49,7 +49,7 @@ jobs:
    name: Ops scripts (unittest)
    runs-on: ubuntu-latest
    # Phase 3 (RFC #219 §1): surface broken workflows without blocking.
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
@@ -58,20 +58,14 @@ jobs:
          python-version: '3.11'
      - name: Install .gitea script test dependencies
        run: python -m pip install --quiet 'pytest==9.0.2' 'PyYAML==6.0.2'
-      - name: Run scripts/ unittests, if any
-        # Top-level scripts/ tests live alongside their target file. The
-        # runtime packaging tests moved to molecule-ai-workspace-runtime, so
-        # this pass may legitimately find no tests.
+      - name: Run scripts/ unittests (build_runtime_package, ...)
+        # Top-level scripts/ tests live alongside their target file
+        # (e.g. scripts/test_build_runtime_package.py exercises
+        # scripts/build_runtime_package.py). discover from scripts/
+        # picks up only top-level test_*.py because scripts/ops/ has
+        # no __init__.py — that's intentional, so we run two passes.
        working-directory: scripts
-        run: |
-          set +e
-          python -m unittest discover -t . -p 'test_*.py' -v
-          rc=$?
-          if [ "$rc" -eq 5 ]; then
-            echo "No top-level scripts/ unittest files found; skipping."
-            exit 0
-          fi
-          exit "$rc"
+        run: python -m unittest discover -t . -p 'test_*.py' -v
      - name: Run scripts/ops/ unittests (sweep_cf_decide, ...)
        working-directory: scripts/ops
        run: python -m unittest discover -p 'test_*.py' -v
@@ -1,107 +0,0 @@
-name: verify-providers-gen
-
-# Provider-registry SSOT enforcement gate — molecule-core side (internal#718
-# P2-A, CTO 2026-05-27 "Distribution = SDK via codegen + verify-CI").
-#
-# The canonical schema SSOT is molecule-controlplane
-# internal/providers/providers.yaml. molecule-core carries a SYNCED COPY at
-# workspace-server/internal/providers/providers.yaml (kept in sync by the
-# companion sync-providers-yaml.yml gate), and cmd/gen-providers emits the
-# checked-in Go projection workspace-server/internal/providers/gen/registry_gen.go.
-#
-# This workflow regenerates the artifact into the working tree and fails RED if
-# it differs from what is committed — catching BOTH:
-#   * a providers.yaml (synced-copy) change that wasn't followed by `go generate ./...`, and
-#   * a hand-edit of the generated artifact (it carries a DO NOT EDIT header).
-#
-# It is the molecule-core mirror of molecule-controlplane's verify-providers-gen
-# workflow. Together with sync-providers-yaml (canonical↔synced-copy drift) it
-# closes the codegen-checked-into-each-repo + verify-CI loop the RFC mandates.
-#
-# ENFORCEMENT GATING (deliberate, per dev-SOP "implementation gating"):
-# this is a STANDALONE workflow, NOT a job inside ci.yml, and is NOT yet in any
-# branch-protection status_check_contexts. Rationale (identical to the CP P0
-# rollout):
-#   * It runs + reports RED on every PR/push immediately (visible signal).
-#   * It is intentionally absent from ci.yml's job set so the ci-required-drift
-#     sentinel (jobs ↔ branch-protection ↔ audit-env) does NOT fire on it, and
-#     from branch protection (turning it into a hard merge gate has blast radius
-#     — operator GO required, same pattern as sop-tier-check / verify-providers-gen
-#     on controlplane). Promote it into branch protection in a follow-up once
-#     P2 has soaked.
-# Until then it behaves like secret-scan / block-internal-paths: a standalone
-# advisory-to-hard gate the author is expected to keep green.
-
-on:
-  pull_request:
-    types: [opened, synchronize, reopened]
-    # CI-scheduler-overload fix (fix/ci-scheduler-fanout, 2026-06-01):
-    # this gate only verifies that the generated providers artifact is in
-    # sync with the schema SSOT. Its verdict can ONLY change when one of
-    # the codegen inputs/outputs changes, so firing the Go toolchain on
-    # every unrelated PR (docs, canvas, scripts) is pure fan-out cost.
-    # Scoped to the codegen surface. SAFE because this workflow is NOT a
-    # branch-protection status_check_context (see header §ENFORCEMENT
-    # GATING) — lint-required-no-paths only forbids paths filters on
-    # REQUIRED workflows; this is advisory, so a paths filter is allowed.
-    # Mirrors the sibling sync-providers-yaml.yml scoping convention.
-    paths:
-      - 'workspace-server/internal/providers/**'
-      - 'workspace-server/cmd/gen-providers/**'
-      - '.gitea/workflows/verify-providers-gen.yml'
-  push:
-    branches: [main, staging]
-    paths:
-      - 'workspace-server/internal/providers/**'
-      - 'workspace-server/cmd/gen-providers/**'
-      - '.gitea/workflows/verify-providers-gen.yml'
-
-env:
-  GITHUB_SERVER_URL: https://git.moleculesai.app
-
-permissions:
-  contents: read
-
-concurrency:
-  group: verify-providers-gen-${{ github.ref }}
-  cancel-in-progress: true
-
-jobs:
-  verify:
-    name: Regenerate providers artifact and fail on drift
-    runs-on: ubuntu-latest
-    timeout-minutes: 8
-    defaults:
-      run:
-        working-directory: workspace-server
-    steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-      - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6.4.0
-        with:
-          go-version: 'stable'
-          cache: true
-          cache-dependency-path: workspace-server/go.sum
-
-      - name: Verify generated artifact is in sync with providers.yaml
-        run: |
-          set -euo pipefail
-          # -check regenerates in memory and byte-compares against the
-          # checked-in artifact; exit 1 (RED) on any drift. This is the
-          # single source of the gate's verdict — the same code path
-          # `go test ./cmd/gen-providers` exercises.
-          go run ./cmd/gen-providers -check
-
-      - name: Belt-and-braces — regenerate in place and assert clean tree
-        run: |
-          set -euo pipefail
-          # Independent confirmation that does not trust the -check path:
-          # actually write the artifact and assert git sees no change. If
-          # this and the step above ever disagree, the gate is suspect.
-          go generate ./...
-          if ! git diff --quiet -- internal/providers/gen/registry_gen.go; then
-            echo "::error::workspace-server/internal/providers/gen/registry_gen.go drifted from providers.yaml."
-            echo "Run 'go generate ./...' (or 'go run ./cmd/gen-providers') in workspace-server/ and commit the result."
-            git --no-pager diff -- internal/providers/gen/registry_gen.go | head -80
-            exit 1
-          fi
-          echo "OK — generated providers artifact is in sync with the schema SSOT."
@@ -31,7 +31,7 @@ jobs:
    name: Weekly Platform-Go Surface
    runs-on: ubuntu-latest
    # continue-on-error: surface only, never block
-    # mc#1982: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
+    # mc#774: pre-existing continue-on-error mask; root-fix and remove, do not renew silently.
    continue-on-error: true
    defaults:
      run:
@@ -106,7 +106,7 @@ jobs:
              [[ "$file" == *_test.go ]] && continue
              [[ "$file" == *"$path"* ]] || continue
              awk "BEGIN{exit !(\$pct < 10)}" || continue
-              rel=$(echo "$file" | sed 's|^git.moleculesai.app/molecule-ai/molecule-core/workspace-server/workspace-server/||; s|^git.moleculesai.app/molecule-ai/molecule-core/workspace-server/||')
+              rel=$(echo "$file" | sed 's|^github.com/molecule-ai/molecule-monorepo/platform/workspace-server/||; s|^github.com/molecule-ai/molecule-monorepo/platform/||')
              if echo "$ALLOWLIST" | grep -qxF "$rel"; then
                continue
              fi
@@ -0,0 +1,154 @@
+name: Block internal-flavored paths
+
+# Hard CI gate. Internal content (positioning, competitive briefs, sales
+# playbooks, PMM/press drip, draft campaigns) lives in molecule-ai/internal —
+# this public monorepo must never re-acquire those paths. CEO directive
+# 2026-04-23 after a fleet-wide audit found 79 internal files leaked here.
+#
+# Failure mode without this gate: agents (PMM, Research, DevRel, Sales) drop
+# briefs into the easiest path their cwd resolves to (root /research,
+# /marketing, /docs/marketing) and gitignore alone won't catch a `git add -f`
+# or a stale gitignore line. This workflow is the mechanical backstop.
+
+on:
+  pull_request:
+    types: [opened, synchronize, reopened]
+  push:
+    branches: [main, staging]
+  # Required for GitHub merge queue: the queue's pre-merge CI run on
+  # `gh-readonly-queue/...` refs needs this check to fire so the queue
+  # gets a real result instead of stalling forever AWAITING_CHECKS.
+  merge_group:
+    types: [checks_requested]
+
+jobs:
+  check:
+    name: Block forbidden paths
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          fetch-depth: 2  # need previous commit to diff against on push events
+
+      # For pull_request events the diff base is github.event.pull_request.base.sha,
+      # which may be many commits behind HEAD and therefore absent from the
+      # shallow clone above.  Fetch it explicitly (depth=1 keeps it fast).
+      - name: Fetch PR base SHA (pull_request events only)
+        if: github.event_name == 'pull_request'
+        run: git fetch --depth=1 origin ${{ github.event.pull_request.base.sha }}
+
+      # For merge_group events the queue's pre-merge ref is a commit on
+      # `gh-readonly-queue/...` whose parent is the queue's base_sha.
+      # That parent isn't part of the queue branch's shallow clone, so
+      # we fetch it explicitly. Mirrors the equivalent step in
+      # secret-scan.yml (#2120) — same shallow-clone bug class.
+      - name: Fetch merge_group base SHA (merge_group events only)
+        if: github.event_name == 'merge_group'
+        run: git fetch --depth=1 origin ${{ github.event.merge_group.base_sha }}
+
+      - name: Refuse if forbidden paths appear
+        env:
+          # Plumb event-specific SHAs through env so the script doesn't
+          # need conditional `${{ ... }}` interpolation per event type.
+          # github.event.before/after only exist on push events;
+          # merge_group has its own base_sha/head_sha; pull_request has
+          # pull_request.base.sha / pull_request.head.sha.
+          PR_BASE_SHA: ${{ github.event.pull_request.base.sha }}
+          PR_HEAD_SHA: ${{ github.event.pull_request.head.sha }}
+          MG_BASE_SHA: ${{ github.event.merge_group.base_sha }}
+          MG_HEAD_SHA: ${{ github.event.merge_group.head_sha }}
+          PUSH_BEFORE: ${{ github.event.before }}
+          PUSH_AFTER: ${{ github.event.after }}
+        run: |
+          # Paths that must NEVER live in the public monorepo. Add to this
+          # list narrowly — broader patterns belong in .gitignore so day-to-day
+          # docs work isn't accidentally blocked.
+          FORBIDDEN_PATTERNS=(
+            "^research/"
+            "^marketing/"
+            "^docs/marketing/"
+            "^comment-[0-9]+\.json$"
+            "^test-pmm.*\.(txt|md)$"
+            "^tick-reflections.*\.(txt|md)$"
+            ".*-temp\.(md|txt)$"
+          )
+
+          # Determine the diff base. Each event type stores its SHAs in
+          # a different place — see the env block above.
+          case "${{ github.event_name }}" in
+            pull_request)
+              BASE="$PR_BASE_SHA"
+              HEAD="$PR_HEAD_SHA"
+              ;;
+            merge_group)
+              BASE="$MG_BASE_SHA"
+              HEAD="$MG_HEAD_SHA"
+              ;;
+            *)
+              BASE="$PUSH_BEFORE"
+              HEAD="$PUSH_AFTER"
+              ;;
+          esac
+
+          # On push events with shallow clones, BASE may be present in
+          # the event payload but absent from the local object DB
+          # (fetch-depth=2 doesn't always reach the previous commit
+          # across true merges). Try fetching it on demand. If the
+          # fetch fails — e.g. the SHA was force-overwritten — we fall
+          # through to the empty-BASE branch below, which scans the
+          # entire tree as if every file were new. Correct, just slow.
+          # Same recovery shape as secret-scan.yml (#2120 — incident
+          # 2026-04-27 06:50Z block-internal-paths exit 128 with
+          # "fatal: bad object <sha>" on staging push).
+          if [ -n "$BASE" ] && ! echo "$BASE" | grep -qE '^0+$'; then
+            if ! git cat-file -e "$BASE" 2>/dev/null; then
+              git fetch --depth=1 origin "$BASE" 2>/dev/null || true
+            fi
+          fi
+
+          # Files added or modified in this change.
+          if [ -z "$BASE" ] || echo "$BASE" | grep -qE '^0+$' || ! git cat-file -e "$BASE" 2>/dev/null; then
+            # New branch / no previous SHA / BASE unreachable — check
+            # the entire tree as if every file were new. Slower but
+            # correct on first push or post-fetch-failure recovery.
+            CHANGED=$(git ls-tree -r --name-only HEAD)
+          else
+            CHANGED=$(git diff --name-only --diff-filter=AM "$BASE" "$HEAD")
+          fi
+
+          if [ -z "$CHANGED" ]; then
+            echo "No changed files to inspect."
+            exit 0
+          fi
+
+          OFFENDING=""
+          for path in $CHANGED; do
+            for pattern in "${FORBIDDEN_PATTERNS[@]}"; do
+              if echo "$path" | grep -qE "$pattern"; then
+                OFFENDING="${OFFENDING}${path} (matched: ${pattern})\n"
+                break
+              fi
+            done
+          done
+
+          if [ -n "$OFFENDING" ]; then
+            echo "::error::Forbidden internal-flavored paths detected:"
+            printf "$OFFENDING"
+            echo ""
+            echo "These paths belong in molecule-ai/internal, not this public repo."
+            echo "See docs/internal-content-policy.md for canonical locations."
+            echo ""
+            echo "If your file is genuinely public-facing (e.g. a blog post"
+            echo "ready to ship), use one of these alternatives instead:"
+            echo "  • Public-bound blog posts:  docs/blog/<slug>.md"
+            echo "  • Public-bound tutorials:   docs/tutorials/<slug>.md"
+            echo "  • Public devrel content:    docs/devrel/<slug>.md"
+            echo ""
+            echo "If you legitimately need to add a new top-level path that"
+            echo "happens to match a forbidden pattern, edit"
+            echo ".github/workflows/block-internal-paths.yml and update the"
+            echo "FORBIDDEN_PATTERNS list with reviewer signoff."
+            exit 1
+          fi
+
+          echo "✓ No forbidden paths in this change."
@@ -0,0 +1,320 @@
+name: Canary — staging SaaS smoke (every 30 min)
+
+# Minimum viable health check: provisions one Hermes workspace on a fresh
+# staging org, sends one A2A message, verifies PONG, tears down. ~8 min
+# wall clock. Pages on failure by opening a GitHub issue; auto-closes the
+# issue on the next green run.
+#
+# The full-SaaS workflow (e2e-staging-saas.yml) covers the broader surface
+# but runs only on provisioning-critical pushes + nightly — this one
+# catches drift in the 30-min window between those runs (AMI health, CF
+# cert rotation, WorkOS session stability, etc.).
+#
+# Lean mode: E2E_MODE=canary skips the child workspace + HMA memory +
+# peers/activity checks. One parent workspace + one A2A turn is enough
+# to signal "SaaS stack end-to-end is alive."
+
+on:
+  schedule:
+    # Every 30 min. Cron on GitHub-hosted runners has a known drift of
+    # a few minutes under load — that's fine for a canary.
+    - cron: '*/30 * * * *'
+  workflow_dispatch:
+    inputs:
+      keep_on_failure:
+        description: >-
+          Skip teardown when the canary fails (debugging only). The
+          tenant org + EC2 + CF tunnel + DNS stay alive so an operator
+          can SSM into the workspace EC2 and capture docker logs of the
+          failing claude-code container. REMEMBER to manually delete
+          via DELETE /cp/admin/tenants/<slug> when done so the org
+          doesn't accumulate cost. Only honored on workflow_dispatch;
+          cron runs always tear down (we don't want unattended cron
+          to leak resources).
+        type: boolean
+        default: false
+
+# Serialise with the full-SaaS workflow so they don't contend for the
+# same org-create quota on staging. Different group key from
+# e2e-staging-saas since we don't mind queueing canaries behind one
+# full run, but two canaries SHOULD queue against each other.
+concurrency:
+  group: canary-staging
+  cancel-in-progress: false
+
+permissions:
+  # Needed to open / close the alerting issue.
+  issues: write
+  contents: read
+
+jobs:
+  canary:
+    name: Canary smoke
+    runs-on: ubuntu-latest
+    # 25 min headroom over the 15-min TLS-readiness deadline in
+    # tests/e2e/test_staging_full_saas.sh (#2107). Without the buffer
+    # the job is killed at the wall-clock 15:00 mark BEFORE the bash
+    # `fail` + diagnostic burst can fire, leaving every cancellation
+    # silent. Sibling staging E2E jobs run at 20-45 min — keeping
+    # canary tighter than them so a true wedge still surfaces here
+    # first.
