test(canvas): add MemoryTab tests (42 cases)

Cover awareness dashboard expand/collapse, iframe with workspaceId in URL, status grid, KV memory list, expand/collapse entries, add/edit/delete memory entries, JSON parsing, TTL support, 409 conflict retry hint, error states, and refresh. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
chore: retrigger CI after rebase to main
2026-05-12 03:46:50 +00:00 · 2026-05-12 03:46:50 +00:00 · 2026-05-12 03:38:40 +00:00 · 2026-05-12 03:37:52 +00:00 · 2026-05-12 03:34:57 +00:00 · 2026-05-12 03:34:13 +00:00
34 changed files with 4029 additions and 1163 deletions
@@ -49,11 +49,11 @@ if [ "$MERGED" != "true" ]; then
  exit 0
 fi

-MERGE_SHA=$(echo "$PR" | jq -r '.merge_commit_sha // empty')
-MERGED_BY=$(echo "$PR" | jq -r '.merged_by.login // "unknown"')
-TITLE=$(echo "$PR" | jq -r '.title // ""')
-BASE_BRANCH=$(echo "$PR" | jq -r '.base.ref // "main"')
-HEAD_SHA=$(echo "$PR" | jq -r '.head.sha // empty')
+MERGE_SHA=$(echo "$PR" | jq -r '.merge_commit_sha // empty') || true
+MERGED_BY=$(echo "$PR" | jq -r '.merged_by.login // "unknown"') || true
+TITLE=$(echo "$PR" | jq -r '.title // ""') || true
+BASE_BRANCH=$(echo "$PR" | jq -r '.base.ref // "main"') || true
+HEAD_SHA=$(echo "$PR" | jq -r '.head.sha // empty') || true

 if [ -z "$MERGE_SHA" ]; then
  echo "::warning::PR #${PR_NUMBER} merged=true but no merge_commit_sha — cannot evaluate force-merge."
@@ -75,7 +75,7 @@ STATUS=$(curl -sS -H "$AUTH" \
 declare -A CHECK_STATE
 while IFS=$'\t' read -r ctx state; do
  [ -n "$ctx" ] && CHECK_STATE[$ctx]="$state"
-done < <(echo "$STATUS" | jq -r '.statuses // [] | .[] | "\(.context)\t\(.status)"')
+done < <(echo "$STATUS" | jq -r '.statuses // [] | .[] | "\(.context)\t\(.status)"') || true

 # 4. For each required check, was it green at merge? YAML block scalars
 #    (`|`) leave a trailing newline; skip blank/whitespace-only lines.
@@ -97,7 +97,7 @@ fi

 # 5. Emit structured audit event.
 NOW=$(date -u +%Y-%m-%dT%H:%M:%SZ)
-FAILED_JSON=$(printf '%s\n' "${FAILED_CHECKS[@]}" | jq -R . | jq -s .)
+FAILED_JSON=$(printf '%s\n' "${FAILED_CHECKS[@]}" | jq -R . | jq -s .) || true

 # Print as a single-line JSON so Vector's parse_json transform can pick
 # it up cleanly from docker_logs.
@@ -301,7 +301,19 @@ def expected_context(job_key: str, workflow_name: str = "ci") -> str:
 # Drift detection
 # --------------------------------------------------------------------------
 def detect_drift(branch: str) -> tuple[list[str], dict]:
-    """Returns (findings, debug). Empty findings == no drift."""
+    """Returns (findings, debug). Empty findings == no drift.
+
+    Raises:
+        ApiError: propagated from the protection fetch only when the
+                  failure is likely a transient Gitea outage (5xx).
+                  403/404 from the protection endpoint is treated as
+                  "cannot determine drift for this branch" — a token-
+                  scope issue (missing repo-admin on DRIFT_BOT_TOKEN) or
+                  a repo with no protection set should not turn the
+                  hourly cron red. The workflow continues to the next
+                  branch; no [ci-drift] issue is filed for a branch
+                  whose protection cannot be read.
+    """
    findings: list[str] = []

    ci_doc = load_yaml(CI_WORKFLOW_PATH)
@@ -313,9 +325,50 @@ def detect_drift(branch: str) -> tuple[list[str], dict]:
    env_set = required_checks_env(audit_doc)

    # Protection
-    # api() raises ApiError on non-2xx; let it propagate so a transient
-    # 500 fails the run loudly rather than producing a "no drift" lie.
-    _, protection = api("GET", f"/repos/{OWNER}/{NAME}/branch_protections/{branch}")
+    # api() raises ApiError on non-2xx. Transient 5xx should fail loud.
+    # 403/404 means the token lacks repo-admin scope (Gitea 1.22.6's
+    # branch_protections endpoint requires it — see DRIFT_BOT_TOKEN
+    # provisioning trail in ci-required-drift.yml). Treat as
+    # "cannot determine drift for this branch" — skip without turning
+    # the workflow red. Surface a clear diagnostic so the operator
+    # knows what to fix.
+    contexts: set[str] = set()
+    protection_path = f"/repos/{OWNER}/{NAME}/branch_protections/{branch}"
+    try:
+        _, protection = api("GET", protection_path)
+    except ApiError as e:
+        # Isolate the HTTP status from the error message.
+        http_status: int | None = None
+        msg = str(e)
+        # ApiError message format: "{method} {path} → HTTP {status}: {body}"
+        import re as _re
+
+        m = _re.search(r"HTTP (\d{3})", msg)
+        if m:
+            http_status = int(m.group(1))
+        if http_status in (403, 404):
+            # Token lacks scope OR branch has no protection. Cannot
+            # determine drift — skip this branch. Do NOT exit non-zero;
+            # the issue IS the alarm, not a red workflow.
+            sys.stderr.write(
+                f"::error::GET {protection_path} returned HTTP {http_status} — "
+                f"DRIFT_BOT_TOKEN lacks repo-admin scope (Gitea 1.22.6 "
+                f"requires it for this endpoint) OR branch has no protection "
+                f"configured. Cannot determine drift for {branch}; "
+                f"skipping. Fix: grant repo-admin to mc-drift-bot or "
+                f"configure protection on {branch}.\n"
+            )
+            debug = {
+                "branch": branch,
+                "ci_jobs": sorted(jobs),
+                "sentinel_needs": sorted(needs),
+                "protection_contexts_skipped": True,
+                "protection_http_status": http_status,
+                "audit_env_checks": sorted(env_set),
+            }
+            return [], debug
+        # 5xx — propagate (transient outage, fail loud per design).
+        raise
    if not isinstance(protection, dict):
        sys.stderr.write(
            f"::error::protection response for {branch} not a JSON object\n"
@@ -96,16 +96,27 @@ API="https://${GITEA_HOST}/api/v1"
 AUTH="Authorization: token ${GITEA_TOKEN}"
 echo "::notice::tier-check start: repo=$OWNER/$NAME pr=$PR_NUMBER author=$PR_AUTHOR"