+    timeout-minutes: 25
+
+    env:
+      MOLECULE_CP_URL: https://staging-api.moleculesai.app
+      MOLECULE_ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+      # MiniMax is the canary's PRIMARY LLM auth path post-2026-05-04.
+      # Switched from hermes+OpenAI after #2578 (the staging OpenAI key
+      # account went over quota and stayed dead for 36+ hours, taking
+      # the canary red the entire time). claude-code template's
+      # `minimax` provider routes ANTHROPIC_BASE_URL to
+      # api.minimax.io/anthropic and reads MINIMAX_API_KEY at boot —
+      # ~5-10x cheaper per token than gpt-4.1-mini AND on a separate
+      # billing account, so OpenAI quota collapse no longer wedges the
+      # canary. Mirrors the migration continuous-synth-e2e.yml made on
+      # 2026-05-03 (#265) for the same reason. tests/e2e/test_staging_
+      # full_saas.sh branches SECRETS_JSON on which key is present —
+      # MiniMax wins when set.
+      E2E_MINIMAX_API_KEY: ${{ secrets.MOLECULE_STAGING_MINIMAX_API_KEY }}
+      # Direct-Anthropic alternative for operators who don't want to
+      # set up a MiniMax account (priority below MiniMax — first
+      # non-empty wins in test_staging_full_saas.sh's secrets-injection
+      # block). See #2578 PR comment for the rationale.
+      E2E_ANTHROPIC_API_KEY: ${{ secrets.MOLECULE_STAGING_ANTHROPIC_API_KEY }}
+      # OpenAI fallback — kept wired so an operator-dispatched run with
+      # E2E_RUNTIME=hermes overridden via workflow_dispatch can still
+      # exercise the OpenAI path without re-editing the workflow.
+      E2E_OPENAI_API_KEY: ${{ secrets.MOLECULE_STAGING_OPENAI_KEY }}
+      E2E_MODE: canary
+      E2E_RUNTIME: claude-code
+      # Pin the canary to a specific MiniMax model rather than relying
+      # on the per-runtime default (which could resolve to "sonnet" →
+      # direct Anthropic and defeat the cost saving). M2.7-highspeed
+      # is "Token Plan only" but cheap-per-token and fast.
+      E2E_MODEL_SLUG: MiniMax-M2.7-highspeed
+      E2E_RUN_ID: "canary-${{ github.run_id }}"
+      # Debug-only: when an operator dispatches with keep_on_failure=true,
+      # the canary script's E2E_KEEP_ORG=1 path skips teardown so the
+      # tenant org + EC2 stay alive for SSM-based log capture. Cron runs
+      # never set this (the input only exists on workflow_dispatch) so
+      # unattended cron always tears down. See molecule-core#129
+      # failure mode #1 — capturing the actual exception requires
+      # docker logs from the live container.
+      E2E_KEEP_ORG: ${{ github.event.inputs.keep_on_failure == 'true' && '1' || '0' }}
+
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Verify admin token present
+        run: |
+          if [ -z "$MOLECULE_ADMIN_TOKEN" ]; then
+            echo "::error::MOLECULE_STAGING_ADMIN_TOKEN not set"
+            exit 2
+          fi
+
+      - name: Verify LLM key present
+        run: |
+          # Per-runtime key check — claude-code uses MiniMax; hermes /
+          # langgraph (operator-dispatched only) use OpenAI. Hard-fail
+          # rather than soft-skip per the lesson from synth E2E #2578:
+          # an empty key silently falls through to the wrong
+          # SECRETS_JSON branch and the canary fails 5 min later with
+          # a confusing auth error instead of the clean "secret
+          # missing" message at the top.
+          case "${E2E_RUNTIME}" in
+            claude-code)
+              # Either MiniMax OR direct-Anthropic works — first
+              # non-empty wins in the test script's secrets-injection
+              # priority chain. Operators only need to set ONE of these
+              # secrets; we don't force a choice between them.
+              if [ -n "${E2E_MINIMAX_API_KEY:-}" ]; then
+                required_secret_name="MOLECULE_STAGING_MINIMAX_API_KEY"
+                required_secret_value="${E2E_MINIMAX_API_KEY}"
+              elif [ -n "${E2E_ANTHROPIC_API_KEY:-}" ]; then
+                required_secret_name="MOLECULE_STAGING_ANTHROPIC_API_KEY"
+                required_secret_value="${E2E_ANTHROPIC_API_KEY}"
+              else
+                required_secret_name="MOLECULE_STAGING_MINIMAX_API_KEY or MOLECULE_STAGING_ANTHROPIC_API_KEY"
+                required_secret_value=""
+              fi
+              ;;
+            langgraph|hermes)
+              required_secret_name="MOLECULE_STAGING_OPENAI_KEY"
+              required_secret_value="${E2E_OPENAI_API_KEY:-}"
+              ;;
+            *)
+              echo "::warning::Unknown E2E_RUNTIME='${E2E_RUNTIME}' — skipping LLM-key check"
+              required_secret_name=""
+              required_secret_value="present"
+              ;;
+          esac
+          if [ -n "$required_secret_name" ] && [ -z "$required_secret_value" ]; then
+            echo "::error::${required_secret_name} secret not set for runtime=${E2E_RUNTIME} — A2A will fail at request time with 'No LLM provider configured'"
+            exit 2
+          fi
+          echo "LLM key present ✓ (runtime=${E2E_RUNTIME}, key=${required_secret_name}, len=${#required_secret_value})"
+
+      - name: Canary run
+        id: canary
+        run: bash tests/e2e/test_staging_full_saas.sh
+
+      # Alerting: open a sticky issue on the FIRST failure; comment on
+      # subsequent failures; auto-close on next green. Comment-on-existing
+      # de-duplicates so a single open issue accumulates the streak —
+      # ops sees one issue with N comments rather than N issues.
+      #
+      # Why no consecutive-failures threshold (e.g., wait 3 runs before
+      # filing): the prior threshold check used
+      # `github.rest.actions.listWorkflowRuns()` which Gitea 1.22.6 does
+      # not expose (returns 404). On Gitea Actions the threshold call
+      # ALWAYS failed, breaking the entire alerting step and going days
+      # silent on real regressions (38h+ chronic red on 2026-05-07/08
+      # before this fix; tracked in molecule-core#129). Filing on first
+      # failure is also better UX — we want to know about the first red,
+      # not wait 90 min for it to "count." Real flakes get one issue +
+      # a quick close-on-green; persistent reds accumulate comments.
+      - name: Open issue on failure
+        if: failure()
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        with:
+          script: |
+            const title = '🔴 Canary failing: staging SaaS smoke';
+            const runURL = `${context.serverUrl}/${context.repo.owner}/${context.repo.repo}/actions/runs/${context.runId}`;
+
+            // Find an existing open canary issue (stable title match).
+            // If one exists, this isn't a "first failure" — comment and exit.
+            const { data: existing } = await github.rest.issues.listForRepo({
+              owner: context.repo.owner, repo: context.repo.repo,
+              state: 'open', labels: 'canary-staging',
+              per_page: 10,
+            });
+            const match = existing.find(i => i.title === title);
+            if (match) {
+              await github.rest.issues.createComment({
+                owner: context.repo.owner, repo: context.repo.repo,
+                issue_number: match.number,
+                body: `Canary still failing. ${runURL}`,
+              });
+              core.info(`Commented on existing issue #${match.number}`);
+              return;
+            }
+
+            // No open issue yet — file one on this first failure. The
+            // comment-on-existing branch above means subsequent failures
+            // accumulate as comments on this same issue, so we don't
+            // spam new issues per run.
+            const body =
+              `Canary run failed at ${new Date().toISOString()}.\n\n` +
+              `Run: ${runURL}\n\n` +
+              `This issue auto-closes on the next green canary run. ` +
+              `Consecutive failures add a comment here rather than a new issue.`;
+            await github.rest.issues.create({
+              owner: context.repo.owner, repo: context.repo.repo,
+              title, body,
+              labels: ['canary-staging', 'bug'],
+            });
+            core.info('Opened canary failure issue (first red)');
+
+      - name: Auto-close canary issue on success
+        if: success()
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        with:
+          script: |
+            const title = '🔴 Canary failing: staging SaaS smoke';
+            const { data: open } = await github.rest.issues.listForRepo({
+              owner: context.repo.owner, repo: context.repo.repo,
+              state: 'open', labels: 'canary-staging',
+              per_page: 10,
+            });
+            const match = open.find(i => i.title === title);
+            if (match) {
+              await github.rest.issues.createComment({
+                owner: context.repo.owner, repo: context.repo.repo,
+                issue_number: match.number,
+                body: `Canary recovered at ${new Date().toISOString()}. Closing.`,
+              });
+              await github.rest.issues.update({
+                owner: context.repo.owner, repo: context.repo.repo,
+                issue_number: match.number,
+                state: 'closed',
+              });
+              core.info(`Closed recovered canary issue #${match.number}`);
+            }
+
+      - name: Teardown safety net
+        if: always()
+        env:
+          ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+        run: |
+          set +e
+          # Slug prefix matches what test_staging_full_saas.sh emits
+          # in canary mode:
+          #   SLUG="e2e-canary-$(date +%Y%m%d)-${RUN_ID_SUFFIX}"
+          # Earlier this was `e2e-{today}-canary-` — that was the
+          # full-mode pattern (date FIRST, mode SECOND); canary slugs
+          # have mode FIRST, date SECOND. The mismatch silently
+          # never matched, leaving every cancelled-canary EC2 alive
+          # until the once-an-hour sweep eventually caught it
+          # (incident 2026-04-26 21:03Z: 1h25m EC2 leak before manual
+          # cleanup; same gap on three earlier cancellations today).
+          orgs=$(curl -sS "$MOLECULE_CP_URL/cp/admin/orgs" \
+            -H "Authorization: Bearer $ADMIN_TOKEN" 2>/dev/null \
+            | python3 -c "
+          import json, sys, os, datetime
+          run_id = os.environ.get('GITHUB_RUN_ID', '')
+          d = json.load(sys.stdin)
+          # Scope to slugs from THIS canary run when GITHUB_RUN_ID is
+          # available; the canary workflow sets E2E_RUN_ID='canary-\${run_id}'
+          # so the slug suffix is '-canary-\${run_id}-...'. Mirrors the
+          # full-mode safety net's per-run scoping (e2e-staging-saas.yml)
+          # added after the 2026-04-21 cross-run cleanup incident.
+          # Sweep both today AND yesterday's UTC dates so a run that
+          # crosses midnight still cleans up its own slug — see the
+          # 2026-04-26→27 canvas-safety-net incident.
+          today = datetime.date.today()
+          yesterday = today - datetime.timedelta(days=1)
+          dates = (today.strftime('%Y%m%d'), yesterday.strftime('%Y%m%d'))
+          if run_id:
+              prefixes = tuple(f'e2e-canary-{d}-canary-{run_id}' for d in dates)
+          else:
+              prefixes = tuple(f'e2e-canary-{d}-' for d in dates)
+          candidates = [o['slug'] for o in d.get('orgs', [])
+                        if any(o.get('slug','').startswith(p) for p in prefixes)
+                        and o.get('status') not in ('purged',)]
+          print('\n'.join(candidates))
+          " 2>/dev/null)
+          # Per-slug DELETE with HTTP-code verification. The previous
+          # `... >/dev/null || true` swallowed every failure, so a 5xx
+          # or timeout from CP looked identical to "successfully cleaned
+          # up" and the tenant kept eating ~2 vCPU until the hourly
+          # stale sweep caught it (up to 2h later). Now we capture the
+          # response code and surface non-2xx as a workflow warning, so
+          # the run page shows which slug leaked. We still don't `exit 1`
+          # on cleanup failure — a single-canary cleanup miss shouldn't
+          # fail-flag the canary itself when the actual smoke check
+          # passed. The sweep-stale-e2e-orgs cron (now every 15 min,
+          # 30-min threshold) is the safety net for whatever slips past.
+          # See molecule-controlplane#420.
+          leaks=()
+          for slug in $orgs; do
+            # Tempfile-routed -w + set +e/-e prevents curl-exit-code
+            # pollution of the captured status (lint-curl-status-capture.yml).
+            set +e
+            curl -sS -o /tmp/canary-cleanup.out -w "%{http_code}" \
+              -X DELETE "$MOLECULE_CP_URL/cp/admin/tenants/$slug" \
+              -H "Authorization: Bearer $ADMIN_TOKEN" \
+              -H "Content-Type: application/json" \
+              -d "{\"confirm\":\"$slug\"}" >/tmp/canary-cleanup.code
+            set -e
+            code=$(cat /tmp/canary-cleanup.code 2>/dev/null || echo "000")
+            if [ "$code" = "200" ] || [ "$code" = "204" ]; then
+              echo "[teardown] deleted $slug (HTTP $code)"
+            else
+              echo "::warning::canary teardown for $slug returned HTTP $code — sweep-stale-e2e-orgs will catch it within ~45 min. Body: $(head -c 300 /tmp/canary-cleanup.out 2>/dev/null)"
+              leaks+=("$slug")
+            fi
+          done
+          if [ ${#leaks[@]} -gt 0 ]; then
+            echo "::warning::canary teardown left ${#leaks[@]} leak(s): ${leaks[*]}"
+          fi
+          exit 0
@@ -0,0 +1,255 @@
+name: canary-verify
+
+# Runs the canary smoke suite against the staging canary tenant fleet
+# after a new :staging-<sha> image lands in ECR. On green, calls the
+# CP redeploy-fleet endpoint to promote :staging-<sha> → :latest so
+# the prod tenant fleet's 5-minute auto-updater picks up the verified
+# digest. On red, :latest stays on the prior known-good digest and
+# prod is untouched.
+#
+# Registry note (2026-05-10): This workflow previously used GHCR
+# (ghcr.io/molecule-ai/platform-tenant) — that registry was retired
+# during the 2026-05-06 Gitea suspension migration when publish-
+# workspace-server-image.yml switched to the operator's ECR org
+# (153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/
+# platform-tenant). The GHCR → ECR migration was never applied to
+# this file, so canary-verify was silently smoke-testing the stale
+# GHCR image while the actual staging/prod tenants ran the ECR image.
+# Result: smoke tests could not catch a broken ECR build. Fix:
+#   - Wait step: reads SHA from running canary /health (tenant-
+#     agnostic, works regardless of registry).
+#   - Promote step: calls CP redeploy-fleet endpoint with target_tag=
+#     staging-<sha>, same mechanism as redeploy-tenants-on-main.yml.
+#     No longer attempts GHCR crane ops.
+#
+# Dependencies:
+#   - publish-workspace-server-image.yml publishes :staging-<sha>
+#     to ECR on staging and main merges.
+#   - Canary tenants are configured to pull :staging-<sha> from ECR
+#     (TENANT_IMAGE env set to the ECR :staging-<sha> tag).
+#   - Repo secrets CANARY_TENANT_URLS / CANARY_ADMIN_TOKENS /
+#     CANARY_CP_SHARED_SECRET are populated.
+
+on:
+  workflow_run:
+    workflows: ["publish-workspace-server-image"]
+    types: [completed]
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  packages: write
+  actions: read
+
+env:
+  # ECR registry (post-2026-05-06 SSOT for tenant images).
+  # publish-workspace-server-image.yml pushes here.
+  IMAGE_NAME: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/platform
+  TENANT_IMAGE_NAME: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/platform-tenant
+  # CP endpoint for redeploy-fleet (used in promote step below).
+  CP_URL: ${{ vars.CP_URL || 'https://staging-api.moleculesai.app' }}
+
+jobs:
+  canary-smoke:
+    # Skip when the upstream workflow failed — no image to test against.
+    if: ${{ github.event.workflow_run.conclusion == 'success' || github.event_name == 'workflow_dispatch' }}
+    runs-on: ubuntu-latest
+    outputs:
+      sha: ${{ steps.compute.outputs.sha }}
+      smoke_ran: ${{ steps.smoke.outputs.ran }}
+    steps:
+      - name: Checkout
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Compute sha
+        id: compute
+        run: echo "sha=${GITHUB_SHA::7}" >> "$GITHUB_OUTPUT"
+
+      - name: Wait for canary tenants to pick up :staging-<sha>
+        # Poll canary health endpoints every 30s for up to 7 min instead
+        # of a fixed 6-min sleep. Exits as soon as ALL canaries report
+        # the new SHA (~2-3 min typical vs 6 min fixed). Falls back to
+        # proceeding after 7 min even if not all canaries responded —
+        # the smoke suite will catch any that didn't update.
+        #
+        # NOTE: The SHA is read from the running tenant's /health response,
+        # NOT from a registry lookup. This is registry-agnostic and works
+        # regardless of whether the tenant pulls from ECR, GHCR, or any
+        # other registry — the canary is telling us what it's actually
+        # running, which is the ground truth for smoke testing.
+        env:
+          CANARY_TENANT_URLS: ${{ secrets.CANARY_TENANT_URLS }}
+          EXPECTED_SHA: ${{ steps.compute.outputs.sha }}
+        run: |
+          if [ -z "$CANARY_TENANT_URLS" ]; then
+            echo "No canary URLs configured — falling back to 60s wait"
+            sleep 60
+            exit 0
+          fi
+          IFS=',' read -ra URLS <<< "$CANARY_TENANT_URLS"
+          MAX_WAIT=420  # 7 minutes
+          INTERVAL=30
+          ELAPSED=0
+          while [ $ELAPSED -lt $MAX_WAIT ]; do
+            ALL_READY=true
+            for url in "${URLS[@]}"; do
+              HEALTH=$(curl -s --max-time 5 "${url}/health" 2>/dev/null || echo "{}")
+              SHA=$(echo "$HEALTH" | grep -o "\"sha\":\"[^\"]*\"" | head -1 | cut -d'"' -f4)
+              if [ "$SHA" != "$EXPECTED_SHA" ]; then
+                ALL_READY=false
+                break
+              fi
+            done
+            if $ALL_READY; then
+              echo "All canaries running staging-${EXPECTED_SHA} after ${ELAPSED}s"
+              exit 0
+            fi
+            echo "Waiting for canaries... (${ELAPSED}s / ${MAX_WAIT}s)"
+            sleep $INTERVAL
+            ELAPSED=$((ELAPSED + INTERVAL))
+          done
+          echo "Timeout after ${MAX_WAIT}s — proceeding anyway (smoke suite will validate)"
+
+      - name: Run canary smoke suite
+        id: smoke
+        # Graceful-skip when no canary fleet is configured (Phase 2 not yet
+        # stood up — see molecule-controlplane/docs/canary-tenants.md).
+        # Sets `ran=false` on skip so promote-to-latest stays off (we don't
+        # want every main merge auto-promoting without gating). Manual
+        # promote-latest.yml is the release gate while canary is absent.
+        # Once the fleet is real: delete the early-exit branch.
+        env:
+          CANARY_TENANT_URLS: ${{ secrets.CANARY_TENANT_URLS }}
+          CANARY_ADMIN_TOKENS: ${{ secrets.CANARY_ADMIN_TOKENS }}
+          CANARY_CP_BASE_URL: https://staging-api.moleculesai.app
+          CANARY_CP_SHARED_SECRET: ${{ secrets.CANARY_CP_SHARED_SECRET }}
+        run: |
+          set -euo pipefail
+          if [ -z "${CANARY_TENANT_URLS:-}" ] \
+            || [ -z "${CANARY_ADMIN_TOKENS:-}" ] \
+            || [ -z "${CANARY_CP_SHARED_SECRET:-}" ]; then
+            {
+              echo "## ⚠️ canary-verify skipped"
+              echo
+              echo "One or more canary secrets are unset (\`CANARY_TENANT_URLS\`, \`CANARY_ADMIN_TOKENS\`, \`CANARY_CP_SHARED_SECRET\`)."
+              echo "Phase 2 canary fleet has not been stood up yet —"
+              echo "see [canary-tenants.md](https://git.moleculesai.app/molecule-ai/molecule-controlplane/blob/main/docs/canary-tenants.md)."
+              echo
+              echo "**Skipped — promote-to-latest will NOT auto-fire.** Dispatch \`promote-latest.yml\` manually when ready."
+            } >> "$GITHUB_STEP_SUMMARY"
+            echo "ran=false" >> "$GITHUB_OUTPUT"
+            echo "::notice::canary-verify: skipped — no canary fleet configured"
+            exit 0
+          fi
+          bash scripts/canary-smoke.sh
+          echo "ran=true" >> "$GITHUB_OUTPUT"
+
+      - name: Summary on failure
+        if: ${{ failure() }}
+        run: |
+          {
+            echo "## Canary smoke FAILED"
+            echo
+            echo "Canary tenants rejected image \`staging-${{ steps.compute.outputs.sha }}\`."
+            echo ":latest stays pinned to the prior good digest — prod is untouched."
+            echo
+            echo "Fix forward and merge again, or investigate the specific failed"
+            echo "assertions in the canary-smoke step log above."
+          } >> "$GITHUB_STEP_SUMMARY"
+
+  promote-to-latest:
+    # On green, calls the CP redeploy-fleet endpoint with target_tag=
+    # staging-<sha> to promote the verified ECR image. This is the same
+    # mechanism as redeploy-tenants-on-main.yml — no GHCR crane ops.
+    #
+    # Pre-fix history: the old GHCR promote step used `crane tag` against
+    # ghcr.io/molecule-ai/platform-tenant, but publish-workspace-server-
+    # image.yml had already migrated to ECR on 2026-05-07 (commit
+    # 10e510f5). The GHCR tags were never updated, so this step was
+    # silently promoting a stale GHCR image while actual prod tenants
+    # pulled from ECR. Canary smoke tests were GHCR-targeted and could
+    # not catch a broken ECR build.
+    needs: canary-smoke
+    if: ${{ needs.canary-smoke.result == 'success' && needs.canary-smoke.outputs.smoke_ran == 'true' }}
+    runs-on: ubuntu-latest
+    env:
+      SHA: ${{ needs.canary-smoke.outputs.sha }}
+      CP_URL: ${{ vars.CP_URL || 'https://staging-api.moleculesai.app' }}
+      # CP_ADMIN_API_TOKEN gates write access to the redeploy endpoint.
+      # Stored at the repo level so all workflows pick it up automatically.
+      CP_ADMIN_API_TOKEN: ${{ secrets.CP_ADMIN_API_TOKEN }}
+      # canary_slug pin: deploy the verified :staging-<sha> to the canary
+      # first (soak 120s), then fan out to the rest of the fleet.
+      CANARY_SLUG: ${{ vars.CANARY_PROMOTE_SLUG || '' }}
+      SOAK_SECONDS: ${{ vars.CANARY_PROMOTE_SOAK || '120' }}
+      BATCH_SIZE: ${{ vars.CANARY_PROMOTE_BATCH || '3' }}
+    steps:
+      - name: Check CP credentials
+        run: |
+          if [ -z "${CP_ADMIN_API_TOKEN:-}" ]; then
+            echo "::error::CP_ADMIN_API_TOKEN secret is not set — promote step cannot call redeploy-fleet."
+            echo "::error::Set it at: repo Settings → Actions → Variables and Secrets → New Secret."
+            exit 1
+          fi
+
+      - name: Promote verified ECR image to :latest
+        run: |
+          set -euo pipefail
+
+          TARGET_TAG="staging-${SHA}"
+          BODY=$(jq -nc \
+            --arg tag "$TARGET_TAG" \
+            --argjson soak "${SOAK_SECONDS:-120}" \
+            --argjson batch "${BATCH_SIZE:-3}" \
+            --argjson dry false \
+            '{
+              target_tag: $tag,
+              soak_seconds: $soak,
+              batch_size: $batch,
+              dry_run: $dry
+            }')
+
+          if [ -n "${CANARY_SLUG:-}" ]; then
+            BODY=$(jq '. * {canary_slug: $slug}' --arg slug "$CANARY_SLUG" <<<"$BODY")
+          fi
+
+          echo "Calling: POST $CP_URL/cp/admin/tenants/redeploy-fleet"
+          echo "  target_tag: $TARGET_TAG"
+          echo "  body: $BODY"
+
+          HTTP_RESPONSE=$(mktemp)
+          HTTP_CODE_FILE=$(mktemp)
+          set +e
+          curl -sS -o "$HTTP_RESPONSE" -w '%{http_code}' \
+            -m 1200 \
+            -H "Authorization: Bearer $CP_ADMIN_API_TOKEN" \
+            -H "Content-Type: application/json" \
+            -X POST "$CP_URL/cp/admin/tenants/redeploy-fleet" \
+            -d "$BODY" >"$HTTP_CODE_FILE"
+          CURL_EXIT=$?
+          set -e
+
+          HTTP_CODE=$(cat "$HTTP_CODE_FILE" 2>/dev/null || echo "000")
+          [ -z "$HTTP_CODE" ] && HTTP_CODE="000"
+
+          echo "HTTP $HTTP_CODE (curl exit $CURL_EXIT)"
+          cat "$HTTP_RESPONSE" | jq . || cat "$HTTP_RESPONSE"
+
+          if [ "$HTTP_CODE" -ge 400 ]; then
+            echo "::error::CP redeploy-fleet returned HTTP $HTTP_CODE — refusing to proceed."
+            exit 1
+          fi
+
+      - name: Summary
+        run: |
+          {
+            echo "## Canary verified — :latest promoted via CP redeploy-fleet"
+            echo ""
+            echo "- **Target tag:** \`staging-${{ needs.canary-smoke.outputs.sha }}\`"
+            echo "- **Registry:** ECR (\`${TENANT_IMAGE_NAME}\`)"
+            echo "- **Canary slug:** \`${CANARY_SLUG:-<none>}\` (soak ${SOAK_SECONDS}s)"
+            echo "- **Batch size:** ${BATCH_SIZE:-3}"
+            echo ""
+            echo "CP redeploy-fleet is rolling out the verified image across the prod fleet."
+            echo "The fleet's 5-minute health-check loop will pick up the update automatically."
+          } >> "$GITHUB_STEP_SUMMARY"
@@ -0,0 +1,39 @@
+name: cascade-list-drift-gate
+
+# Structural gate: TEMPLATES list in publish-runtime.yml must match
+# manifest.json's workspace_templates exactly. Closes the recurrence
+# path of PR #2556 (the data fix) and is the first concrete deliverable
+# of RFC #388 PR-3.
+#
+# Why a gate, not just discipline: PR #2536 pruned the manifest, but the
+# cascade list wasn't updated for ~weeks before someone (PR #2556)
+# noticed during an unrelated audit. During that window, codex never
+# rebuilt on a runtime publish. A structural gate catches the drift
+# the same day either file changes.
+#
+# Triggers narrowly to keep CI quiet: only on PRs that actually change
+# one of the two files. The path-filtered split + always-emit-result
+# pattern (memory: "Required check names need a job that always runs")
+# is unnecessary here because the workflow IS the check name and PR
+# branch protection should require it directly. Future-proof: if this
+# becomes a required check, add a no-op aggregator with always() so the
+# name still emits when paths don't match.
+
+on:
+  pull_request:
+    branches: [staging, main]
+    paths:
+      - manifest.json
+      - .github/workflows/publish-runtime.yml
+      - scripts/check-cascade-list-vs-manifest.sh
+
+permissions:
+  contents: read
+
+jobs:
+  check:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - name: Check cascade list matches manifest
+        run: bash scripts/check-cascade-list-vs-manifest.sh
@@ -0,0 +1,58 @@
+name: Check migration collisions
+
+# Hard gate (#2341): fails a PR that adds a migration prefix already
+# claimed by the base branch or another open PR. Caught manually 2026-04-30
+# during PR #2276 rebase: 044_runtime_image_pins collided with
+# 044_platform_inbound_secret from RFC #2312. This workflow makes that
+# check automatic.
+#
+# Trigger model: pull_request only — there's no value running this on
+# pushes to staging or main (those are post-merge; the gate must fire
+# pre-merge to be useful). Path filter scopes to PRs that actually touch
+# migrations.
+
+on:
+  pull_request:
+    types: [opened, synchronize, reopened]
+    paths:
+      - 'workspace-server/migrations/**'
+      - 'scripts/ops/check_migration_collisions.py'
+      - '.github/workflows/check-migration-collisions.yml'
+
+permissions:
+  contents: read
+  # gh pr list/diff need read access to other PRs
+  pull-requests: read
+
+jobs:
+  check:
+    name: Migration version collision check
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          # Need history to diff against base ref
+          fetch-depth: 0
+
+      - name: Detect collisions
+        env:
+          PR_NUMBER: ${{ github.event.pull_request.number }}
+          BASE_REF: origin/${{ github.event.pull_request.base.ref }}
+          HEAD_REF: ${{ github.event.pull_request.head.sha }}
+          GITHUB_REPOSITORY: ${{ github.repository }}
+          # gh CLI uses GH_TOKEN from env
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          # Ensure the named base ref exists locally. checkout@v4 with
+          # fetch-depth=0 pulls full history, but the explicit fetch is
+          # cheap insurance against form-of-ref differences across runs.
+          #
+          # IMPORTANT: do NOT pass --depth=1 here. The script below uses
+          # `git diff origin/<base>...<head>` (three-dot, merge-base form),
+          # which fails with "fatal: no merge base" if the base ref is
+          # shallow. The auto-promote staging→main PR (#2361) was blocked
+          # by exactly this for ~5h on 2026-04-30 — the depth=1 fetch
+          # overwrote checkout@v4's full-history clone with a shallow tip.
+          git fetch origin "${{ github.event.pull_request.base.ref }}" || true
+          python3 scripts/ops/check_migration_collisions.py
@@ -0,0 +1,443 @@
+name: CI
+
+on:
+  push:
+    branches: [main, staging]
+  pull_request:
+    branches: [main, staging]
+  # GitHub merge queue fires `merge_group` for the queue's pre-merge CI run.
+  # Required so the queue gets a real check result instead of a false-green
+  # from the absence of a triggered workflow. Safe to add unconditionally —
+  # the event simply doesn't fire until the queue is enabled on the branch.
+  merge_group:
+    types: [checks_requested]
+
+# Cancel in-progress CI runs when a new commit arrives on the same ref.
+# This prevents stale runs from queuing behind each other. The merge_group
+# refs (refs/heads/gh-readonly-queue/...) get their own concurrency group
+# automatically because github.ref differs from the PR ref.
+concurrency:
+  group: ci-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  # Detect which paths changed so downstream jobs can skip when only
+  # docs/markdown files were modified.
+  changes:
+    name: Detect changes
+    runs-on: ubuntu-latest
+    outputs:
+      platform: ${{ steps.check.outputs.platform }}
+      canvas: ${{ steps.check.outputs.canvas }}
+      python: ${{ steps.check.outputs.python }}
+      scripts: ${{ steps.check.outputs.scripts }}
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          fetch-depth: 0
+      - id: check
+        run: |
+          # For PR events: diff against the base branch (not HEAD~1 of the branch,
+          # which may be unrelated after force-pushes). When a push updates a PR,
+          # both pull_request and push events fire — prefer the PR base so that
+          # the diff is always computed against the actual merge base, not the
+          # previous SHA on the branch which may be on a different history line.
+          BASE="${GITHUB_BASE_REF:-${{ github.event.before }}}"
+          # GITHUB_BASE_REF is set by GitHub for PR events (the base branch name).
+          # For pull_request events we use the stored base.sha; for push events
+          # (or when base.sha is unavailable) fall back to github.event.before.
+          if [ "${{ github.event_name }}" = "pull_request" ] && [ -n "${{ github.event.pull_request.base.sha }}" ]; then
+            BASE="${{ github.event.pull_request.base.sha }}"
+          fi
+          # Fallback: if BASE is empty or all zeros (new branch), run everything
+          if [ -z "$BASE" ] || echo "$BASE" | grep -qE '^0+$'; then
+            echo "platform=true" >> "$GITHUB_OUTPUT"
+            echo "canvas=true" >> "$GITHUB_OUTPUT"
+            echo "python=true" >> "$GITHUB_OUTPUT"
+            echo "scripts=true" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          DIFF=$(git diff --name-only "$BASE" HEAD 2>/dev/null || echo ".github/workflows/ci.yml")
+          echo "platform=$(echo "$DIFF" | grep -qE '^workspace-server/|^\.github/workflows/ci\.yml$' && echo true || echo false)" >> "$GITHUB_OUTPUT"
+          echo "canvas=$(echo "$DIFF" | grep -qE '^canvas/|^\.github/workflows/ci\.yml$' && echo true || echo false)" >> "$GITHUB_OUTPUT"
+          echo "python=$(echo "$DIFF" | grep -qE '^workspace/|^\.github/workflows/ci\.yml$' && echo true || echo false)" >> "$GITHUB_OUTPUT"
+          echo "scripts=$(echo "$DIFF" | grep -qE '^tests/e2e/|^scripts/|^infra/scripts/|^\.github/workflows/ci\.yml$' && echo true || echo false)" >> "$GITHUB_OUTPUT"
+
+  # Platform (Go) is a required check on staging. Always-run + per-step
+  # gating (see Canvas (Next.js) for the rationale and the failure mode
+  # this avoids).
+  platform-build:
+    name: Platform (Go)
+    needs: changes
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        working-directory: workspace-server
+    steps:
+      - if: needs.changes.outputs.platform != 'true'
+        working-directory: .
+        run: echo "No platform/** changes — skipping real build steps; this job always runs to satisfy the required-check name on branch protection."
+      - if: needs.changes.outputs.platform == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - if: needs.changes.outputs.platform == 'true'
+        uses: actions/setup-go@40f1582b2485089dde7abd97c1529aa768e1baff # v5
+        with:
+          go-version: 'stable'
+      - if: needs.changes.outputs.platform == 'true'
+        run: go mod download
+      - if: needs.changes.outputs.platform == 'true'
+        run: go build ./cmd/server
+      # CLI (molecli) moved to standalone repo: github.com/molecule-ai/molecule-cli
+      - if: needs.changes.outputs.platform == 'true'
+        run: go vet ./... || true
+      - if: needs.changes.outputs.platform == 'true'
+        name: Run golangci-lint
+        run: golangci-lint run --timeout 3m ./... || true
+      - if: needs.changes.outputs.platform == 'true'
+        name: Run tests with race detection and coverage
+        run: go test -race -coverprofile=coverage.out ./...