-# Sanity: token resolves to a user
-WHOAMI=$(curl -sS -H "$AUTH" "${API}/user" | jq -r '.login // ""')
+# Sanity: token resolves to a user.
+# Use || true on the jq pipeline so that set -euo pipefail (line 45) does not
+# cause the script to exit prematurely when the token is empty/invalid — the
+# if check below handles that case gracefully. Without || true, a 401 from an
+# empty/invalid token causes jq to exit 1, triggering set -e and exiting the
+# entire script before SOP_FAIL_OPEN can be evaluated (the check is in the jq-
+# install block; if jq is already on PATH, that block is skipped entirely).
+WHOAMI=$(curl -sS -H "$AUTH" "${API}/user" | jq -r '.login // ""') || true
 if [ -z "$WHOAMI" ]; then
  echo "::error::GITEA_TOKEN cannot resolve a user via /api/v1/user — check the token scope and that the secret is wired correctly."
+  if [ "${SOP_FAIL_OPEN:-}" = "1" ]; then
+    echo "::warning::SOP_FAIL_OPEN=1 — exiting 0 so CI does not block."
+    exit 0
+  fi
  exit 1
 fi
 echo "::notice::token resolves to user: $WHOAMI"

-# 1. Read tier label
-LABELS=$(curl -sS -H "$AUTH" "${API}/repos/${OWNER}/${NAME}/issues/${PR_NUMBER}/labels" | jq -r '.[].name')
+# 1. Read tier label. || true ensures set -euo pipefail does not abort the
+# script if curl or jq fails (e.g. 401 from empty token).
+LABELS=$(curl -sS -H "$AUTH" "${API}/repos/${OWNER}/${NAME}/issues/${PR_NUMBER}/labels" | jq -r '.[].name') || true
 TIER=""
 for L in $LABELS; do
  case "$L" in
@@ -176,17 +187,25 @@ fi
 # 4. Resolve all team names → IDs
 # /orgs/{org}/teams/{slug}/... endpoints don't exist on Gitea 1.22;
 # we use /teams/{id}.
+# set +e prevents set -e from aborting the script if curl fails (e.g. empty token).
 ORG_TEAMS_FILE=$(mktemp)
 trap 'rm -f "$ORG_TEAMS_FILE"' EXIT
+set +e
 HTTP_CODE=$(curl -sS -o "$ORG_TEAMS_FILE" -w '%{http_code}' -H "$AUTH" \
  "${API}/orgs/${OWNER}/teams")
-debug "teams-list HTTP=$HTTP_CODE size=$(wc -c <"$ORG_TEAMS_FILE")"
+_HTTP_EXIT=$?
+set -e
+debug "teams-list HTTP=$HTTP_CODE (curl exit=$_HTTP_EXIT) size=$(wc -c <"$ORG_TEAMS_FILE")"
 if [ "${SOP_DEBUG:-}" = "1" ]; then
  echo "  [debug] teams-list body (first 300 chars):" >&2
  head -c 300 "$ORG_TEAMS_FILE" >&2; echo >&2
 fi
-if [ "$HTTP_CODE" != "200" ]; then
-  echo "::error::GET /orgs/${OWNER}/teams returned HTTP $HTTP_CODE — token likely lacks read:org scope."
+if [ "$_HTTP_EXIT" -ne 0 ] || [ "$HTTP_CODE" != "200" ]; then
+  echo "::error::GET /orgs/${OWNER}/teams failed (curl exit=$_HTTP_EXIT HTTP=$HTTP_CODE) — token may lack read:org scope or be invalid."
+  if [ "${SOP_FAIL_OPEN:-}" = "1" ]; then
+    echo "::warning::SOP_FAIL_OPEN=1 — exiting 0 so CI does not block."
+    exit 0
+  fi
  exit 1
 fi

@@ -231,9 +250,22 @@ for _t in $_all_teams; do
  debug "team-id: $_t → $_id"
 done

-# 5. Read approving reviewers
+# 5. Read approving reviewers. set +e disables set -e temporarily so that curl
+# failures (e.g. empty/invalid token → HTTP 401) do not abort the script before
+# SOP_FAIL_OPEN is evaluated. set -e is restored immediately after.
+set +e
 REVIEWS=$(curl -sS -H "$AUTH" "${API}/repos/${OWNER}/${NAME}/pulls/${PR_NUMBER}/reviews")
-APPROVERS=$(echo "$REVIEWS" | jq -r '[.[] | select(.state=="APPROVED") | .user.login] | unique | .[]')
+_REVIEWS_EXIT=$?
+set -e
+if [ $_REVIEWS_EXIT -ne 0 ] || [ -z "$REVIEWS" ]; then
+  echo "::error::Failed to fetch reviews (curl exit=$_REVIEWS_EXIT) — token may be invalid or unreachable."
+  if [ "${SOP_FAIL_OPEN:-}" = "1" ]; then
+    echo "::warning::SOP_FAIL_OPEN=1 — exiting 0 so CI does not block."
+    exit 0
+  fi
+  exit 1
+fi
+APPROVERS=$(echo "$REVIEWS" | jq -r '[.[] | select(.state=="APPROVED") | .user.login] | unique | .[]') || true
 if [ -z "$APPROVERS" ]; then
  echo "::error::No approving reviews on this PR. Set SOP_DEBUG=1 and re-run for diagnostics."
  exit 1
@@ -0,0 +1,688 @@
+#!/usr/bin/env python3
+"""status-reaper — Option B compensating-status POST for Gitea 1.22.6's
+hardcoded `(push)` suffix on default-branch commit statuses.
+
+Tracking: this PR (workflow + script + tests + audit issue). Sibling
+bots: internal#327 (publish-runtime-bot), internal#328 (mc-drift-bot).
+Upstream RFC: internal#80. Persona provisioned by sub-agent aefaac1b
+(2026-05-11 21:39Z; Gitea uid 94, scope=write:repository).
+
+What this script does, per `.gitea/workflows/status-reaper.yml` invocation:
+
+  1. Walk `.gitea/workflows/*.yml`. For each file, build the workflow_id
+     using this resolution (per hongming-pc 22:08Z review):
+       - If YAML has top-level `name:` → use that.
+       - Else → use filename stem (basename minus `.yml`).
+     Fail-LOUD on:
+       - Two workflows resolving to the SAME identifier (collision).
+       - Any identifier containing `/` (it would break context parsing
+         downstream — Gitea uses ` / ` as the workflow/job separator).
+     Classify each by whether `on:` contains a `push:` trigger.
+
+  2. List the last N (=30, rev3 — widened from 10) commits on
+     WATCH_BRANCH via GET /repos/{o}/{r}/commits?sha={branch}&limit={N}.
+     rev2 sweeps N commits per tick instead of HEAD only — schedule
+     workflows post `failure` to whatever SHA was HEAD when they
+     COMPLETED, so by the next */5 tick main has often moved forward
+     and the red gets stranded on a stale commit. rev3 widens the
+     window from 10 → 30 because schedule workflows post `failure`
+     RETROACTIVELY (5-15 min after their merge); a 10-commit window
+     is narrower than the merge-cadence during a burst, so reds land
+     OUTSIDE the window before reaper sees them (Phase 1+2 evidence:
+     rev2 run 17057 at 02:46Z saw 185/0 contexts on 10 SHAs; direct
+     probe ~30min later showed ~25 fails on those same 10 SHAs).
+
+  3. For EACH SHA in the list:
+       - GET combined commit status. Per-SHA error isolation
+         (refinement #7): if this call raises ApiError or any 5xx,
+         LOG `::warning::` + continue to the next SHA. Different from
+         the single-HEAD pre-rev2 path where fail-loud was correct;
+         the sweep is best-effort across historical commits, so one
+         transient blip on a stale SHA must not strand reds on the
+         OTHER stale SHAs.
+       - If combined.state == "success": skip — cost optimization
+         (refinement #2), common case (most commits are green).
+       - Otherwise iterate per-context entries. For each entry where:
+           state == "failure" AND context.endswith(" (push)")
+         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 → 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
+             documenting the workaround + this script's path.
+
+  4. Exit 0. Re-running is idempotent — Gitea's commit-status table
+     stores the LATEST state-per-context, so the success POST sticks
+     even if another tick happens before the runner finishes.
+
+What it does NOT do:
+  - Touch any context NOT ending in ` (push)`. The required-checks on
+    main (verified 2026-05-11) all have ` (pull_request)` suffixes;
+    they CANNOT be reached by this code path.
+  - Compensate `error`/`pending` states. Only `failure` — the only one
+    Gitea emits for the hardcoded-suffix bug.
+  - Write to non-default branches. WATCH_BRANCH is sourced from
+    `github.event.repository.default_branch` in the workflow.
+  - Mutate workflows or runs. The Actions UI still shows the
+    underlying schedule-triggered run as failed; this script edits
+    the commit-status surface only.
+
+Halt conditions (script-level — orchestrator-level halts are in the
+workflow comments):
+  - PyYAML missing → fail-loud at import (no fallback parse).
+  - Workflow `name:` collision → exit 1 with ::error:: message.
+  - Workflow `name:` containing `/` → exit 1 with ::error:: message.
+  - Ambiguous `on:` shape (e.g. neither str/list/dict) → treat as
+    "has_push_trigger=True" and log ::notice:: (preserve, never
+    compensate the unknown).
+  - api() non-2xx → raise ApiError, fail the workflow run loudly so
+    a subsequent tick retries (per
+    `feedback_api_helper_must_raise_not_return_dict`).
+
+Local dry-run (no network):
+    GITEA_TOKEN=... GITEA_HOST=git.moleculesai.app REPO=owner/repo \\
+      WATCH_BRANCH=main WORKFLOWS_DIR=.gitea/workflows \\
+      python3 .gitea/scripts/status-reaper.py --dry-run
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import sys
+import urllib.error
+import urllib.parse
+import urllib.request
+from pathlib import Path
+from typing import Any
+
+import yaml  # PyYAML 6.0.2 — installed by the workflow before this runs.
+
+
+# --------------------------------------------------------------------------
+# Environment
+# --------------------------------------------------------------------------
+def _env(key: str, *, default: str = "") -> str:
+    """Read an env var with a default. Module-import-safe — tests can
+    import this script without setting the full env contract."""
+    return os.environ.get(key, default)
+
+
+GITEA_TOKEN = _env("GITEA_TOKEN")
+GITEA_HOST = _env("GITEA_HOST")
+REPO = _env("REPO")
+WATCH_BRANCH = _env("WATCH_BRANCH", default="main")
+WORKFLOWS_DIR = _env("WORKFLOWS_DIR", default=".gitea/workflows")
+
+OWNER, NAME = (REPO.split("/", 1) + [""])[:2] if REPO else ("", "")
+API = f"https://{GITEA_HOST}/api/v1" if GITEA_HOST else ""
+
+# Compensating-status description prefix. Used as the marker so a human
+# auditing commit statuses can tell at a glance that the green was
+# synthetic, not a real CI pass. Kept stable; downstream tooling
+# (e.g. main-red-watchdog visual diff) MAY key on it.
+COMPENSATION_DESCRIPTION = (
+    "Compensated by status-reaper (workflow has no push: trigger; "
+    "Gitea 1.22.6 hardcoded-suffix bug — see .gitea/scripts/status-reaper.py)"
+)
+
+# Context suffix the reaper acts on. Gitea hardcodes this for ALL
+# default-branch workflow runs.
+PUSH_SUFFIX = " (push)"
+
+
+def _require_runtime_env() -> None:
+    """Enforce env contract — called from `main()` only.
+
+    Tests import individual functions without setting the full env
+    contract. Mirrors `main-red-watchdog.py`/`ci-required-drift.py`.
+    """
+    for key in ("GITEA_TOKEN", "GITEA_HOST", "REPO", "WATCH_BRANCH", "WORKFLOWS_DIR"):
+        if not os.environ.get(key):
+            sys.stderr.write(f"::error::missing required env var: {key}\n")
+            sys.exit(2)
+
+
+# --------------------------------------------------------------------------
+# Tiny HTTP helper — raises on non-2xx + on JSON-decode-of-expected-JSON.
+# --------------------------------------------------------------------------
+class ApiError(RuntimeError):
+    """Raised when a Gitea API call cannot be trusted to have succeeded.
+
+    Per `feedback_api_helper_must_raise_not_return_dict`: soft-failure is
+    opt-in via `expect_json=False`, never the default. A pre-fix
+    implementation that returned `{}` on non-2xx would skip the
+    compensating POST on a transient outage AND silently lose the
+    failed-status enumeration, painting main green via omission.
+    """
+
+
+def api(
+    method: str,
+    path: str,
+    *,
+    body: dict | None = None,
+    query: dict[str, str] | None = None,
+    expect_json: bool = True,
+) -> tuple[int, Any]:
+    """Tiny HTTP helper around urllib. Same contract as
+    `main-red-watchdog.py` and `ci-required-drift.py` so behaviour
+    is cross-checkable."""
+    url = f"{API}{path}"
+    if query:
+        url = f"{url}?{urllib.parse.urlencode(query)}"
+    data = None
+    headers = {
+        "Authorization": f"token {GITEA_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, method=method, data=data, headers=headers)
+    try:
+        with urllib.request.urlopen(req, timeout=30) as resp:
+            raw = resp.read()
+            status = resp.status
+    except urllib.error.HTTPError as e:
+        raw = e.read()
+        status = e.code
+
+    if not (200 <= status < 300):
+        snippet = raw[:500].decode("utf-8", errors="replace") if raw else ""