+
+      - if: needs.changes.outputs.platform == 'true'
+        name: Per-file coverage report
+        # Advisory — lists every source file with its coverage so reviewers
+        # can see at-a-glance where gaps are. Sorted ascending so the worst
+        # offenders float to the top. Does NOT fail the build; the hard
+        # gate is the threshold check below. (#1823)
+        run: |
+          echo "=== Per-file coverage (worst first) ==="
+          go tool cover -func=coverage.out \
+            | grep -v '^total:' \
+            | awk '{file=$1; sub(/:[0-9][0-9.]*:.*/, "", file); pct=$NF; gsub(/%/,"",pct); s[file]+=pct; c[file]++}
+                   END {for (f in s) printf "%6.1f%%  %s\n", s[f]/c[f], f}' \
+            | sort -n
+
+      - if: needs.changes.outputs.platform == 'true'
+        name: Check coverage thresholds
+        # Enforces two gates from #1823 Layer 1:
+        #   1. Total floor (25% — ratchet plan in COVERAGE_FLOOR.md).
+        #   2. Per-file floor — non-test .go files in security-critical
+        #      paths with coverage <10% fail the build, UNLESS the file
+        #      path is listed in .coverage-allowlist.txt (acknowledged
+        #      historical debt with a tracking issue + expiry).
+        run: |
+          set -e
+          TOTAL_FLOOR=25
+          # Security-critical paths where a 0%-coverage file is a real risk.
+          CRITICAL_PATHS=(
+            "internal/handlers/tokens"
+            "internal/handlers/workspace_provision"
+            "internal/handlers/a2a_proxy"
+            "internal/handlers/registry"
+            "internal/handlers/secrets"
+            "internal/middleware/wsauth"
+            "internal/crypto"
+          )
+
+          TOTAL=$(go tool cover -func=coverage.out | grep '^total:' | awk '{print $3}' | sed 's/%//')
+          echo "Total coverage: ${TOTAL}%"
+          if awk "BEGIN{exit !($TOTAL < $TOTAL_FLOOR)}"; then
+            echo "::error::Total coverage ${TOTAL}% is below the ${TOTAL_FLOOR}% floor. See COVERAGE_FLOOR.md for ratchet plan."
+            exit 1
+          fi
+
+          # Aggregate per-file coverage → /tmp/perfile.txt: "<fullpath> <pct>"
+          go tool cover -func=coverage.out \
+            | grep -v '^total:' \
+            | awk '{file=$1; sub(/:[0-9][0-9.]*:.*/, "", file); pct=$NF; gsub(/%/,"",pct); s[file]+=pct; c[file]++}
+                   END {for (f in s) printf "%s %.1f\n", f, s[f]/c[f]}' \
+            > /tmp/perfile.txt
+
+          # Build allowlist — paths relative to workspace-server, one per line.
+          # Lines starting with # are comments.
+          ALLOWLIST=""
+          if [ -f ../.coverage-allowlist.txt ]; then
+            ALLOWLIST=$(grep -vE '^(#|[[:space:]]*$)' ../.coverage-allowlist.txt || true)
+          fi
+
+          FAILED=0
+          WARNED=0
+          for path in "${CRITICAL_PATHS[@]}"; do
+            while read -r file pct; do
+              [[ "$file" == *_test.go ]] && continue
+              [[ "$file" == *"$path"* ]] || continue
+              awk "BEGIN{exit !($pct < 10)}" || continue
+
+              # Strip the package-import prefix so we can match .coverage-allowlist.txt
+              # entries written as paths relative to workspace-server/.
+              # Handle both module paths: platform/workspace-server/... and platform/...
+              rel=$(echo "$file" | sed 's|^github.com/molecule-ai/molecule-monorepo/platform/workspace-server/||; s|^github.com/molecule-ai/molecule-monorepo/platform/||')
+
+              if echo "$ALLOWLIST" | grep -qxF "$rel"; then
+                echo "::warning file=workspace-server/$rel::Critical file at ${pct}% coverage (allowlisted, #1823) — fix before expiry."
+                WARNED=$((WARNED+1))
+              else
+                echo "::error file=workspace-server/$rel::Critical file at ${pct}% coverage — must be >=10% (target 80%). See #1823. To acknowledge as known debt, add this path to .coverage-allowlist.txt."
+                FAILED=$((FAILED+1))
+              fi
+            done < /tmp/perfile.txt
+          done
+
+          echo ""
+          echo "Critical-path check: $FAILED new failures, $WARNED allowlisted warnings."
+
+          if [ "$FAILED" -gt 0 ]; then
+            echo ""
+            echo "$FAILED security-critical file(s) have <10% test coverage and are"
+            echo "NOT in the allowlist. These paths handle auth, tokens, secrets, or"
+            echo "workspace provisioning — a 0% file here is the exact gap that let"
+            echo "CWE-22, CWE-78, KI-005 slip through in past incidents. Either:"
+            echo "  (a) add tests to raise coverage above 10%, or"
+            echo "  (b) add the path to .coverage-allowlist.txt with an expiry date"
+            echo "      and a tracking issue reference."
+            exit 1
+          fi
+
+  # Canvas (Next.js) — required check, always runs. See platform-build
+  # comment above for the rationale.
+  #
+  # Supersedes the canvas-build-noop pattern attempted in PR #2321: two
+  # jobs sharing `name:` doesn't actually satisfy branch protection
+  # because the SKIPPED check run sibling is treated as not-passed
+  # regardless of how many SUCCESS siblings it has. Verified empirically
+  # on PR #2314 — mergeStateStatus stayed BLOCKED until I collapsed to
+  # a single-job-with-conditional-steps shape.
+  canvas-build:
+    name: Canvas (Next.js)
+    needs: changes
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        working-directory: canvas
+    steps:
+      - if: needs.changes.outputs.canvas != 'true'
+        working-directory: .
+        run: echo "No canvas/** changes — skipping real build steps; this job always runs to satisfy the required-check name on branch protection."
+      - if: needs.changes.outputs.canvas == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - if: needs.changes.outputs.canvas == 'true'
+        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
+        with:
+          node-version: '22'
+      - if: needs.changes.outputs.canvas == 'true'
+        run: rm -f package-lock.json && npm install
+      - if: needs.changes.outputs.canvas == 'true'
+        run: npm run build
+      - if: needs.changes.outputs.canvas == 'true'
+        name: Run tests with coverage
+        # Coverage instrumentation is configured in canvas/vitest.config.ts
+        # (provider: v8, reporters: text + html + json-summary). Step 2 of
+        # #1815 — wires coverage into CI so we get a baseline visible on
+        # every PR. No threshold gate yet; thresholds dial in (Step 3, also
+        # tracked in #1815) after the team sees what current coverage is.
+        # Per the inline comment in vitest.config.ts: "first land
+        # observability so we can see the baseline, then dial in
+        # thresholds + a hard gate" — this PR ships the observability half.
+        run: npx vitest run --coverage
+      - name: Upload coverage summary as artifact
+        if: needs.changes.outputs.canvas == 'true' && always()
+        # Pinned to v3 for Gitea act_runner v0.6 compatibility — v4+ uses
+        # the GHES 3.10+ artifact protocol that Gitea 1.22.x does NOT
+        # implement, surfacing as `GHESNotSupportedError: @actions/artifact
+        # v2.0.0+, upload-artifact@v4+ and download-artifact@v4+ are not
+        # currently supported on GHES`. Drop this pin when Gitea ships
+        # the v4 protocol (tracked: post-Gitea-1.23 followup).
+        uses: actions/upload-artifact@c6a366c94c3e0affe28c06c8df20a878f24da3cf # v3.2.2
+        with:
+          name: canvas-coverage-${{ github.run_id }}
+          path: canvas/coverage/
+          retention-days: 7
+          if-no-files-found: warn
+
+  # MCP Server + SDK removed from CI — now in standalone repos:
+  # - github.com/molecule-ai/molecule-mcp-server (npm CI)
+  # - github.com/molecule-ai/molecule-sdk-python (PyPI CI)
+
+  # e2e-api job moved to .github/workflows/e2e-api.yml (issue #458).
+  # It now has workflow-level concurrency (cancel-in-progress: false) so
+  # new pushes queue the E2E run rather than cancelling it at the run level.
+
+  # Shellcheck (E2E scripts) — required check, always runs. See
+  # platform-build for the rationale.
+  shellcheck:
+    name: Shellcheck (E2E scripts)
+    needs: changes
+    runs-on: ubuntu-latest
+    steps:
+      - if: needs.changes.outputs.scripts != 'true'
+        run: echo "No tests/e2e/ or infra/scripts/ changes — skipping real shellcheck; this job always runs to satisfy the required-check name on branch protection."
+      - if: needs.changes.outputs.scripts == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - if: needs.changes.outputs.scripts == 'true'
+        name: Run shellcheck on tests/e2e/*.sh and infra/scripts/*.sh
+        # shellcheck is pre-installed on ubuntu-latest runners (via apt).
+        # infra/scripts/ is included because setup.sh + nuke.sh gate the
+        # README quickstart — a shellcheck regression there silently breaks
+        # new-user onboarding. scripts/ is intentionally excluded until its
+        # pre-existing SC3040/SC3043 warnings are cleaned up.
+        run: |
+          find tests/e2e infra/scripts -type f -name '*.sh' -print0 \
+            | xargs -0 shellcheck --severity=warning
+
+      - if: needs.changes.outputs.scripts == 'true'
+        name: Lint cleanup-trap hygiene (RFC #2873)
+        # Asserts every shell E2E test that calls `mktemp` also installs
+        # an EXIT trap. Catches the /tmp-leak class — a missing trap
+        # silently leaks scratch into CI runners (~10-100KB per run).
+        # See tests/e2e/lint_cleanup_traps.sh for the rule + fix pattern.
+        run: bash tests/e2e/lint_cleanup_traps.sh
+
+      - if: needs.changes.outputs.scripts == 'true'
+        name: Run E2E bash unit tests (no live infra)
+        # Pure-bash unit tests for E2E helper libs (lib/*.sh). These pin
+        # behavior of dispatch logic that — when broken — silently masks as
+        # "Could not resolve authentication method" only after a successful
+        # tenant + workspace provision (PR #2571 incident, 2026-05-03). Add
+        # new self-contained unit tests here as the lib/ directory grows;
+        # tests requiring live CP/tenant credentials belong in the dedicated
+        # e2e-staging-* workflows, not this job.
+        run: |
+          bash tests/e2e/test_model_slug.sh
+
+  canvas-deploy-reminder:
+    name: Canvas Deploy Reminder
+    runs-on: ubuntu-latest
+    needs: [changes, canvas-build]
+    # Only fires on direct pushes to main (i.e. after staging→main promotion).
+    if: needs.changes.outputs.canvas == 'true' && github.event_name == 'push' && github.ref == 'refs/heads/main'
+    steps:
+      - name: Write deploy reminder to step summary
+        env:
+          COMMIT_SHA: ${{ github.sha }}
+          RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+        run: |
+          # Write body to a temp file — avoids backtick escaping in shell.
+          cat > /tmp/deploy-reminder.md << 'BODY'
+          ## Canvas build passed ✅ — deploy required
+
+          The `publish-canvas-image` workflow is now building a fresh Docker image
+          (`ghcr.io/molecule-ai/canvas:latest`) in the background.
+
+          Once it completes (~3–5 min), apply on the host machine with:
+          ```bash
+          cd <runner-workspace>
+          git pull origin main
+          docker compose pull canvas && docker compose up -d canvas
+          ```
+
+          If you need to rebuild from local source instead (e.g. testing unreleased
+          changes or a new `NEXT_PUBLIC_*` URL), use:
+          ```bash
+          docker compose build canvas && docker compose up -d canvas
+          ```
+          BODY
+          printf '\n> Posted automatically by CI · commit `%s` · [build log](%s)\n' \
+            "$COMMIT_SHA" "$RUN_URL" >> /tmp/deploy-reminder.md
+
+          # Gitea has no commit-comments API (no equivalent of
+          # POST /repos/{owner}/{repo}/commits/{commit_sha}/comments).
+          # Write to GITHUB_STEP_SUMMARY instead — both GitHub Actions and
+          # Gitea Actions render this as the workflow run's summary page,
+          # which is where operators look for post-deploy action items.
+          # (#75 / PR-D)
+          cat /tmp/deploy-reminder.md >> "$GITHUB_STEP_SUMMARY"
+
+  # Python Lint & Test — required check, always runs. See platform-build
+  # for the rationale.
+  python-lint:
+    name: Python Lint & Test
+    needs: changes
+    runs-on: ubuntu-latest
+    env:
+      WORKSPACE_ID: test
+    defaults:
+      run:
+        working-directory: workspace
+    steps:
+      - if: needs.changes.outputs.python != 'true'
+        working-directory: .
+        run: echo "No workspace/** changes — skipping real lint+test; this job always runs to satisfy the required-check name on branch protection."
+      - if: needs.changes.outputs.python == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - if: needs.changes.outputs.python == 'true'
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: '3.11'
+          cache: pip
+          cache-dependency-path: workspace/requirements.txt
+      - if: needs.changes.outputs.python == 'true'
+        run: pip install -r requirements.txt pytest pytest-asyncio pytest-cov sqlalchemy>=2.0.0
+      # Coverage flags + fail-under floor moved into workspace/pytest.ini
+      # (issue #1817) so local `pytest` and CI use identical config.
+      - if: needs.changes.outputs.python == 'true'
+        run: python -m pytest --tb=short
+
+      - if: needs.changes.outputs.python == 'true'
+        name: Per-file critical-path coverage (MCP / inbox / auth)
+        # MCP-critical Python files have a per-file floor on top of the
+        # 86% total floor in pytest.ini. Rationale (issue #2790, after
+        # the PR #2766 → PR #2771 cycle): the total floor averages ~6000
+        # lines, so a single MCP file could regress to ~50% with no
+        # complaint as long as other modules compensate. These five
+        # files handle multi-tenant routing + auth + inbox dispatch —
+        # a coverage drop here is the same risk shape as a Go-side
+        # workspace-server token/secrets file dropping below 10%.
+        #
+        # Floor 75% sits below current actuals (80-96%) so this gate is
+        # strictly additive — no existing PR fails. Ratchet plan in
+        # COVERAGE_FLOOR.md.
+        run: |
+          set -e
+          PER_FILE_FLOOR=75
+          CRITICAL_FILES=(
+            "a2a_mcp_server.py"
+            "mcp_cli.py"
+            "a2a_tools.py"
+            "a2a_tools_inbox.py"
+            "inbox.py"
+            "platform_auth.py"
+          )
+
+          # pytest already wrote .coverage; emit a JSON view scoped to
+          # the critical files so jq/python can read the per-file pct
+          # without parsing tabular text. --include uses fnmatch, and
+          # the leading "*" allows the file to live anywhere under the
+          # workspace root (today they sit at workspace/<name>.py).
+          INCLUDES=$(printf '*%s,' "${CRITICAL_FILES[@]}")
+          INCLUDES="${INCLUDES%,}"
+          python -m coverage json -o /tmp/critical-cov.json --include="$INCLUDES"
+
+          FAILED=0
+          for f in "${CRITICAL_FILES[@]}"; do
+            # Match by top-level path key (e.g. "a2a_tools.py", not
+            # "builtin_tools/a2a_tools.py" — different file at 100%).
+            # The keys in coverage.json are paths relative to the run
+            # cwd (workspace/), so the critical-path entry sits at the
+            # bare basename.
+            pct=$(jq -r --arg f "$f" '.files | to_entries | map(select(.key == $f)) | .[0].value.summary.percent_covered // "MISSING"' /tmp/critical-cov.json)
+            if [ "$pct" = "MISSING" ]; then
+              echo "::error file=workspace/$f::No coverage data — file may have moved or test exclusion mis-set."
+              FAILED=$((FAILED+1))
+              continue
+            fi
+            echo "$f: ${pct}%"
+            if awk "BEGIN{exit !($pct < $PER_FILE_FLOOR)}"; then
+              echo "::error file=workspace/$f::${pct}% < ${PER_FILE_FLOOR}% per-file floor (MCP critical path). See COVERAGE_FLOOR.md."
+              FAILED=$((FAILED+1))
+            fi
+          done
+
+          if [ "$FAILED" -gt 0 ]; then
+            echo ""
+            echo "$FAILED MCP critical-path file(s) below the ${PER_FILE_FLOOR}% per-file floor."
+            echo "These paths handle multi-tenant routing, auth tokens, and inbox dispatch."
+            echo "A coverage drop here is the same risk shape as Go-side tokens/secrets files"
+            echo "dropping below 10% (see COVERAGE_FLOOR.md). Either:"
+            echo "  (a) add tests to raise coverage back above ${PER_FILE_FLOOR}%, or"
+            echo "  (b) if this is unavoidable historical debt, file an issue and propose"
+            echo "      adjusting the floor with rationale in COVERAGE_FLOOR.md."