+        raise ApiError(f"{method} {path} -> HTTP {status}: {snippet}")
+
+    if not raw:
+        return status, None
+    try:
+        return status, json.loads(raw)
+    except json.JSONDecodeError as e:
+        if expect_json:
+            raise ApiError(
+                f"{method} {path} -> HTTP {status} but body is not JSON: {e}"
+            ) from e
+        return status, {"_raw": raw.decode("utf-8", errors="replace")}
+
+
+# --------------------------------------------------------------------------
+# Workflow scan + classification
+# --------------------------------------------------------------------------
+def _on_block(doc: dict) -> Any:
+    """Extract the `on:` block from a parsed YAML doc.
+
+    PyYAML parses bareword `on:` as Python `True` (YAML 1.1 boolean
+    spec — `on/off/yes/no` are booleans). The actual key in the dict
+    is therefore `True`, NOT the string `"on"`. We accept both for
+    forward-compat with YAML 1.2 loaders (which keep it as `"on"`).
+    """
+    if True in doc:
+        return doc[True]
+    return doc.get("on")
+
+
+def _has_push_trigger(on_block: Any, workflow_id: str) -> bool:
+    """Return True if `on:` block declares a `push` trigger.
+
+    Accepts the three common shapes:
+      - str: `on: push` → True only if == "push"
+      - list: `on: [push, pull_request]` → True if "push" in list
+      - dict: `on: { push: {...}, schedule: ... }` → True if "push" key
+
+    Defensive: for anything else (including None/empty), return True
+    so we preserve rather than over-compensate. Logged via ::notice::.
+    """
+    if isinstance(on_block, str):
+        return on_block == "push"
+    if isinstance(on_block, list):
+        return "push" in on_block
+    if isinstance(on_block, dict):
+        return "push" in on_block
+    # None or unexpected shape — preserve, log.
+    print(
+        f"::notice::ambiguous on: for {workflow_id}; preserving "
+        f"(value={on_block!r}, type={type(on_block).__name__})"
+    )
+    return True
+
+
+def scan_workflows(workflows_dir: str) -> dict[str, bool]:
+    """Walk `workflows_dir` and return `{workflow_id: has_push_trigger}`.
+
+    Workflow ID resolution (per hongming-pc 22:08Z review):
+      - Top-level `name:` if present.
+      - Else filename stem (basename minus `.yml`).
+
+    Fail-LOUD on:
+      - Two workflows resolving to the same ID (collision).
+      - Any ID containing `/` (would break ` / `-separated context
+        parsing on the downstream side).
+
+    Returns a dict for O(1) lookup in the per-status loop.
+    """
+    path = Path(workflows_dir)
+    if not path.is_dir():
+        # Workflow dir missing → no workflows to classify. Empty map is
+        # safe: per-status loop will hit "unknown workflow; skip" for
+        # every entry, which is correct (we cannot tell if a push
+        # trigger exists, so we preserve).
+        print(f"::warning::workflows dir not found: {workflows_dir}")
+        return {}
+
+    out: dict[str, bool] = {}
+    sources: dict[str, str] = {}  # workflow_id -> source file (for collision msg)
+
+    for yml in sorted(path.glob("*.yml")):
+        try:
+            with yml.open() as f:
+                doc = yaml.safe_load(f)
+        except yaml.YAMLError as e:
+            # A malformed YAML in the workflows dir is a real defect
+            # (the workflow wouldn't load on Gitea either). Surface it
+            # and keep going — the reaper's job is to compensate the
+            # OTHER workflows even if one is broken.
+            print(f"::warning::yaml parse failed for {yml.name}: {e}; skip")
+            continue
+        if not isinstance(doc, dict):
+            print(f"::warning::workflow {yml.name} not a dict; skip")
+            continue
+
+        # Resolve workflow_id.
+        name_field = doc.get("name")
+        if isinstance(name_field, str) and name_field.strip():
+            workflow_id = name_field.strip()
+        else:
+            workflow_id = yml.stem  # basename minus .yml
+
+        # Halt-loud: `/` in workflow_id breaks ` / ` context parsing.
+        if "/" in workflow_id:
+            sys.stderr.write(
+                f"::error::workflow name contains '/' which breaks "
+                f"context parsing: {workflow_id} (file={yml.name})\n"
+            )
+            sys.exit(1)
+
+        # Halt-loud: ID collision.
+        if workflow_id in out:
+            sys.stderr.write(
+                f"::error::workflow name collision detected: {workflow_id} "
+                f"(files: {sources[workflow_id]} + {yml.name})\n"
+            )
+            sys.exit(1)
+
+        on_block = _on_block(doc)
+        out[workflow_id] = _has_push_trigger(on_block, workflow_id)
+        sources[workflow_id] = yml.name
+
+    return out
+
+
+# --------------------------------------------------------------------------
+# Gitea reads
+# --------------------------------------------------------------------------
+def get_head_sha(branch: str) -> str:
+    """HEAD SHA of `branch`. Raises ApiError on non-2xx."""
+    _, body = api("GET", f"/repos/{OWNER}/{NAME}/branches/{branch}")
+    if not isinstance(body, dict):
+        raise ApiError(f"branch {branch} response not a JSON object")
+    commit = body.get("commit")
+    if not isinstance(commit, dict):
+        raise ApiError(f"branch {branch} response missing `commit` object")
+    sha = commit.get("id") or commit.get("sha")
+    if not isinstance(sha, str) or len(sha) < 7:
+        raise ApiError(f"branch {branch} response has no usable commit SHA")
+    return sha
+
+
+def get_combined_status(sha: str) -> dict:
+    """Combined commit status for `sha`. Gitea returns:
+        {
+          "state": "success" | "failure" | "pending" | "error",
+          "statuses": [
+            {"context": "...", "state": "...", "target_url": "...",
+             "description": "..."},
+            ...
+          ],
+          ...
+        }
+    Raises ApiError on non-2xx.
+    """
+    _, body = api("GET", f"/repos/{OWNER}/{NAME}/commits/{sha}/status")
+    if not isinstance(body, dict):
+        raise ApiError(f"status for {sha} response not a JSON object")
+    return body
+
+
+# --------------------------------------------------------------------------
+# Context parsing
+# --------------------------------------------------------------------------
+def parse_push_context(context: str) -> tuple[str, str] | None:
+    """Parse `<workflow_name> / <job_name> (push)` into
+    (workflow_name, job_name).
+
+    Returns None if the context doesn't match the shape (caller skips).
+    Strict: requires the trailing ` (push)` and at least one ` / `
+    separator. Anything else is left alone.
+    """
+    if not context.endswith(PUSH_SUFFIX):
+        return None
+    head = context[: -len(PUSH_SUFFIX)]  # strip " (push)"
+    if " / " not in head:
+        # No workflow/job separator — not the bug shape we compensate.
+        return None
+    workflow_name, job_name = head.split(" / ", 1)
+    return workflow_name, job_name
+
+
+# --------------------------------------------------------------------------
+# Compensating POST
+# --------------------------------------------------------------------------
+def post_compensating_status(
+    sha: str,
+    context: str,
+    target_url: str | None,
+    *,
+    dry_run: bool = False,
+) -> None:
+    """POST a `state=success` to /repos/{o}/{r}/statuses/{sha} with the
+    given context. Gitea de-dups by context (latest write wins).
+
+    Description references this script so the compensation is
+    self-documenting on the commit's status view.
+    """
+    payload: dict[str, Any] = {
+        "context": context,
+        "state": "success",
+        "description": COMPENSATION_DESCRIPTION,
+    }
+    # Echo the original target_url when present so a human auditing
+    # the (now-green) compensated status can still reach the run logs
+    # that produced the original red.
+    if target_url:
+        payload["target_url"] = target_url
+
+    if dry_run:
+        print(
+            f"::notice::[dry-run] would compensate {context!r} on {sha[:10]} "
+            f"with state=success"
+        )
+        return
+
+    api("POST", f"/repos/{OWNER}/{NAME}/statuses/{sha}", body=payload)
+    print(f"::notice::compensated {context!r} on {sha[:10]} (state=success)")
+
+
+# --------------------------------------------------------------------------
+# Main reap loop
+# --------------------------------------------------------------------------
+def reap(
+    workflow_trigger_map: dict[str, bool],
+    combined: dict,
+    sha: str,
+    *,
+    dry_run: bool = False,
+) -> dict[str, Any]:
+    """Walk `combined.statuses[]` and compensate where appropriate.
+
+    Per-SHA worker. The multi-SHA orchestrator (`reap_branch`) calls
+    this once per stale main commit each tick.
+
+    Returns counters for observability:
+      {compensated, preserved_real_push, preserved_unknown,
+       preserved_non_failure, preserved_non_push_suffix,
+       preserved_unparseable,
+       compensated_contexts: [<context>, ...]}
+
+    `compensated_contexts` is rev2-added so `reap_branch` can build
+    `compensated_per_sha` without re-deriving it from the POST stream.
+    """
+    counters: dict[str, Any] = {
+        "compensated": 0,
+        "preserved_real_push": 0,
+        "preserved_unknown": 0,
+        "preserved_non_failure": 0,
+        "preserved_non_push_suffix": 0,
+        "preserved_unparseable": 0,
+        "compensated_contexts": [],
+    }
+
+    statuses = combined.get("statuses") or []
+    for s in statuses:
+        if not isinstance(s, dict):
+            continue
+        context = s.get("context") or ""
+        state = s.get("state") or ""
+
+        # Only `failure` is the bug shape. `error`/`pending`/`success`
+        # left alone — they have other meanings.
+        if state != "failure":
+            counters["preserved_non_failure"] += 1
+            continue
+
+        # Only `(push)`-suffix contexts hit the hardcoded-suffix bug.
+        # Branch-protection required checks (e.g. `Secret scan / Scan
+        # diff (pull_request)`) are NOT reachable from this path.
+        if not context.endswith(PUSH_SUFFIX):
+            counters["preserved_non_push_suffix"] += 1
+            continue
+
+        parsed = parse_push_context(context)
+        if parsed is None:
+            # Has ` (push)` suffix but missing ` / ` separator — not
+            # the bug shape. Preserve.
+            counters["preserved_unparseable"] += 1
+            continue
+        workflow_name, _job_name = parsed
+
+        if workflow_name not in workflow_trigger_map:
+            # Real workflow but renamed/deleted/external — we can't
+            # tell if it has push trigger. Conservative: preserve.
+            print(f"::notice::unknown workflow {workflow_name!r}; skip")
+            counters["preserved_unknown"] += 1
+            continue
+
+        if workflow_trigger_map[workflow_name]:
+            # Real push trigger → real defect signal. Preserve.
+            counters["preserved_real_push"] += 1
+            continue
+
+        # Class-O: schedule/dispatch/etc.-only workflow with a fake
+        # (push) status from Gitea's hardcoded-suffix bug. Compensate.
+        post_compensating_status(
+            sha, context, s.get("target_url"), dry_run=dry_run
+        )
+        counters["compensated"] += 1
+        counters["compensated_contexts"].append(context)
+
+    return counters
+
+
+# --------------------------------------------------------------------------
+# rev2: multi-SHA sweep over the last N commits on WATCH_BRANCH
+# --------------------------------------------------------------------------
+# How many main commits to sweep per tick. Sized to cover a burst-merge
+# window where multiple PRs land in the 5-min interval between reaper
+# ticks. Older reds falling off the window is acceptable — they were
+# already stale enough that the schedule-run that posted them has long
+# since been overwritten by a real push trigger. See `reference_post_
+# suspension_pipeline` for the merge-cadence baseline.
+#
+# rev3 (2026-05-12, hongming-pc2 GO 03:25Z): widened from 10 → 30.
+# rev2 (limit=10) shipped 01:48Z and ran 6/6 ticks post-merge with
+# `compensated:0` despite ~25 stranded reds visible on those same 10
+# SHAs ~30min later. Root cause: schedule workflows post `failure`
+# RETROACTIVELY 5-15 min after their merge, so by the time reaper's
+# next */5 tick lands, the stranded red is on a SHA that has already
+# fallen out of a 10-commit window during a burst-merge period.
+# Trades window-width-cheap for cadence-loady (per hongming-pc2):
+# kept `*/5` cron unchanged; only the window-N is widened.
+DEFAULT_SWEEP_LIMIT = 30
+
+
+def list_recent_commit_shas(branch: str, limit: int) -> list[str]:
+    """List the most recent `limit` commit SHAs on `branch`, newest
+    first.
+
+    Wraps GET /repos/{o}/{r}/commits?sha={branch}&limit={limit}. Gitea
+    1.22.6 returns a JSON list of commit objects each with a `sha` key
+    (verified via vendor-truth probe 2026-05-11 against
+    git.moleculesai.app — `feedback_smoke_test_vendor_truth_not_shape_match`).
+
+    Raises ApiError on non-2xx OR on unexpected response shape. This is
+    a HARD halt — without the commit list the sweep can't proceed. (The
+    per-SHA error isolation downstream is a different concern: tolerating
+    a transient 5xx on ONE commit's status is best-effort; losing the
+    commit list itself means we don't even know which commits to try.)
+    """
+    _, body = api(
+        "GET",
+        f"/repos/{OWNER}/{NAME}/commits",
+        query={"sha": branch, "limit": str(limit)},
+    )
+    if not isinstance(body, list):
+        raise ApiError(
+            f"commits listing for {branch} not a JSON array "
+            f"(got {type(body).__name__})"
+        )
+    shas: list[str] = []
+    for entry in body:
+        if not isinstance(entry, dict):
+            continue
+        sha = entry.get("sha")
+        if isinstance(sha, str) and len(sha) >= 7:
+            shas.append(sha)
+    if not shas:
+        raise ApiError(
+            f"commits listing for {branch} returned no usable SHAs"
+        )
+    return shas
+
+
+def reap_branch(
+    workflow_trigger_map: dict[str, bool],
+    branch: str,
+    *,
+    limit: int = DEFAULT_SWEEP_LIMIT,
+    dry_run: bool = False,
+) -> dict[str, Any]:
+    """Sweep the last `limit` commits on `branch`, applying `reap()`
+    to each (with per-SHA error isolation).
+
+    Returns aggregated counters PLUS rev2 observability fields:
+      - scanned_shas: how many SHAs we actually iterated
+      - compensated_per_sha: {<sha_full>: [<context>, ...]} — only
+        SHAs that actually got at least one compensation are included
+    """
+    shas = list_recent_commit_shas(branch, limit)
+
+    aggregate: dict[str, Any] = {
+        "scanned_shas": 0,
+        "compensated": 0,
+        "preserved_real_push": 0,
+        "preserved_unknown": 0,
+        "preserved_non_failure": 0,
+        "preserved_non_push_suffix": 0,
+        "preserved_unparseable": 0,
+        "compensated_per_sha": {},
+    }
+
+    for sha in shas:
+        aggregate["scanned_shas"] += 1
+
+        # Per-SHA error isolation (refinement #7). One transient blip
+        # on a historical commit must NOT abort the whole tick — the
+        # OTHER stale SHAs may still hold strandable reds.
+        try:
+            combined = get_combined_status(sha)
+        except ApiError as e:
+            print(
+                f"::warning::get_combined_status({sha[:10]}) failed; "
+                f"skipping this SHA: {e}"
+            )
+            continue
+
+        # Cost optimization (refinement #2): the common case is a green
+        # commit. Skip the per-context loop entirely when combined is
+        # already success — saves a tight loop over ~20 statuses per SHA
+        # on green commits, the dominant majority.
+        if combined.get("state") == "success":
+            continue
+
+        per_sha = reap(
+            workflow_trigger_map, combined, sha, dry_run=dry_run
+        )
+
+        # Aggregate scalar counters.
+        for key in (
+            "compensated",
+            "preserved_real_push",
+            "preserved_unknown",
+            "preserved_non_failure",
+            "preserved_non_push_suffix",
+            "preserved_unparseable",
+        ):
+            aggregate[key] += per_sha[key]
+
+        # Record per-SHA compensated contexts (only when non-empty —
+        # keep the summary readable when most SHAs are no-ops).
+        contexts = per_sha.get("compensated_contexts") or []
+        if contexts:
+            aggregate["compensated_per_sha"][sha] = list(contexts)
+
+    return aggregate
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--dry-run",
+        action="store_true",
+        help="Skip the compensating POST; print what would be done.",
+    )
+    parser.add_argument(
+        "--limit",
+        type=int,
+        default=DEFAULT_SWEEP_LIMIT,
+        help=(
+            "How many recent commits on WATCH_BRANCH to sweep per tick "
+            f"(default: {DEFAULT_SWEEP_LIMIT})."
+        ),
+    )
+    args = parser.parse_args()
+
+    _require_runtime_env()
+
+    workflow_trigger_map = scan_workflows(WORKFLOWS_DIR)
+    print(
+        f"::notice::scanned {len(workflow_trigger_map)} workflows; "
+        f"push-triggered={sum(1 for v in workflow_trigger_map.values() if v)}, "
+        f"class-O candidates={sum(1 for v in workflow_trigger_map.values() if not v)}"
+    )
+
+    counters = reap_branch(
+        workflow_trigger_map,
+        WATCH_BRANCH,
+        limit=args.limit,
+        dry_run=args.dry_run,
+    )
+
+    # Observability: print one JSON line summarising the tick. Loki
+    # ingestion via the runner's stdout (`source="gitea-actions"`).
+    print(
+        "status-reaper summary: "
+        + json.dumps(
+            {
+                "branch": WATCH_BRANCH,
+                "dry_run": args.dry_run,
+                "limit": args.limit,
+                **counters,
+            },
+            sort_keys=True,
+        )
+    )
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -317,7 +317,8 @@ JQ_FILTER='.[]