+            exit 1
+          fi
+
+      # SDK + plugin validation moved to standalone repo:
+      # github.com/molecule-ai/molecule-sdk-python
+
@@ -0,0 +1,257 @@
+name: Continuous synthetic E2E (staging)
+
+# Hard gate (#2342): cron-driven full-lifecycle E2E that catches
+# regressions visible only at runtime — schema drift, deployment-pipeline
+# gaps, vendor outages, env-var rotations, DNS / CF / Railway side-effects.
+#
+# Why this gate exists:
+#   PR-time CI catches code-level regressions but not deployment-time or
+#   integration-time ones. Today's empirical data:
+#     • #2345 (A2A v0.2 silent drop) — passed all unit tests, broke at
+#       JSON-RPC parse layer between sender and receiver. Visible only
+#       to a sender exercising the full path.
+#     • RFC #2312 chat upload — landed on staging-branch but never
+#       reached staging tenants because publish-workspace-server-image
+#       was main-only. Caught by manual dogfooding hours after deploy.
+#   Both would have surfaced within 15-20 min of regression if a
+#   continuous synth-E2E was running.
+#
+# Cadence: every 20 min (3x/hour). The script is conservatively
+# bounded at 10 min wall-clock; even on degraded staging it should
+# finish before the next firing. cron-overlap is guarded by the
+# concurrency group below.
+#
+# Cost: ~3 runs/hour × 5-10 min × $0.008/min GHA = ~$0.50-$1/day.
+# Plus a fresh tenant provisioned + torn down each run (Railway +
+# AWS pennies). Negligible.
+#
+# Failure handling: when the run fails, the workflow exits non-zero
+# and GitHub's standard email/notification path fires. Operators
+# can subscribe to this workflow's failure channel for paging-grade
+# alerting.
+
+on:
+  schedule:
+    # Every 10 minutes, on :02 :12 :22 :32 :42 :52. Three constraints:
+    #   1. Stay off the top-of-hour. GitHub Actions scheduler drops
+    #      :00 firings under high load (own docs:
+    #      https://docs.github.com/en/actions/using-workflows/events-that-trigger-workflows#schedule).
+    #      Prior history: cron was '0,20,40' (2026-05-02) — only :00
+    #      ever survived. Bumped to '10,30,50' (2026-05-03) on the
+    #      theory that further-from-:00 wins. Empirically 2026-05-04
+    #      that ALSO dropped to ~60 min effective cadence (only ~1
+    #      schedule fire per hour — see molecule-core#2726). Detection
+    #      latency was claimed 20 min, actual 60 min.
+    #   2. Avoid colliding with the existing :15 sweep-cf-orphans
+    #      and :45 sweep-cf-tunnels — both hit the CF API and we
+    #      don't want to fight for rate-limit tokens.
+    #   3. Avoid the :30 heavy slot (canary-staging /30, sweep-aws-
+    #      secrets, sweep-stale-e2e-orgs every :15) — multiple
+    #      overlapping cron registrations on the same minute is part
+    #      of what GH drops under load.
+    # Solution: bump fires-per-hour 3 → 6 AND keep all slots in clean
+    # lanes (1-3 min away from any other cron). Even with empirically-
+    # observed ~67% GH drop ratio, 6 attempts/hour yields ~2 effective
+    # fires = ~30 min cadence; closer to the 20-min target than the
+    # current shape and provides a real degradation alarm if drops
+    # get worse.
+    - cron: '2,12,22,32,42,52 * * * *'
+  workflow_dispatch:
+    inputs:
+      runtime:
+        description: "Runtime to provision (claude-code = default + cheapest via MiniMax; langgraph = OpenAI-only; hermes = SDK-native path, slower)"
+        required: false
+        default: "claude-code"
+        type: string
+      model_slug:
+        description: "Model id to provision the workspace with (default MiniMax-M2.7-highspeed; e.g. 'sonnet' to test direct Anthropic, 'openai/gpt-4o' for hermes)"
+        required: false
+        default: "MiniMax-M2.7-highspeed"
+        type: string
+      keep_org:
+        description: "Skip teardown for post-mortem debugging (only manual dispatch — never set this for cron runs)"
+        required: false
+        default: false
+        type: boolean
+
+permissions:
+  contents: read
+  # No issue-write here — failures surface as red runs in the workflow
+  # history. If you want auto-issue-on-fail, add a follow-up step that
+  # uses gh issue create gated on `if: failure()`. Keeping the surface
+  # minimal until that's actually wanted.
+
+# Serialize so two firings can never overlap. Cron firing every 20 min
+# but scripts conservatively bounded at 10 min — overlap shouldn't
+# happen in steady state, but if a run hangs we don't want N more
+# stacking up.
+concurrency:
+  group: continuous-synth-e2e
+  cancel-in-progress: false
+
+jobs:
+  synth:
+    name: Synthetic E2E against staging
+    runs-on: ubuntu-latest
+    # Bumped from 12 → 20 (2026-05-04). Tenant user-data install phase
+    # (apt-get update + install docker.io/jq/awscli/caddy + snap install
+    # ssm-agent) runs from raw Ubuntu on every boot — none of it is
+    # pre-baked into the tenant AMI. Empirical fetch_secrets/ok timing
+    # across today's canaries: 51s → 82s → 143s → 625s. apt-mirror tail
+    # latency drives the boot-to-fetch_secrets phase from ~1min to >10min.
+    # A 12min budget leaves only ~2min for the workspace (which needs
+    # ~3.5min for claude-code cold boot) on slow-apt days, blowing the
+    # budget. 20min absorbs the worst tenant tail so the workspace probe
+    # gets the full ~7min it needs even on a slow apt day. Real fix:
+    # pre-bake caddy + ssm-agent into the tenant AMI (controlplane#TBD).
+    timeout-minutes: 20
+    env:
+      # claude-code default: cold-start ~5 min (comparable to langgraph),
+      # but uses MiniMax-M2.7-highspeed via the template's third-party-
+      # Anthropic-compat path (workspace-configs-templates/claude-code-
+      # default/config.yaml:64-69). MiniMax is ~5-10x cheaper than
+      # gpt-4.1-mini per token AND avoids the recurring OpenAI quota-
+      # exhaustion class that took the canary down 2026-05-03 (#265).
+      # Operators can pick langgraph / hermes via workflow_dispatch
+      # when they specifically need to exercise the OpenAI or SDK-
+      # native paths.
+      E2E_RUNTIME: ${{ github.event.inputs.runtime || 'claude-code' }}
+      # Pin the canary to a specific MiniMax model rather than relying
+      # on the per-runtime default ("sonnet" → routes to direct
+      # Anthropic, defeats the cost saving). Operators can override
+      # via workflow_dispatch by setting a different E2E_MODEL_SLUG
+      # input if they need to exercise a specific model. M2.7-highspeed
+      # is "Token Plan only" but cheap-per-token and fast.
+      E2E_MODEL_SLUG: ${{ github.event.inputs.model_slug || 'MiniMax-M2.7-highspeed' }}
+      # Bound to 10 min so a stuck provision fails the run instead of
+      # holding up the next cron firing. 15-min default in the script
+      # is for the on-PR full lifecycle where we have more headroom.
+      E2E_PROVISION_TIMEOUT_SECS: '600'
+      # Slug suffix — namespaced "synth-" so these runs are
+      # distinguishable from PR-driven runs in CP admin.
+      E2E_RUN_ID: synth-${{ github.run_id }}
+      # Forced false for cron; respected for manual dispatch
+      E2E_KEEP_ORG: ${{ github.event.inputs.keep_org == 'true' && '1' || '' }}
+      MOLECULE_CP_URL: ${{ vars.STAGING_CP_URL || 'https://staging-api.moleculesai.app' }}
+      MOLECULE_ADMIN_TOKEN: ${{ secrets.CP_STAGING_ADMIN_API_TOKEN }}
+      # MiniMax key is the canary's PRIMARY auth path. claude-code
+      # template's `minimax` provider routes ANTHROPIC_BASE_URL to
+      # api.minimax.io/anthropic and reads MINIMAX_API_KEY at boot.
+      # tests/e2e/test_staging_full_saas.sh branches SECRETS_JSON on
+      # which key is present — MiniMax wins when set.
+      E2E_MINIMAX_API_KEY: ${{ secrets.MOLECULE_STAGING_MINIMAX_API_KEY }}
+      # Direct-Anthropic alternative for operators who don't want to
+      # set up a MiniMax account (priority below MiniMax — first
+      # non-empty wins in test_staging_full_saas.sh's secrets-injection
+      # block). See #2578 PR comment for the rationale.
+      E2E_ANTHROPIC_API_KEY: ${{ secrets.MOLECULE_STAGING_ANTHROPIC_API_KEY }}
+      # OpenAI fallback — kept wired so operators can dispatch with
+      # E2E_RUNTIME=langgraph or =hermes and still have a working
+      # canary path. The script picks the right blob shape based on
+      # which key is non-empty.
+      E2E_OPENAI_API_KEY: ${{ secrets.MOLECULE_STAGING_OPENAI_KEY }}
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Verify required secrets present
+        run: |
+          # Hard-fail on missing secret REGARDLESS of trigger. Previously
+          # this step soft-skipped on workflow_dispatch via `exit 0`, but
+          # `exit 0` only ends the STEP — subsequent steps still ran with
+          # the empty secret, the synth script fell through to the wrong
+          # SECRETS_JSON branch, and the canary failed 5 min later with a
+          # confusing "Agent error (Exception)" instead of the clean
+          # "secret missing" message at the top. Caught 2026-05-04 by
+          # dispatched run 25296530706: claude-code + missing MINIMAX
+          # silently used OpenAI keys but kept model=MiniMax-M2.7, then
+          # the workspace 401'd against MiniMax once it tried to call.
+          # Fix: exit 1 in both cron and dispatch paths. Operators who
+          # want to verify a YAML change without setting up the secret
+          # can read the verify-secrets step's stderr — the failure is
+          # itself the verification signal.
+          if [ -z "${MOLECULE_ADMIN_TOKEN:-}" ]; then
+            echo "::error::CP_STAGING_ADMIN_API_TOKEN secret missing — synth E2E cannot run"
+            echo "::error::Set it at Settings → Secrets and Variables → Actions; pull from staging-CP's CP_ADMIN_API_TOKEN env in Railway."
+            exit 1
+          fi
+
+          # LLM-key requirement is per-runtime: claude-code accepts
+          # EITHER MiniMax OR direct-Anthropic (whichever is set first),
+          # langgraph + hermes use OpenAI (MOLECULE_STAGING_OPENAI_KEY).
+          case "${E2E_RUNTIME}" in
+            claude-code)
+              if [ -n "${E2E_MINIMAX_API_KEY:-}" ]; then
+                required_secret_name="MOLECULE_STAGING_MINIMAX_API_KEY"
+                required_secret_value="${E2E_MINIMAX_API_KEY}"
+              elif [ -n "${E2E_ANTHROPIC_API_KEY:-}" ]; then
+                required_secret_name="MOLECULE_STAGING_ANTHROPIC_API_KEY"
+                required_secret_value="${E2E_ANTHROPIC_API_KEY}"
+              else
+                required_secret_name="MOLECULE_STAGING_MINIMAX_API_KEY or MOLECULE_STAGING_ANTHROPIC_API_KEY"
+                required_secret_value=""
+              fi
+              ;;
+            langgraph|hermes)
+              required_secret_name="MOLECULE_STAGING_OPENAI_KEY"
+              required_secret_value="${E2E_OPENAI_API_KEY:-}"
+              ;;
+            *)
+              echo "::warning::Unknown E2E_RUNTIME='${E2E_RUNTIME}' — skipping LLM-key check"
+              required_secret_name=""
+              required_secret_value="present"
+              ;;
+          esac
+          if [ -n "$required_secret_name" ] && [ -z "$required_secret_value" ]; then
+            echo "::error::${required_secret_name} secret missing — runtime=${E2E_RUNTIME} cannot authenticate against its LLM provider"
+            echo "::error::Set it at Settings → Secrets and Variables → Actions, OR dispatch with a different runtime"
+            exit 1
+          fi
+
+      - name: Install required tools
+        run: |
+          # The script depends on jq + curl (already on ubuntu-latest)
+          # and python3 (likewise). Verify they're all present so we
+          # fail fast on a runner image regression rather than mid-script.
+          for cmd in jq curl python3; do
+            command -v "$cmd" >/dev/null 2>&1 || {
+              echo "::error::required tool '$cmd' not on PATH — runner image regression?"
+              exit 1
+            }
+          done
+
+      - name: Run synthetic E2E
+        # The script handles its own teardown via EXIT trap; even on
+        # failure (timeout, assertion), the org is deprovisioned and
+        # leaks are reported. Exit code propagates from the script.
+        run: |
+          bash tests/e2e/test_staging_full_saas.sh
+
+      - name: Failure summary
+        # Runs only on failure. Adds a job summary so the workflow run
+        # page shows a quick "what happened" instead of forcing readers
+        # to scroll through script output.
+        if: failure()
+        run: |
+          {
+            echo "## Continuous synth E2E failed"
+            echo ""
+            echo "**Run ID:** ${{ github.run_id }}"
+            echo "**Trigger:** ${{ github.event_name }}"
+            echo "**Runtime:** ${E2E_RUNTIME}"
+            echo "**Slug:** synth-${{ github.run_id }}"
+            echo ""
+            echo "### What this means"
+            echo ""
+            echo "Staging just regressed on a path that previously worked. Likely classes:"
+            echo "- Schema mismatch between sender and receiver (#2345 class)"
+            echo "- Deployment-pipeline gap (RFC #2312 / staging-tenant-image-stale class)"
+            echo "- Vendor outage (Cloudflare, Railway, AWS, GHCR)"
+            echo "- Staging-CP env var rotation"
+            echo ""
+            echo "### Next steps"
+            echo ""
+            echo "1. Check the script output above for the assertion that failed"
+            echo "2. If it's a vendor outage, no action needed — next firing in ~20 min"
+            echo "3. If it's a code regression, find the causing PR via \`git log\` against last green run and revert/fix"
+            echo "4. Keep an eye on the next 1-2 firings — flake vs persistent fail differs in priority"
+          } >> "$GITHUB_STEP_SUMMARY"
@@ -0,0 +1,307 @@
+name: E2E API Smoke Test
+# Extracted from ci.yml so workflow-level concurrency can protect this job
+# from run-level cancellation (issue #458).
+#
+# Trigger model (revised 2026-04-29):
+#
+# Always FIRES on push/pull_request to staging+main. Real work is gated
+# per-step on `needs.detect-changes.outputs.api` — when paths under
+# `workspace-server/`, `tests/e2e/`, or this workflow file haven't
+# changed, the no-op step alone runs and emits SUCCESS for the
+# `E2E API Smoke Test` check, satisfying branch protection without
+# spending CI cycles. See the in-job comment on the `e2e-api` job for
+# why this is one job (not two-jobs-sharing-name) and the 2026-04-29
+# PR #2264 incident that drove the consolidation.
+#
+# Parallel-safety (Class B Hongming-owned CICD red sweep, 2026-05-08)
+# -------------------------------------------------------------------
+# Same substrate hazard as PR #98 (handlers-postgres-integration). Our
+# Gitea act_runner runs with `container.network: host` (operator host
+# `/opt/molecule/runners/config.yaml`), which means:
+#
+#   * Two concurrent runs both try to bind their `-p 15432:5432` /
+#     `-p 16379:6379` host ports — the second postgres/redis FATALs
+#     with `Address in use` and `docker run` returns exit 125 with
+#     `Conflict. The container name "/molecule-ci-postgres" is already
+#     in use by container ...`. Verified in run a7/2727 on 2026-05-07.
+#   * The fixed container names `molecule-ci-postgres` / `-redis` (the
+#     pre-fix shape) collide on name AS WELL AS port. The cleanup-with-
+#     `docker rm -f` at the start of the second job KILLS the first
+#     job's still-running postgres/redis.
+#
+# Fix shape (mirrors PR #98's bridge-net pattern, adapted because
+# platform-server is a Go binary on the host, not a containerised
+# step):
+#
+#   1. Unique container names per run:
+#         pg-e2e-api-${RUN_ID}-${RUN_ATTEMPT}
+#         redis-e2e-api-${RUN_ID}-${RUN_ATTEMPT}
+#      `${RUN_ID}-${RUN_ATTEMPT}` is unique even across reruns of the
+#      same run_id.
+#   2. Ephemeral host port per run (`-p 0:5432`), then read the actual
+#      bound port via `docker port` and export DATABASE_URL/REDIS_URL
+#      pointing at it. No fixed host-port → no port collision.
+#   3. `127.0.0.1` (NOT `localhost`) in URLs — IPv6 first-resolve was
+#      the original flake fixed in #92 and the script's still IPv6-
+#      enabled.
+#   4. `if: always()` cleanup so containers don't leak when test steps
+#      fail.