 T12_INPUT='[{"state":"APPROVED","dismissed":false,"user":{"login":"core-devops"}},{"state":"CHANGES_REQUESTED","dismissed":false,"user":{"login":"bob"}},{"state":"APPROVED","dismissed":false,"user":{"login":"alice"}},{"state":"APPROVED","dismissed":true,"user":{"login":"carol"}}]'

-T12_CANDIDATES=$(echo "$T12_INPUT" | /tmp/jq -r "$JQ_FILTER" 2>/dev/null | sort -u)
+JQ_CMD=$(command -v jq 2>/dev/null || echo /tmp/jq)
+T12_CANDIDATES=$(echo "$T12_INPUT" | "$JQ_CMD" -r "$JQ_FILTER" 2>/dev/null | sort -u)
 assert_contains "T12 jq: core-devops (non-author APPROVED) in candidates" "core-devops" "$T12_CANDIDATES"
 assert_eq "T12 jq: alice (author) NOT in candidates" "" "$(echo "$T12_CANDIDATES" | grep '^alice$' || true)"
 assert_eq "T12 jq: carol (dismissed) NOT in candidates" "" "$(echo "$T12_CANDIDATES" | grep '^carol$' || true)"
@@ -23,11 +23,11 @@
 # `feedback_behavior_based_ast_gates` — NOT grep-by-name. That way
 # job renames or matrix-expansion-induced churn produce honest signal.
 #
-# IMPORTANT — TRANSITIONAL STATE: molecule-core's ci.yml does NOT yet
-# contain the `all-required` sentinel job (RFC §4 Phase 4 adds it).
-# Until Phase 4 lands the detector will hard-fail with exit 3 on the
-# missing sentinel. That's intentional: a red workflow on a 5-min cron
-# is louder than a silent issue and forces Phase 4 to land soon.
+# NOTE on protection endpoint scope: `GET /repos/.../branch_protections/{branch}`
+# requires repo-admin role in Gitea 1.22.6. If DRIFT_BOT_TOKEN lacks it,
+# the script skips that branch with a clear ::error:: diagnostic and exits 0
+# (the issue IS the alarm, not a red workflow). See provisioning trail in
+# the run step's GITEA_TOKEN env comment.