+#
+# Issue #94 items #2 + #3 (also fixed here):
+#   * Pre-pull `alpine:latest` so the platform-server's provisioner
+#     (`internal/handlers/container_files.go`) can stand up its
+#     ephemeral token-write helper without a daemon.io round-trip.
+#   * Create `molecule-core-net` bridge network if missing so the
+#     provisioner's container.HostConfig {NetworkMode: ...} attach
+#     succeeds.
+# Item #1 (timeouts) — evidence on recent runs (77/3191, ae/4270, 0e/
+# 2318) shows Postgres ready in 3s, Redis in 1s, Platform in 1s when
+# they DO come up. Timeouts are not the bottleneck; not bumped.
+#
+# Item explicitly NOT fixed here: failing test `Status back online`
+# fails because the platform's langgraph workspace template image
+# (ghcr.io/molecule-ai/workspace-template-langgraph:latest) returns
+# 403 Forbidden post-2026-05-06 GitHub org suspension. That is a
+# template-registry resolution issue (ADR-002 / local-build mode) and
+# belongs in a separate change that touches workspace-server, not
+# this workflow file.
+
+on:
+  push:
+    branches: [main, staging]
+  pull_request:
+    branches: [main, staging]
+  workflow_dispatch:
+
+concurrency:
+  # Per-SHA grouping (changed 2026-04-28 from per-ref). Per-ref had the
+  # same auto-promote-staging brittleness as e2e-staging-canvas — back-
+  # to-back staging pushes share refs/heads/staging, so the older push's
+  # queued run gets cancelled when a newer push lands. Auto-promote-
+  # staging then sees `completed/cancelled` for the older SHA and stays
+  # put; the newer SHA's gates may eventually save the day, but if the
+  # newer push gets cancelled too, we deadlock.
+  #
+  # See e2e-staging-canvas.yml's identical concurrency block for the full
+  # rationale and the 2026-04-28 incident reference.
+  group: e2e-api-${{ github.event.pull_request.head.sha || github.sha }}
+  cancel-in-progress: false
+
+jobs:
+  detect-changes:
+    runs-on: ubuntu-latest
+    outputs:
+      api: ${{ steps.decide.outputs.api }}
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - uses: dorny/paths-filter@fbd0ab8f3e69293af611ebaee6363fc25e6d187d # v4.0.1
+        id: filter
+        with:
+          filters: |
+            api:
+              - 'workspace-server/**'
+              - 'tests/e2e/**'
+              - '.github/workflows/e2e-api.yml'
+      - id: decide
+        # Always run real work for manual dispatch — no diff context to
+        # filter against and ops dispatching this expects the suite to
+        # actually exercise the platform.
+        run: |
+          if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
+            echo "api=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "api=${{ steps.filter.outputs.api }}" >> "$GITHUB_OUTPUT"
+          fi
+
+  # ONE job (no job-level `if:`) that always runs and reports under the
+  # required-check name `E2E API Smoke Test`. Real work is gated per-step
+  # on `needs.detect-changes.outputs.api`. Reason: GitHub registers a
+  # check run for every job that matches `name:`, and a job-level
+  # `if: false` produces a SKIPPED check run. Branch protection treats
+  # all check runs with a matching context name on the latest commit as a
+  # SET — any SKIPPED in the set fails the required-check eval, even with
+  # SUCCESS siblings. Verified 2026-04-29 on PR #2264 (staging→main):
+  # 4 check runs (2 SKIPPED + 2 SUCCESS) at the head SHA blocked
+  # promotion despite all real work succeeding. Collapsing to a single
+  # always-running job with conditional steps emits exactly one SUCCESS
+  # check run regardless of paths filter — branch-protection-clean.
+  e2e-api:
+    needs: detect-changes
+    name: E2E API Smoke Test
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+    env:
+      # Unique per-run container names so concurrent runs on the host-
+      # network act_runner don't collide on name OR port.
+      # `${RUN_ID}-${RUN_ATTEMPT}` stays unique across reruns of the
+      # same run_id. PORT is set later (after docker port lookup) since
+      # we let Docker assign an ephemeral host port.
+      PG_CONTAINER: pg-e2e-api-${{ github.run_id }}-${{ github.run_attempt }}
+      REDIS_CONTAINER: redis-e2e-api-${{ github.run_id }}-${{ github.run_attempt }}
+      PORT: "8080"
+    steps:
+      - name: No-op pass (paths filter excluded this commit)
+        if: needs.detect-changes.outputs.api != 'true'
+        run: |
+          echo "No workspace-server / tests/e2e / workflow changes — E2E API gate satisfied without running tests."
+          echo "::notice::E2E API Smoke Test no-op pass (paths filter excluded this commit)."
+      - if: needs.detect-changes.outputs.api == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - if: needs.detect-changes.outputs.api == 'true'
+        uses: actions/setup-go@40f1582b2485089dde7abd97c1529aa768e1baff # v5
+        with:
+          go-version: 'stable'
+          cache: true
+          cache-dependency-path: workspace-server/go.sum
+      - name: Pre-pull alpine + ensure provisioner network (Issue #94 items #2 + #3)
+        if: needs.detect-changes.outputs.api == 'true'
+        run: |
+          # Provisioner uses alpine:latest for ephemeral token-write
+          # containers (workspace-server/internal/handlers/container_files.go).
+          # Pre-pull so the first provision in test_api.sh doesn't race
+          # the daemon's pull cache. Idempotent — `docker pull` is a no-op
+          # when the image is already present.
+          docker pull alpine:latest >/dev/null
+          # Provisioner attaches workspace containers to
+          # molecule-core-net (workspace-server/internal/provisioner/
+          # provisioner.go::DefaultNetwork). The bridge already exists on
+          # the operator host's docker daemon — `network create` is
+          # idempotent via `|| true`.
+          docker network create molecule-core-net >/dev/null 2>&1 || true
+          echo "alpine:latest pre-pulled; molecule-core-net ensured."
+      - name: Start Postgres (docker)
+        if: needs.detect-changes.outputs.api == 'true'
+        run: |
+          # Defensive cleanup — only matches THIS run's container name,
+          # so it cannot kill a sibling run's postgres. (Pre-fix the
+          # name was static and this rm hit other runs' containers.)
+          docker rm -f "$PG_CONTAINER" 2>/dev/null || true
+          # `-p 0:5432` requests an ephemeral host port; we read it back
+          # below and export DATABASE_URL.
+          docker run -d --name "$PG_CONTAINER" \
+            -e POSTGRES_USER=dev -e POSTGRES_PASSWORD=dev -e POSTGRES_DB=molecule \
+            -p 0:5432 postgres:16 >/dev/null
+          # Resolve the host-side port assignment. `docker port` prints
+          # `0.0.0.0:NNNN` (and on host-net runners may also print an
+          # IPv6 line — take the first IPv4 line).
+          PG_PORT=$(docker port "$PG_CONTAINER" 5432/tcp | awk -F: '/^0\.0\.0\.0:/ {print $2; exit}')
+          if [ -z "$PG_PORT" ]; then
+            # Fallback: any first line. Some Docker versions print only
+            # one line.
+            PG_PORT=$(docker port "$PG_CONTAINER" 5432/tcp | head -1 | awk -F: '{print $NF}')
+          fi
+          if [ -z "$PG_PORT" ]; then
+            echo "::error::Could not resolve host port for $PG_CONTAINER"
+            docker port "$PG_CONTAINER" 5432/tcp || true
+            docker logs "$PG_CONTAINER" || true
+            exit 1
+          fi
+          # 127.0.0.1 (NOT localhost) — IPv6 first-resolve flake (#92).
+          echo "PG_PORT=${PG_PORT}" >> "$GITHUB_ENV"
+          echo "DATABASE_URL=postgres://dev:dev@127.0.0.1:${PG_PORT}/molecule?sslmode=disable" >> "$GITHUB_ENV"
+          echo "Postgres host port: ${PG_PORT}"
+          for i in $(seq 1 30); do
+            if docker exec "$PG_CONTAINER" pg_isready -U dev >/dev/null 2>&1; then
+              echo "Postgres ready after ${i}s"
+              exit 0
+            fi
+            sleep 1
+          done
+          echo "::error::Postgres did not become ready in 30s"
+          docker logs "$PG_CONTAINER" || true
+          exit 1
+      - name: Start Redis (docker)
+        if: needs.detect-changes.outputs.api == 'true'
+        run: |
+          docker rm -f "$REDIS_CONTAINER" 2>/dev/null || true
+          docker run -d --name "$REDIS_CONTAINER" -p 0:6379 redis:7 >/dev/null
+          REDIS_PORT=$(docker port "$REDIS_CONTAINER" 6379/tcp | awk -F: '/^0\.0\.0\.0:/ {print $2; exit}')
+          if [ -z "$REDIS_PORT" ]; then
+            REDIS_PORT=$(docker port "$REDIS_CONTAINER" 6379/tcp | head -1 | awk -F: '{print $NF}')
+          fi
+          if [ -z "$REDIS_PORT" ]; then
+            echo "::error::Could not resolve host port for $REDIS_CONTAINER"
+            docker port "$REDIS_CONTAINER" 6379/tcp || true
+            docker logs "$REDIS_CONTAINER" || true
+            exit 1
+          fi
+          echo "REDIS_PORT=${REDIS_PORT}" >> "$GITHUB_ENV"
+          echo "REDIS_URL=redis://127.0.0.1:${REDIS_PORT}" >> "$GITHUB_ENV"
+          echo "Redis host port: ${REDIS_PORT}"
+          for i in $(seq 1 15); do
+            if docker exec "$REDIS_CONTAINER" redis-cli ping 2>/dev/null | grep -q PONG; then
+              echo "Redis ready after ${i}s"
+              exit 0
+            fi
+            sleep 1
+          done
+          echo "::error::Redis did not become ready in 15s"
+          docker logs "$REDIS_CONTAINER" || true
+          exit 1
+      - name: Build platform
+        if: needs.detect-changes.outputs.api == 'true'
+        working-directory: workspace-server
+        run: go build -o platform-server ./cmd/server
+      - name: Start platform (background)
+        if: needs.detect-changes.outputs.api == 'true'
+        working-directory: workspace-server
+        run: |
+          # DATABASE_URL + REDIS_URL exported by the start-postgres /
+          # start-redis steps point at this run's per-run host ports.
+          ./platform-server > platform.log 2>&1 &
+          echo $! > platform.pid
+      - name: Wait for /health
+        if: needs.detect-changes.outputs.api == 'true'
+        run: |
+          for i in $(seq 1 30); do
+            if curl -sf http://127.0.0.1:8080/health > /dev/null; then
+              echo "Platform up after ${i}s"
+              exit 0
+            fi
+            sleep 1
+          done
+          echo "::error::Platform did not become healthy in 30s"
+          cat workspace-server/platform.log || true
+          exit 1
+      - name: Assert migrations applied
+        if: needs.detect-changes.outputs.api == 'true'
+        run: |
+          tables=$(docker exec "$PG_CONTAINER" psql -U dev -d molecule -tAc "SELECT count(*) FROM information_schema.tables WHERE table_schema='public' AND table_name='workspaces'")
+          if [ "$tables" != "1" ]; then
+            echo "::error::Migrations did not apply"
+            cat workspace-server/platform.log || true
+            exit 1
+          fi
+          echo "Migrations OK"
+      - name: Run E2E API tests
+        if: needs.detect-changes.outputs.api == 'true'
+        run: bash tests/e2e/test_api.sh
+      - name: Run notify-with-attachments E2E
+        if: needs.detect-changes.outputs.api == 'true'
+        run: bash tests/e2e/test_notify_attachments_e2e.sh
+      - name: Run priority-runtimes E2E (claude-code + hermes — skips when keys absent)
+        if: needs.detect-changes.outputs.api == 'true'
+        run: bash tests/e2e/test_priority_runtimes_e2e.sh
+      - name: Run poll-mode + since_id cursor E2E (#2339)
+        if: needs.detect-changes.outputs.api == 'true'
+        run: bash tests/e2e/test_poll_mode_e2e.sh
+      - name: Run poll-mode chat upload E2E (RFC #2891)
+        if: needs.detect-changes.outputs.api == 'true'
+        run: bash tests/e2e/test_poll_mode_chat_upload_e2e.sh
+      - name: Dump platform log on failure
+        if: failure() && needs.detect-changes.outputs.api == 'true'
+        run: cat workspace-server/platform.log || true
+      - name: Stop platform
+        if: always() && needs.detect-changes.outputs.api == 'true'
+        run: |
+          if [ -f workspace-server/platform.pid ]; then
+            kill "$(cat workspace-server/platform.pid)" 2>/dev/null || true
+          fi
+      - name: Stop service containers
+        # always() so containers don't leak when test steps fail. The
+        # cleanup is best-effort: if the container is already gone
+        # (e.g. concurrent rerun race), don't fail the job.
+        if: always() && needs.detect-changes.outputs.api == 'true'
+        run: |
+          docker rm -f "$PG_CONTAINER" 2>/dev/null || true
+          docker rm -f "$REDIS_CONTAINER" 2>/dev/null || true
@@ -0,0 +1,216 @@
+name: E2E Staging Canvas (Playwright)
+
+# Playwright test suite that provisions a fresh staging org per run and
+# verifies every workspace-panel tab renders without crashing. Complements
+# e2e-staging-saas.yml (which tests the API shape) by exercising the
+# actual browser + canvas bundle against live staging.
+#
+# Triggers: push to main/staging or PR touching canvas sources + this workflow,
+# manual dispatch, and weekly cron to catch browser/runtime drift even
+# when canvas is quiet.
+# Added staging to push/pull_request branches so the auto-promote gate
+# check (--event push --branch staging) can see a completed run for this
+# workflow — mirrors what PR #1891 does for e2e-api.yml.
+
+on:
+  # Trigger model (revised 2026-04-29):
+  #
+  # Always fires on push/pull_request; real work is gated per-step on
+  # `needs.detect-changes.outputs.canvas`. When canvas/ paths haven't
+  # changed, the no-op step alone runs and emits SUCCESS for the
+  # `Canvas tabs E2E` check, satisfying branch protection without
+  # spending CI cycles. See e2e-api.yml for the rationale on why this
+  # is a single job rather than two-jobs-sharing-name.
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+  workflow_dispatch:
+  schedule:
+    # Weekly on Sunday 08:00 UTC — catches Chrome / Playwright / Next.js
+    # release-note-shaped regressions that don't ride in with a PR.
+    - cron: '0 8 * * 0'
+
+concurrency:
+  # Per-SHA grouping (changed 2026-04-28 from a single global group). The
+  # global group made auto-promote-staging brittle: when a staging push
+  # queued behind an in-flight run and a third entrant (a PR run, a
+  # follow-on push) entered the group, the staging push got cancelled —
+  # leaving auto-promote-staging looking at `completed/cancelled` for a
+  # required gate and refusing to advance main. Observed 2026-04-28
+  # 23:51-23:53 on staging tip 3f99fede.
+  #
+  # The original intent of the global group was to throttle parallel
+  # E2E provisions (each spins a fresh EC2). At our scale that throttle
+  # isn't worth the correctness cost — fresh-org-per-run isolates the
+  # state, and the cost of two parallel runs (~$0.001/min × 10min × 2)
+  # is rounding error vs. the cost of a stuck pipeline.
+  #
+  # Per-SHA still dedupes accidental double-triggers for the SAME SHA.
+  # It does NOT cancel obsolete-PR-version runs on force-push; that
+  # wasted CI is acceptable given the alternative is losing staging-tip
+  # data that auto-promote-staging needs.
+  group: e2e-staging-canvas-${{ github.event.pull_request.head.sha || github.sha }}
+  cancel-in-progress: false
+
+jobs:
+  detect-changes:
+    runs-on: ubuntu-latest
+    outputs:
+      canvas: ${{ steps.decide.outputs.canvas }}
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - uses: dorny/paths-filter@fbd0ab8f3e69293af611ebaee6363fc25e6d187d # v4.0.1
+        id: filter
+        with:
+          filters: |
+            canvas:
+              - 'canvas/**'
+              - '.github/workflows/e2e-staging-canvas.yml'
+      - id: decide
+        # Always run real tests for manual dispatch and the weekly cron —
+        # both exist precisely to exercise the suite, regardless of diff.
+        run: |
+          if [ "${{ github.event_name }}" = "workflow_dispatch" ] || [ "${{ github.event_name }}" = "schedule" ]; then
+            echo "canvas=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "canvas=${{ steps.filter.outputs.canvas }}" >> "$GITHUB_OUTPUT"
+          fi
+
+  # ONE job (no job-level `if:`) that always runs and reports under the
+  # required-check name `Canvas tabs E2E`. Real work is gated per-step on
+  # `needs.detect-changes.outputs.canvas`. See e2e-api.yml for the full
+  # rationale — same path-filter check-name parity issue blocked PR #2264
+  # (staging→main) on 2026-04-29 because branch protection treats matching-
+  # name check runs as a SET, and any SKIPPED member fails the eval.
+  playwright:
+    needs: detect-changes
+    name: Canvas tabs E2E
+    runs-on: ubuntu-latest
+    timeout-minutes: 40
+
+    env:
+      CANVAS_E2E_STAGING: '1'
+      MOLECULE_CP_URL: https://staging-api.moleculesai.app
+      MOLECULE_ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+
+    defaults:
+      run:
+        working-directory: canvas
+
+    steps:
+      - name: No-op pass (paths filter excluded this commit)
+        if: needs.detect-changes.outputs.canvas != 'true'
+        working-directory: .
+        run: |
+          echo "No canvas / workflow changes — E2E Staging Canvas gate satisfied without running tests."
+          echo "::notice::E2E Staging Canvas no-op pass (paths filter excluded this commit)."
+
+      - if: needs.detect-changes.outputs.canvas == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Verify admin token present
+        if: needs.detect-changes.outputs.canvas == 'true'
+        run: |
+          if [ -z "$MOLECULE_ADMIN_TOKEN" ]; then
+            echo "::error::Missing MOLECULE_STAGING_ADMIN_TOKEN"
+            exit 2
+          fi
+
+      - name: Set up Node
+        if: needs.detect-changes.outputs.canvas == 'true'
+        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
+        with:
+          node-version: '20'
+          cache: 'npm'
+          cache-dependency-path: canvas/package-lock.json
+
+      - name: Install canvas deps
+        if: needs.detect-changes.outputs.canvas == 'true'
+        run: npm ci
+
+      - name: Install Playwright browsers
+        if: needs.detect-changes.outputs.canvas == 'true'
+        timeout-minutes: 10
+        run: npx playwright install --with-deps chromium
+
+      - name: Run staging canvas E2E
+        if: needs.detect-changes.outputs.canvas == 'true'
+        run: npx playwright test --config=playwright.staging.config.ts
+
+      - name: Upload Playwright report on failure
+        if: failure() && needs.detect-changes.outputs.canvas == 'true'
+        # Pinned to v3 for Gitea act_runner v0.6 compatibility — v4+ uses
+        # the GHES 3.10+ artifact protocol that Gitea 1.22.x does NOT
+        # implement (see ci.yml upload step for the canonical error
+        # cite). Drop this pin when Gitea ships the v4 protocol.
+        uses: actions/upload-artifact@c6a366c94c3e0affe28c06c8df20a878f24da3cf # v3.2.2
+        with:
+          name: playwright-report-staging
+          path: canvas/playwright-report-staging/
+          retention-days: 14
+
+      - name: Upload screenshots on failure
+        if: failure() && needs.detect-changes.outputs.canvas == 'true'
+        # Pinned to v3 for Gitea act_runner v0.6 compatibility (see above).
+        uses: actions/upload-artifact@c6a366c94c3e0affe28c06c8df20a878f24da3cf # v3.2.2
+        with:
+          name: playwright-screenshots
+          path: canvas/test-results/
+          retention-days: 14
+
+      # Safety-net teardown — fires only when Playwright's globalTeardown
+      # didn't (worker crash, runner cancel). Reads the slug from
+      # canvas/.playwright-staging-state.json (written by staging-setup
+      # as its first action, before any CP call) and deletes only that
+      # slug.
+      #
+      # Earlier versions of this step pattern-swept `e2e-canvas-<today>-*`
+      # orgs to compensate for setup-crash-before-state-file-write. That
+      # over-aggressive cleanup raced concurrent canvas-E2E runs and
+      # poisoned each other's tenants — observed 2026-04-30 when three
+      # real-test runs killed each other mid-test, surfacing as
+      # `getaddrinfo ENOTFOUND` once CP had cleaned up the just-deleted
+      # DNS record. Pattern-sweep removed; setup now writes the state
+      # file before any CP work, so the slug is always recoverable.
+      - name: Teardown safety net
+        if: always() && needs.detect-changes.outputs.canvas == 'true'
+        env:
+          ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+        run: |
+          set +e
+          STATE_FILE=".playwright-staging-state.json"
+          if [ ! -f "$STATE_FILE" ]; then
+            echo "::notice::No state file at canvas/$STATE_FILE — Playwright globalTeardown handled it (or setup never ran)."
+            exit 0
+          fi
+          slug=$(python3 -c "import json; print(json.load(open('$STATE_FILE')).get('slug',''))")
+          if [ -z "$slug" ]; then
+            echo "::warning::State file present but slug missing; nothing to clean up."
+            exit 0
+          fi
+          echo "Deleting orphan tenant: $slug"
+          # Verify HTTP 2xx instead of `>/dev/null || true` swallowing
+          # failures. A 5xx or timeout previously looked identical to
+          # success, leaving the tenant alive for up to ~45 min until
+          # sweep-stale-e2e-orgs caught it. Surface failures as
+          # workflow warnings naming the slug. Don't `exit 1` — a single
+          # cleanup miss shouldn't fail-flag the canvas test when the
+          # actual smoke check passed; the sweeper is the safety net.
+          # See molecule-controlplane#420.
+          # Tempfile-routed -w + set +e/-e prevents curl-exit-code
+          # pollution of the captured status (lint-curl-status-capture.yml).
+          set +e
+          curl -sS -o /tmp/canvas-cleanup.out -w "%{http_code}" \
+            -X DELETE "$MOLECULE_CP_URL/cp/admin/tenants/$slug" \
+            -H "Authorization: Bearer $ADMIN_TOKEN" \
+            -H "Content-Type: application/json" \
+            -d "{\"confirm\":\"$slug\"}" >/tmp/canvas-cleanup.code
+          set -e
+          code=$(cat /tmp/canvas-cleanup.code 2>/dev/null || echo "000")
+          if [ "$code" = "200" ] || [ "$code" = "204" ]; then
+            echo "[teardown] deleted $slug (HTTP $code)"
+          else
+            echo "::warning::canvas teardown for $slug returned HTTP $code — sweep-stale-e2e-orgs will catch it within ~45 min. Body: $(head -c 300 /tmp/canvas-cleanup.out 2>/dev/null)"
+          fi
+          exit 0
@@ -0,0 +1,184 @@
+name: E2E Staging External Runtime
+
+# Regression for the four/five workspaces.status=awaiting_agent transitions
+# that silently failed in production for five days before migration 046
+# extended the workspace_status enum (see
+# workspace-server/migrations/046_workspace_status_awaiting_agent.up.sql).
+#
+# Why this is its own workflow (not folded into e2e-staging-saas.yml):
+#   - The full-saas harness defaults to runtime=hermes, never exercises
+#     external-runtime. Adding an `external` parameter to that script
+#     would force every push to staging through both lifecycles in
+#     series, doubling the EC2 cold-start budget.
+#   - The external lifecycle has unique timing (REMOTE_LIVENESS_STALE_AFTER
+#     window, 90s default + sweep interval), which we wait through
+#     deliberately. Folding it into hermes would make the long path
+#     even longer.
+#   - It can run in parallel with the hermes E2E since both create
+#     fresh tenant orgs with distinct slug prefixes (`e2e-ext-...` vs
+#     `e2e-...`).
+#
+# Triggers:
+#   - Push to staging when any source affecting external runtime,
+#     hibernation, or the migration set changes.
+#   - PR review for the same set.
+#   - Manual workflow_dispatch.
+#   - Daily cron at 07:30 UTC (catches drift on quiet days; staggered
+#     30 min after e2e-staging-saas.yml's 07:00 UTC cron).
+#
+# Concurrency: serialized so two staging pushes don't fight for the
+# same EC2 quota window. cancel-in-progress=false so a half-rolled
+# tenant always finishes its teardown.
+
+on:
+  push:
+    branches: [main]
+    paths:
+      - 'workspace-server/internal/handlers/workspace.go'
+      - 'workspace-server/internal/handlers/registry.go'
+      - 'workspace-server/internal/handlers/workspace_restart.go'
+      - 'workspace-server/internal/registry/healthsweep.go'
+      - 'workspace-server/internal/registry/liveness.go'
+      - 'workspace-server/migrations/**'
+      - 'workspace-server/internal/db/workspace_status_enum_drift_test.go'
+      - 'tests/e2e/test_staging_external_runtime.sh'
+      - '.github/workflows/e2e-staging-external.yml'
+  pull_request:
+    branches: [main]
+    paths:
+      - 'workspace-server/internal/handlers/workspace.go'
+      - 'workspace-server/internal/handlers/registry.go'
+      - 'workspace-server/internal/handlers/workspace_restart.go'
+      - 'workspace-server/internal/registry/healthsweep.go'
+      - 'workspace-server/internal/registry/liveness.go'
+      - 'workspace-server/migrations/**'
+      - 'workspace-server/internal/db/workspace_status_enum_drift_test.go'
+      - 'tests/e2e/test_staging_external_runtime.sh'
+      - '.github/workflows/e2e-staging-external.yml'
+  workflow_dispatch:
+    inputs:
+      keep_org:
+        description: "Skip teardown for debugging (only via manual dispatch)"
+        required: false
+        type: boolean
+        default: false
+      stale_wait_secs:
+        description: "Seconds to wait for the heartbeat-staleness sweep (default 180 = 90s window + 90s buffer)"
+        required: false
+        default: "180"
+  schedule:
+    - cron: '30 7 * * *'
+
+concurrency:
+  group: e2e-staging-external
+  cancel-in-progress: false
+
+permissions:
+  contents: read
+
+jobs:
+  e2e-staging-external:
+    name: E2E Staging External Runtime
+    runs-on: ubuntu-latest
+    timeout-minutes: 25
+
+    env:
+      MOLECULE_CP_URL: https://staging-api.moleculesai.app
+      MOLECULE_ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+      E2E_RUN_ID: "${{ github.run_id }}-${{ github.run_attempt }}"
+      E2E_KEEP_ORG: ${{ github.event.inputs.keep_org && '1' || '0' }}
+      E2E_STALE_WAIT_SECS: ${{ github.event.inputs.stale_wait_secs || '180' }}
+
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Verify admin token present
+        run: |
+          if [ -z "$MOLECULE_ADMIN_TOKEN" ]; then
+            # Schedule + push triggers must hard-fail when the token is
+            # missing — silent skip would mask infra rot. Manual dispatch
+            # gets the same hard-fail; an operator running this on a fork
+            # without secrets configured needs to know up-front.
+            echo "::error::MOLECULE_STAGING_ADMIN_TOKEN secret not set (Railway staging CP_ADMIN_API_TOKEN)"
+            exit 2
+          fi
+          echo "Admin token present ✓"
+
+      - name: CP staging health preflight
+        run: |
+          code=$(curl -sS -o /dev/null -w "%{http_code}" --max-time 10 "$MOLECULE_CP_URL/health")
+          if [ "$code" != "200" ]; then
+            echo "::error::Staging CP unhealthy (got HTTP $code). Skipping — not a workspace bug."
+            exit 1
+          fi
+          echo "Staging CP healthy ✓"
+
+      - name: Run external-runtime E2E
+        id: e2e
+        run: bash tests/e2e/test_staging_external_runtime.sh
+
+      # Mirror the e2e-staging-saas.yml safety net: if the runner is
+      # cancelled (e.g. concurrent staging push), the test script's
+      # EXIT trap may not fire, so we sweep e2e-ext-* slugs scoped to
+      # *this* run id.
+      - name: Teardown safety net (runs on cancel/failure)
+        if: always()
+        env:
+          ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+        run: |
+          set +e
+          orgs=$(curl -sS "$MOLECULE_CP_URL/cp/admin/orgs" \
+            -H "Authorization: Bearer $ADMIN_TOKEN" 2>/dev/null \
+            | python3 -c "
+          import json, sys, os, datetime
+          run_id = os.environ.get('GITHUB_RUN_ID', '')
+          d = json.load(sys.stdin)
+          # Scope STRICTLY to this run id (e2e-ext-YYYYMMDD-<runid>-...)
+          # so concurrent runs and unrelated dev probes are not touched.
+          # Sweep today AND yesterday so a midnight-crossing run still
+          # cleans up its own slug.
+          today = datetime.date.today()
+          yesterday = today - datetime.timedelta(days=1)
+          dates = (today.strftime('%Y%m%d'), yesterday.strftime('%Y%m%d'))
+          if not run_id:
+              # Without a run id we cannot scope safely; bail rather
+              # than risk deleting unrelated tenants.
+              sys.exit(0)
+          prefixes = tuple(f'e2e-ext-{d}-{run_id}-' for d in dates)
+          for o in d.get('orgs', []):
+              s = o.get('slug', '')
+              if s.startswith(prefixes) and o.get('status') != 'purged':
+                  print(s)
+          " 2>/dev/null)
+          if [ -n "$orgs" ]; then
+            echo "Safety-net sweep: deleting leftover orgs:"
+            echo "$orgs"
+            # Per-slug verified DELETE — see molecule-controlplane#420.
+            # `>/dev/null 2>&1` previously hid every failure; surface
+            # non-2xx as workflow warnings so the run page names what
+            # leaked. Sweeper catches the rest within ~45 min.
+            leaks=()
+            for slug in $orgs; do
+              # Tempfile-routed -w + set +e/-e prevents curl-exit-code
+              # pollution of the captured status (lint-curl-status-capture.yml).
+              set +e
+              curl -sS -o /tmp/external-cleanup.out -w "%{http_code}" \
+                -X DELETE "$MOLECULE_CP_URL/cp/admin/tenants/$slug" \
+                -H "Authorization: Bearer $ADMIN_TOKEN" \
+                -H "Content-Type: application/json" \
+                -d "{\"confirm\":\"$slug\"}" >/tmp/external-cleanup.code
+              set -e
+              code=$(cat /tmp/external-cleanup.code 2>/dev/null || echo "000")
+              if [ "$code" = "200" ] || [ "$code" = "204" ]; then
+                echo "[teardown] deleted $slug (HTTP $code)"
+              else
+                echo "::warning::external teardown for $slug returned HTTP $code — sweep-stale-e2e-orgs will catch it within ~45 min. Body: $(head -c 300 /tmp/external-cleanup.out 2>/dev/null)"
+                leaks+=("$slug")
+              fi
+            done
+            if [ ${#leaks[@]} -gt 0 ]; then
+              echo "::warning::external teardown left ${#leaks[@]} leak(s): ${leaks[*]}"
+            fi
+          else
+            echo "Safety-net sweep: no leftover orgs to clean."
+          fi
@@ -0,0 +1,246 @@
+name: E2E Staging SaaS (full lifecycle)
+
+# Dedicated workflow that provisions a fresh staging org per run, exercises
+# the full workspace lifecycle (register → heartbeat → A2A → delegation →
+# HMA memory → activity → peers), then tears down and asserts leak-free.
+#
+# Why a separate workflow (not folded into ci.yml):
+#   - The run takes ~25-35 min (EC2 boot + cloudflared DNS + provision sweeps +
+#     agent bootstrap), way too slow for every PR.
+#   - Needs its own concurrency group so two pushes don't fight over the
+#     same staging org slug prefix.
+#   - Has its own required secrets (session cookie, admin token) that most
+#     PRs don't need to read.
+#
+# Triggers:
+#   - Push to main (regression guard)
+#   - workflow_dispatch (manual re-run from UI)
+#   - Nightly cron (catches drift even when no pushes land)
+#   - Changes to any provisioning-critical file under PR review (opt-in
+#     via the same paths watcher that e2e-api.yml uses)
+
+on:
+  # Trunk-based (Phase 3 of internal#81): main is the only branch.
+  # Previously this fired on staging push too because staging was a
+  # superset of main and ran the gate ahead of auto-promote; with no
+  # staging branch, main is where E2E gates the deploy.
+  push:
+    branches: [main]
+    paths:
+      - 'workspace-server/internal/handlers/registry.go'
+      - 'workspace-server/internal/handlers/workspace_provision.go'
+      - 'workspace-server/internal/handlers/a2a_proxy.go'
+      - 'workspace-server/internal/middleware/**'
+      - 'workspace-server/internal/provisioner/**'
+      - 'tests/e2e/test_staging_full_saas.sh'
+      - '.github/workflows/e2e-staging-saas.yml'
+  pull_request:
+    branches: [main]
+    paths:
+      - 'workspace-server/internal/handlers/registry.go'
+      - 'workspace-server/internal/handlers/workspace_provision.go'
+      - 'workspace-server/internal/handlers/a2a_proxy.go'
+      - 'workspace-server/internal/middleware/**'
+      - 'workspace-server/internal/provisioner/**'
+      - 'tests/e2e/test_staging_full_saas.sh'
+      - '.github/workflows/e2e-staging-saas.yml'
+  workflow_dispatch:
+    inputs:
+      runtime:
+        description: "Runtime to test (claude-code [default, MiniMax] | hermes [OpenAI] | langgraph [OpenAI])"
+        required: false
+        default: "claude-code"
+      keep_org:
+        description: "Skip teardown for debugging (only use via manual dispatch!)"
+        required: false
+        type: boolean
+        default: false
+  schedule:
+    # 07:00 UTC every day — catches AMI drift, WorkOS cert rotation,
+    # Cloudflare API regressions, etc. even on quiet days.
+    - cron: '0 7 * * *'
+
+# Serialize: staging has a finite per-hour org creation quota. Two pushes
+# landing in quick succession should queue, not race. `cancel-in-progress:
+# false` mirrors e2e-api.yml — GitHub would otherwise cancel the running
+# teardown step and leave orphan EC2s.
+concurrency:
+  group: e2e-staging-saas
+  cancel-in-progress: false
+
+jobs:
+  e2e-staging-saas:
+    name: E2E Staging SaaS
+    runs-on: ubuntu-latest
+    timeout-minutes: 45
+    permissions:
+      contents: read
+
+    env:
+      MOLECULE_CP_URL: https://staging-api.moleculesai.app
+      # Single admin-bearer secret drives provision + tenant-token
+      # retrieval + teardown. Configure in
+      # Settings → Secrets and variables → Actions → Repository secrets.
+      MOLECULE_ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+      # MiniMax is the PRIMARY LLM auth path post-2026-05-04. Switched
+      # from hermes+OpenAI default after #2578 (the staging OpenAI key
+      # account went over quota and stayed dead for 36+ hours, taking
+      # the full-lifecycle E2E red on every provisioning-critical push).
+      # claude-code template's `minimax` provider routes
+      # ANTHROPIC_BASE_URL to api.minimax.io/anthropic and reads
+      # MINIMAX_API_KEY at boot — separate billing account so an
+      # OpenAI quota collapse no longer wedges the gate. Mirrors the
+      # canary-staging.yml + continuous-synth-e2e.yml migrations.
+      E2E_MINIMAX_API_KEY: ${{ secrets.MOLECULE_STAGING_MINIMAX_API_KEY }}
+      # Direct-Anthropic alternative for operators who don't want to
+      # set up a MiniMax account (priority below MiniMax — first
+      # non-empty wins in test_staging_full_saas.sh's secrets-injection
+      # block). See #2578 PR comment for the rationale.
+      E2E_ANTHROPIC_API_KEY: ${{ secrets.MOLECULE_STAGING_ANTHROPIC_API_KEY }}
+      # OpenAI fallback — kept wired so an operator-dispatched run with
+      # E2E_RUNTIME=hermes or =langgraph via workflow_dispatch can still
+      # exercise the OpenAI path.
+      E2E_OPENAI_API_KEY: ${{ secrets.MOLECULE_STAGING_OPENAI_KEY }}
+      E2E_RUNTIME: ${{ github.event.inputs.runtime || 'claude-code' }}
+      # Pin the model when running on the default claude-code path —
+      # the per-runtime default ("sonnet") routes to direct Anthropic
+      # and defeats the cost saving. Operators can override via the
+      # workflow_dispatch flow (no input wired here yet — runtime
+      # override is enough for ad-hoc).
+      E2E_MODEL_SLUG: ${{ github.event.inputs.runtime == 'hermes' && 'openai/gpt-4o' || github.event.inputs.runtime == 'langgraph' && 'openai:gpt-4o' || 'MiniMax-M2.7-highspeed' }}
+      E2E_RUN_ID: "${{ github.run_id }}-${{ github.run_attempt }}"
+      E2E_KEEP_ORG: ${{ github.event.inputs.keep_org && '1' || '0' }}
+
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Verify admin token present
+        run: |
+          if [ -z "$MOLECULE_ADMIN_TOKEN" ]; then
+            echo "::error::MOLECULE_STAGING_ADMIN_TOKEN secret not set (Railway staging CP_ADMIN_API_TOKEN)"
+            exit 2
+          fi
+          echo "Admin token present ✓"
+
+      - name: Verify LLM key present
+        run: |
+          # Per-runtime key check — claude-code uses MiniMax; hermes /
+          # langgraph (operator-dispatched only) use OpenAI. Hard-fail
+          # rather than soft-skip per #2578's lesson — empty key
+          # silently falls through to the wrong SECRETS_JSON branch and
+          # produces a confusing auth error 5 min later instead of the
+          # clean "secret missing" message at the top.
+          case "${E2E_RUNTIME}" in
+            claude-code)
+              # Either MiniMax OR direct-Anthropic works — first
+              # non-empty wins in the test script's secrets-injection
+              # priority chain.
+              if [ -n "${E2E_MINIMAX_API_KEY:-}" ]; then
+                required_secret_name="MOLECULE_STAGING_MINIMAX_API_KEY"
+                required_secret_value="${E2E_MINIMAX_API_KEY}"
+              elif [ -n "${E2E_ANTHROPIC_API_KEY:-}" ]; then
+                required_secret_name="MOLECULE_STAGING_ANTHROPIC_API_KEY"
+                required_secret_value="${E2E_ANTHROPIC_API_KEY}"
+              else
+                required_secret_name="MOLECULE_STAGING_MINIMAX_API_KEY or MOLECULE_STAGING_ANTHROPIC_API_KEY"
+                required_secret_value=""
+              fi
+              ;;
+            langgraph|hermes)
+              required_secret_name="MOLECULE_STAGING_OPENAI_KEY"
+              required_secret_value="${E2E_OPENAI_API_KEY:-}"
+              ;;
+            *)
+              echo "::warning::Unknown E2E_RUNTIME='${E2E_RUNTIME}' — skipping LLM-key check"
+              required_secret_name=""
+              required_secret_value="present"
+              ;;
+          esac
+          if [ -n "$required_secret_name" ] && [ -z "$required_secret_value" ]; then
+            echo "::error::${required_secret_name} secret not set for runtime=${E2E_RUNTIME} — workspaces will fail at boot with 'No provider API key found'"
+            exit 2
+          fi
+          echo "LLM key present ✓ (runtime=${E2E_RUNTIME}, key=${required_secret_name}, len=${#required_secret_value})"
+
+      - name: CP staging health preflight
+        run: |
+          code=$(curl -sS -o /dev/null -w "%{http_code}" --max-time 10 "$MOLECULE_CP_URL/health")
+          if [ "$code" != "200" ]; then
+            echo "::error::Staging CP unhealthy (got HTTP $code). Skipping — not a workspace bug."
+            exit 1
+          fi
+          echo "Staging CP healthy ✓"
+
+      - name: Run full-lifecycle E2E
+        id: e2e
+        run: bash tests/e2e/test_staging_full_saas.sh
+
+      # Belt-and-braces teardown: the test script itself installs a trap
+      # for EXIT/INT/TERM, but if the GH runner itself is cancelled (e.g.
+      # someone pushes a new commit and workflow concurrency is set to
+      # cancel), the trap may not fire. This `always()` step runs even on
+      # cancellation and attempts the delete a second time. The admin
+      # DELETE endpoint is idempotent so double-invoking is safe.
+      - name: Teardown safety net (runs on cancel/failure)
+        if: always()
+        env:
+          ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+        run: |
+          # Best-effort: find any e2e-YYYYMMDD-* orgs matching this run and
+          # nuke them. Catches the case where the script died before
+          # exporting its slug.
+          set +e
+          orgs=$(curl -sS "$MOLECULE_CP_URL/cp/admin/orgs" \
+            -H "Authorization: Bearer $ADMIN_TOKEN" 2>/dev/null \
+            | python3 -c "
+          import json, sys, os, datetime
+          run_id = os.environ.get('GITHUB_RUN_ID', '')
+          d = json.load(sys.stdin)
+          # ONLY sweep slugs from *this* CI run. Previously the filter was
+          # f'e2e-{today}-' which stomped on parallel CI runs AND any manual
+          # E2E probes a dev was running against staging (incident 2026-04-21
+          # 15:02Z: this workflow's safety net deleted an unrelated manual
+          # run's tenant 1s after it hit 'running').
+          # Sweep both today AND yesterday's UTC dates so a run that crosses
+          # midnight still matches its own slug — see the 2026-04-26→27
+          # canvas-safety-net incident for the same bug class.
+          today = datetime.date.today()
+          yesterday = today - datetime.timedelta(days=1)
+          dates = (today.strftime('%Y%m%d'), yesterday.strftime('%Y%m%d'))
+          if run_id:
+              prefixes = tuple(f'e2e-{d}-{run_id}-' for d in dates)
+          else:
+              prefixes = tuple(f'e2e-{d}-' for d in dates)
+          candidates = [o['slug'] for o in d.get('orgs', [])
+                        if any(o.get('slug','').startswith(p) for p in prefixes)
+                        and o.get('instance_status') not in ('purged',)]
+          print('\n'.join(candidates))
+          " 2>/dev/null)
+          # Per-slug verified DELETE (was `>/dev/null || true` — see
+          # molecule-controlplane#420). Surface non-2xx as a workflow
+          # warning naming the leaked slug; don't exit 1 (sweeper is
+          # the safety net within ~45 min).
+          leaks=()
+          for slug in $orgs; do
+            echo "Safety-net teardown: $slug"
+            # Tempfile-routed -w + set +e/-e prevents curl-exit-code
+            # pollution of the captured status (lint-curl-status-capture.yml).
+            set +e
+            curl -sS -o /tmp/saas-cleanup.out -w "%{http_code}" \
+              -X DELETE "$MOLECULE_CP_URL/cp/admin/tenants/$slug" \
+              -H "Authorization: Bearer $ADMIN_TOKEN" \
+              -H "Content-Type: application/json" \
+              -d "{\"confirm\":\"$slug\"}" >/tmp/saas-cleanup.code
+            set -e
+            code=$(cat /tmp/saas-cleanup.code 2>/dev/null || echo "000")
+            if [ "$code" = "200" ] || [ "$code" = "204" ]; then
+              echo "[teardown] deleted $slug (HTTP $code)"
+            else
+              echo "::warning::saas teardown for $slug returned HTTP $code — sweep-stale-e2e-orgs will catch it within ~45 min. Body: $(head -c 300 /tmp/saas-cleanup.out 2>/dev/null)"
+              leaks+=("$slug")
+            fi
+          done
+          if [ ${#leaks[@]} -gt 0 ]; then
+            echo "::warning::saas teardown left ${#leaks[@]} leak(s): ${leaks[*]}"
+          fi
+          exit 0
@@ -0,0 +1,171 @@
+name: E2E Staging Sanity (leak-detection self-check)
+
+# Periodic assertion that the teardown safety nets in e2e-staging-saas
+# and canary-staging actually work. Runs the E2E harness with
+# E2E_INTENTIONAL_FAILURE=1, which poisons the tenant admin token after
+# the org is provisioned. The workspace-provision step then fails, the
+# script exits non-zero, and the EXIT trap + workflow always()-step
+# must still tear down cleanly.
+#
+# A green run means:
+#   - The script exited non-zero (intentional failure caught)
+#   - The trap fired teardown
+#   - The leak-detection poll found zero orphan orgs
+#
+# A red run means the teardown path itself is broken — act on this the
+# same way you'd act on a canary failure (the whole E2E safety net is
+# compromised until it's fixed).
+#
+# Cadence: once a week, Monday 06:00 UTC. Drift-slow, not per-PR — the
+# teardown path rarely changes, and a weekly heartbeat is enough to
+# catch silent regressions in cleanup code paths.
+
+on:
+  schedule:
+    - cron: '0 6 * * 1'
+  workflow_dispatch:
+
+concurrency:
+  # Shares the group with canary + full so they don't collide on
+  # staging org-create quota.
+  group: e2e-staging-sanity
+  cancel-in-progress: false
+
+permissions:
+  issues: write
+  contents: read
+
+jobs:
+  sanity:
+    name: Intentional-failure teardown sanity
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+
+    env:
+      MOLECULE_CP_URL: https://staging-api.moleculesai.app
+      MOLECULE_ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+      E2E_MODE: canary            # lean lifecycle; we only need the org to exist
+      E2E_RUNTIME: hermes
+      E2E_RUN_ID: "sanity-${{ github.run_id }}"
+      E2E_INTENTIONAL_FAILURE: "1"
+
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Verify admin token present
+        run: |
+          if [ -z "$MOLECULE_ADMIN_TOKEN" ]; then
+            echo "::error::MOLECULE_STAGING_ADMIN_TOKEN not set"
+            exit 2
+          fi
+
+      # Inverted assertion: the run MUST fail. If it passes, the
+      # E2E_INTENTIONAL_FAILURE path is broken (token not being
+      # poisoned correctly, or the harness silently recovered).
+      - name: Run harness — expecting exit !=0
+        id: harness
+        run: |
+          set +e
+          bash tests/e2e/test_staging_full_saas.sh
+          rc=$?
+          echo "harness_rc=$rc" >> "$GITHUB_OUTPUT"
+          # The only acceptable outcomes:
+          #   1 — harness failed mid-run, teardown ran, leak-check passed
+          #   (exit 4 means teardown left a leak — that's the real bug
+          #    this sanity check exists to catch)
+          if [ "$rc" = "1" ]; then
+            echo "✓ Harness failed as expected (rc=1); teardown trap ran, leak-check passed"
+            exit 0
+          elif [ "$rc" = "0" ]; then
+            echo "::error::Harness succeeded under E2E_INTENTIONAL_FAILURE=1 — the poisoning path is broken"
+            exit 1
+          elif [ "$rc" = "4" ]; then
+            echo "::error::LEAK DETECTED (rc=4) — teardown failed to clean up the org. Safety net broken."
+            exit 4
+          else
+            echo "::error::Unexpected rc=$rc — neither clean-failure nor leak. Investigate harness."
+            exit 1
+          fi
+
+      - name: Open issue if safety net is broken
+        if: failure()
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        with:
+          script: |
+            const title = "🚨 E2E teardown safety net broken";
+            const runURL = `https://github.com/${context.repo.owner}/${context.repo.repo}/actions/runs/${context.runId}`;
+            const body =
+              `The weekly sanity run (E2E_INTENTIONAL_FAILURE=1) did not exit ` +
+              `as expected. This means one of:\n` +
+              `  - poisoning didn't actually cause failure (test harness regression), OR\n` +
+              `  - teardown left an orphan org (leak detection caught a real bug)\n\n` +
+              `Run: ${runURL}\n\n` +
+              `This is higher priority than a canary failure — the whole ` +
+              `E2E safety net can't be trusted until this is resolved.`;
+
+            const { data: existing } = await github.rest.issues.listForRepo({
+              owner: context.repo.owner, repo: context.repo.repo,
+              state: 'open', labels: 'e2e-safety-net',
+            });
+            const match = existing.find(i => i.title === title);
+            if (match) {
+              await github.rest.issues.createComment({
+                owner: context.repo.owner, repo: context.repo.repo,
+                issue_number: match.number,
+                body: `Still broken. ${runURL}`,
+              });
+            } else {
+              await github.rest.issues.create({
+                owner: context.repo.owner, repo: context.repo.repo,
+                title, body,
+                labels: ['e2e-safety-net', 'bug', 'priority-high'],
+              });
+            }
+
+      # Belt-and-braces: if teardown left anything behind, nuke it here
+      # so we don't bleed staging quota. Different label from the
+      # always()-steps in the other workflows so sanity-only orgs get
+      # cleaned up by sanity runs.
+      - name: Teardown safety net
+        if: always()
+        env:
+          ADMIN_TOKEN: ${{ secrets.MOLECULE_STAGING_ADMIN_TOKEN }}
+        run: |
+          set +e
+          orgs=$(curl -sS "$MOLECULE_CP_URL/cp/admin/orgs" \
+            -H "Authorization: Bearer $ADMIN_TOKEN" 2>/dev/null \
+            | python3 -c "
+          import json, sys
+          d = json.load(sys.stdin)
+          today = __import__('datetime').date.today().strftime('%Y%m%d')
+          candidates = [o['slug'] for o in d.get('orgs', [])
+                        if o.get('slug','').startswith(f'e2e-canary-{today}-sanity-')
+                        and o.get('status') not in ('purged',)]
+          print('\n'.join(candidates))
+          " 2>/dev/null)
+          # Per-slug verified DELETE — see molecule-controlplane#420.
+          # Failures surface as workflow warnings; the sweeper is the
+          # safety net within ~45 min.
+          leaks=()
+          for slug in $orgs; do
+            # Tempfile-routed -w + set +e/-e prevents curl-exit-code
+            # pollution of the captured status (lint-curl-status-capture.yml).
+            set +e
+            curl -sS -o /tmp/sanity-cleanup.out -w "%{http_code}" \
+              -X DELETE "$MOLECULE_CP_URL/cp/admin/tenants/$slug" \
+              -H "Authorization: Bearer $ADMIN_TOKEN" \
+              -H "Content-Type: application/json" \
+              -d "{\"confirm\":\"$slug\"}" >/tmp/sanity-cleanup.code
+            set -e
+            code=$(cat /tmp/sanity-cleanup.code 2>/dev/null || echo "000")
+            if [ "$code" = "200" ] || [ "$code" = "204" ]; then
+              echo "[teardown] deleted $slug (HTTP $code)"
+            else
+              echo "::warning::sanity teardown for $slug returned HTTP $code — sweep-stale-e2e-orgs will catch it within ~45 min. Body: $(head -c 300 /tmp/sanity-cleanup.out 2>/dev/null)"
+              leaks+=("$slug")
+            fi
+          done
+          if [ ${#leaks[@]} -gt 0 ]; then
+            echo "::warning::sanity teardown left ${#leaks[@]} leak(s): ${leaks[*]}"
+          fi
+          exit 0
--- a/Show More
+++ b/Show More