 name: ci-required-drift

@@ -148,6 +148,21 @@ jobs:
      - if: needs.changes.outputs.platform == 'true'
        name: Run golangci-lint
        run: golangci-lint run --timeout 3m ./... || true
+      - if: needs.changes.outputs.platform == 'true'
+        name: Diagnostic — per-package verbose 60s
+        run: |
+          set +e
+          go test -race -v -timeout 60s ./internal/handlers/... 2>&1 | tee /tmp/test-handlers.log
+          handlers_exit=$?
+          go test -race -v -timeout 60s ./internal/pendinguploads/... 2>&1 | tee /tmp/test-pu.log
+          pu_exit=$?
+          echo "::group::handlers exit=$handlers_exit (last 100 lines)"
+          tail -100 /tmp/test-handlers.log
+          echo "::endgroup::"
+          echo "::group::pendinguploads exit=$pu_exit (last 100 lines)"
+          tail -100 /tmp/test-pu.log
+          echo "::endgroup::"
+        continue-on-error: true
      - if: needs.changes.outputs.platform == 'true'
        name: Run tests with race detection and coverage
        run: go test -race -coverprofile=coverage.out ./...
@@ -71,8 +71,12 @@ jobs:
        run: |
          set -euo pipefail
          # 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=$(python3 -c "
-            import urllib.request, json, os
+            import socket, urllib.request, json, os
+            socket.setdefaulttimeout(15)
            token = os.environ['GITEA_TOKEN']
            req = urllib.request.Request(
                'https://git.moleculesai.app/api/v1/repos/${{ github.repository }}/pulls?state=open&limit=100',
@@ -220,12 +220,14 @@ jobs:
        run: |
          set -euo pipefail
          if [ -z "${MOLECULE_GITEA_TOKEN}" ]; then
-            echo "::error::AUTO_SYNC_TOKEN secret is empty — register the devops-engineer persona PAT in repo Actions secrets"
-            exit 1
+            echo "::warning::AUTO_SYNC_TOKEN not set — using anonymous clone (repos are public per manifest.json OSS contract)"
          fi
          mkdir -p .tenant-bundle-deps
+          # Strip JSON5 comments before jq parsing — Integration Tester appends
+          # `// Triggered by ...` which breaks `jq` in clone-manifest.sh.
+          sed '/^[[:space:]]*\/\//d' manifest.json > .manifest-stripped.json
          bash scripts/clone-manifest.sh \
-            manifest.json \
+            .manifest-stripped.json \
            .tenant-bundle-deps/workspace-configs-templates \
            .tenant-bundle-deps/org-templates \
            .tenant-bundle-deps/plugins
@@ -37,6 +37,11 @@ name: main-red-watchdog
 # "unknown on type" when `workflow_dispatch.inputs.X` is present. Revisit
 # when Gitea ≥ 1.23 is fleet-wide.
 on:
+  # SCHEDULE RE-ENABLED 2026-05-12 rev3 — interim disable (mc#645) reverted alongside
+  # status-reaper rev3 (widen-window). Job-level timeout-minutes raised 5 → 15 below
+  # to absorb runner-saturation latency without spurious cancels (the original cascade
+  # cause). If runner-saturation root persists, the dedicated-runner-label split
+  # remains the structural next step (tracked separately).
  schedule:
    # Hourly at :05 — task spec calls for "off-zero" (`5 * * * *`),
    # offset from :17 (ci-required-drift) and :00 (peak cron load).
@@ -58,7 +63,12 @@ concurrency:
 jobs:
  watchdog:
    runs-on: ubuntu-latest
-    timeout-minutes: 5
+    # rev3 (2026-05-12, mc#645 revert): raised 5 → 15 to absorb runner-saturation
+    # latency. Original 5min cap was producing 124-style cancels under load,
+    # which fed the very `[main-red]` issues this workflow files (self-poisoning).
+    # 15min is still well below Gitea-default 6h job ceiling; if a real hang
+    # occurs the issue-file path is still the alarm surface.
+    timeout-minutes: 15
    steps:
      - name: Check out repo (script lives at .gitea/scripts/)
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
@@ -54,6 +54,12 @@ env:
 jobs:
  build-and-push:
    name: Build & push canvas image
+    # 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.
    continue-on-error: true
@@ -79,8 +85,10 @@ jobs:
        run: |
          set -euo pipefail
          echo "::group::Docker daemon health check"
+          echo "Runner: ${HOSTNAME:-unknown}"
          docker info 2>&1 | head -5 || {
            echo "::error::Docker daemon is not accessible at /var/run/docker.sock"
+            echo "::error::Runner: ${HOSTNAME:-unknown}"
            echo "::error::Check: (1) daemon running, (2) runner user in docker group, (3) sock perms 660+"
            exit 1
          }
@@ -52,6 +52,12 @@ env:

 jobs:
  build-and-push:
+    # 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
    steps:
      - name: Checkout
@@ -68,8 +74,10 @@ jobs:
        run: |
          set -euo pipefail
          echo "::group::Docker daemon health check"
+          echo "Runner: ${HOSTNAME:-unknown}"
          docker info 2>&1 | head -5 || {
            echo "::error::Docker daemon is not accessible at /var/run/docker.sock"
+            echo "::error::Runner: ${HOSTNAME:-unknown}"
            echo "::error::Check: (1) daemon is running, (2) runner user is in docker group, (3) sock permissions are 660+"
            exit 1
          }
@@ -96,8 +104,11 @@ jobs:
          # 2026-05-08 migration). The token is only needed for private repos.
          # Do NOT require it — a missing secret would fail the build unnecessarily.
          mkdir -p .tenant-bundle-deps
+          # Strip JSON5 comments before jq parsing — Integration Tester appends
+          # `// Triggered by ...` which breaks `jq` in clone-manifest.sh.
+          sed '/^[[:space:]]*\/\//d' manifest.json > .manifest-stripped.json
          bash scripts/clone-manifest.sh \
-            manifest.json \
+            .manifest-stripped.json \
            .tenant-bundle-deps/workspace-configs-templates \
            .tenant-bundle-deps/org-templates \
            .tenant-bundle-deps/plugins
@@ -0,0 +1,70 @@
+name: review-check-tests
+
+# Runs review-check.sh regression tests on every PR + push that touches
+# the evaluator script or its test fixtures.
+#
+# Follows RFC#324 follow-up (issue #540):
+#   .gitea/scripts/review-check.sh is load-bearing for PR merge gates.
+#   It has ZERO production CI coverage. This workflow closes that gap.
+#
+# Design choices:
+#   - Bash test harness (not bats). The existing test_review_check.sh
+#     uses a custom assert_eq/assert_contains framework that is already
+#     working and covers all 13 acceptance criteria (issue #540 §Acceptance).
+#     Converting to bats would be refactoring, not closing the gap.
+#   - No bats dependency: the runner-base image needs no extra tooling.
+#   - continue-on-error: false — these tests must pass; a failure means
+#     the review-gate evaluator is broken and must not be merged.
+
+on:
+  push:
+    branches: [main, staging]
+    paths:
+      - '.gitea/scripts/review-check.sh'
+      - '.gitea/scripts/tests/test_review_check.sh'
+      - '.gitea/scripts/tests/_review_check_fixture.py'
+      - '.gitea/workflows/review-check-tests.yml'
+  pull_request:
+    branches: [main, staging]
+    paths:
+      - '.gitea/scripts/review-check.sh'
+      - '.gitea/scripts/tests/test_review_check.sh'
+      - '.gitea/scripts/tests/_review_check_fixture.py'
+      - '.gitea/workflows/review-check-tests.yml'
+  workflow_dispatch:
+
+env:
+  GITHUB_SERVER_URL: https://git.moleculesai.app
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  test:
+    name: review-check.sh regression tests
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Install jq
+        # Required for T12 jq-filter test case. Gitea Actions runners (ubuntu-latest
+        # label) do not bundle jq. Install via apt-get first (reliable for Ubuntu
+        # 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).
+        continue-on-error: true
+        run: |
+          if apt-get update -qq && apt-get install -y -qq jq; then
+            echo "::notice::jq installed via apt-get: $(jq --version)"
+          elif timeout 120 curl -sSL \
+            "https://github.com/jqlang/jq/releases/download/jq-1.7.1/jq-linux-amd64" \
+            -o /usr/local/bin/jq && chmod +x /usr/local/bin/jq; then
+            echo "::notice::jq binary downloaded: $(/usr/local/bin/jq --version)"
+          else
+            echo "::warning::jq install failed — apt-get and GitHub download both failed."
+          fi
+          jq --version 2>/dev/null || echo "::notice::jq not yet available — continuing"
+
+      - name: Run review-check.sh regression suite
+        run: bash .gitea/scripts/tests/test_review_check.sh
@@ -0,0 +1,121 @@
+# status-reaper — Option B (compensating-status POST) for Gitea 1.22.6's
+# hardcoded `(push)` suffix on default-branch commit statuses.
+#
+# Tracking: molecule-core#? (this PR), internal#327 (sibling publish-runtime-bot),
+# internal#328 (sibling mc-drift-bot), internal#80 (upstream RFC). Sister
+# bots already deployed under the same per-persona-identity contract
+# (`feedback_per_agent_gitea_identity_default`).
+#
+# Root cause:
+#   Gitea 1.22.6 emits commit-status context as
+#     `<workflow_name> / <job_name> (push)`
+#   for ANY workflow run on the default branch's HEAD commit, REGARDLESS
+#   of the trigger event. Schedule- and workflow_dispatch-triggered runs
+#   on `main` therefore appear as `(push)` failures on the latest main
+#   commit, painting main red via a fake-push status. Verified on runs
+#   14525 + 14526 via Phase 1 evidence (3 sub-agents). No upstream fix
+#   in 1.23-1.26.1 (sibling a6f20db1 research).
+#
+# Why a cron-driven reaper, not workflow_run:
+#   Gitea 1.22.6 does NOT support `on: workflow_run` (verified via
+#   modules/actions/workflows.go enumeration; sister a6f20db1). The
+#   only event-shaped option that fires is cron. 5min is chosen to
+#   sit BETWEEN ci-required-drift (`:17` hourly) and main-red-watchdog
+#   (`:05` hourly) so the reaper sweeps red before the watchdog files
+#   a `[main-red]` issue (would-be false-positive).
+#
+# What the reaper does each tick:
+#   1. Parse `.gitea/workflows/*.yml`, classify each by whether `on:`
+#      contains a `push:` trigger (see script for workflow_id resolution
+#      including `name:` collision and `/`-in-name fail-loud lints).
+#   2. GET combined status for main HEAD.
+#   3. For each `failure` status whose context ends ` (push)`:
+#      - if workflow has push trigger: PRESERVE (real defect signal).
+#      - if workflow has no push trigger: POST a compensating
+#        `state=success` with the same context and a description that
+#        documents the workaround.
+#
+# What it does NOT do:
+#   - Mutate non-`(push)`-suffix statuses (e.g. `(pull_request)` from
+#     branch_protections required-checks — verified safe 2026-05-11).
+#   - Auto-revert. Same reasoning as main-red-watchdog.
+#   - Cancel runs. The runs themselves stay visible in Actions UI; the
+#     fix is at the commit-status surface only.
+#
+# Removal path: drop this workflow when Gitea ≥ 1.24 ships with a
+# real fix for the hardcoded-suffix bug. Audit issue (filed post-merge)
+# tracks the deletion as a follow-up sweep.
+
+name: status-reaper
+
+# IMPORTANT — Gitea 1.22.6 parser quirk per
+# `feedback_gitea_workflow_dispatch_inputs_unsupported`: do NOT add an
+# `inputs:` block here. Gitea 1.22.6 rejects the whole workflow as
+# "unknown on type" when `workflow_dispatch.inputs.X` is present.
+on:
+  # 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
+# write surface is touched. checkout still needs `contents: read`.
+permissions:
+  contents: read
+
+# NOTE: NO `concurrency:` block is intentional.
+# Gitea 1.22.6 doesn't honor `cancel-in-progress: false`: queued ticks
+# of the same group get cancelled-with-started=0 instead of waiting
+# (DB-verified 2026-05-12, runs 16053/16085 of status-reaper.yml).
+# The reaper's POST /statuses/{sha} is idempotent — Gitea de-dups by
+# context — so concurrent ticks are safe; accept them rather than
+# serialise via the broken mechanism.
+
+jobs:
+  reap:
+    runs-on: ubuntu-latest
+    timeout-minutes: 3
+    steps:
+      - name: Check out repo at default-branch HEAD
+        # BASE checkout per `feedback_pull_request_target_workflow_from_base`.
+        # The script reads .gitea/workflows/*.yml from the working tree to
+        # classify trigger sets; we must read main's CURRENT state, not
+        # the SHA a stale schedule fired against.
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        with:
+          ref: ${{ github.event.repository.default_branch }}
+
+      - name: Set up Python (PyYAML for workflow `on:` parse)
+        # Pinned to 3.12 to match sibling watchdog / ci-required-drift.
+        uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065  # v5.6.0
+        with:
+          python-version: '3.12'
+
+      - name: Install PyYAML
+        # PyYAML is needed because shell-grep on `on:` misses list/string
+        # forms and nested `push: { paths: ... }`. Same install pattern
+        # as ci-required-drift.yml (sub-2s install, no wheel cache).
+        run: python -m pip install --quiet 'PyYAML==6.0.2'
+
+      - name: Compensate operational push-suffix failures on main
+        env:
+          # claude-status-reaper persona token; provisioned by sibling
+          # aefaac1b 2026-05-11. Owns write:repository scope to POST
+          # /statuses/{sha} but NOTHING ELSE
+          # (`feedback_per_agent_gitea_identity_default`).
+          GITEA_TOKEN: ${{ secrets.STATUS_REAPER_TOKEN }}
+          GITEA_HOST: git.moleculesai.app
+          REPO: ${{ github.repository }}
+          WATCH_BRANCH: ${{ github.event.repository.default_branch }}
+          WORKFLOWS_DIR: .gitea/workflows
+        run: python3 .gitea/scripts/status-reaper.py
@@ -0,0 +1,120 @@
+name: Weekly Platform-Go Surface
+
+# Surface latent vet/test errors on main by running the full Platform-Go
+# suite on a weekly cron regardless of whether the last push touched
+# workspace-server/.
+#
+# Background: ci.yml's `platform-build` job gates real work on
+# `if: needs.changes.outputs.platform == 'true'`. When no push touches
+# workspace-server/, the skip fires and the suite never executes on main.
+# Latent vet errors and test flakes can sit for weeks undetected.
+#
+# This workflow runs the full suite (build, vet, golangci-lint, tests with
+# coverage) every Monday at 04:17 UTC. Results are posted as commit statuses
+# but continue-on-error: true means they never block anything — they're
+# purely a noise-reduction signal for when the next workspace-server push
+# lands and would otherwise trigger the first real suite run.
+#
+# Why 04:17 UTC on Monday: off-peak, before the weekly sprint cycle starts.
+
+on:
+  schedule:
+    - cron: '17 4 * * 1'  # Mondays at 04:17 UTC
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  statuses: write
+
+jobs:
+  weekly-platform-go:
+    name: Weekly Platform-Go Surface
+    runs-on: ubuntu-latest
+    # continue-on-error: surface only, never block
+    continue-on-error: true
+    defaults:
+      run:
+        working-directory: workspace-server
+    steps:
+      - name: Checkout main
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          ref: main
+          fetch-depth: 1
+
+      - name: Set up Go
+        uses: actions/setup-go@40f1582b2485089dde7abd97c1529aa768e1baff # v5
+        with:
+          go-version: stable
+
+      - name: Go mod download
+        run: go mod download
+
+      - name: Build
+        run: go build ./cmd/server
+
+      # `go vet` is NOT `|| true`-guarded: surfacing latent vet errors on main is
+      # the whole point of this workflow (issue #567 — the motivating case was a
+      # `go vet` error in org_external.go that sat undetected on main for weeks).
+      # A vet error here fails the step → fails the job → shows red on the weekly
+      # commit. Per Gitea quirk #10 (job-level continue-on-error is ignored), that
+      # red surfaces on main — which is the intended signal, not a regression.
+      - name: go vet
+        run: go vet ./...
+
+      # golangci-lint stays `|| true`-guarded: lint is noisier (more false-
+      # positives than vet) and golangci-lint may not be pre-installed on every
+      # runner image — a `|| true` here keeps a missing-binary or lint-noise case
+      # from masking the vet/test signal above. Tighten to match ci.yml's lint
+      # gate if/when ci.yml's lint step becomes hard-failing.
+      - name: golangci-lint
+        run: golangci-lint run --timeout 3m ./... || true
+
+      - name: Tests with race detection + coverage
+        run: go test -race -coverprofile=coverage.out ./...
+
+      - name: Check coverage thresholds
+        run: |
+          set -e
+          TOTAL_FLOOR=25
+          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: ${TOTAL}%"
+          if awk "BEGIN{exit !(\$TOTAL < \$TOTAL_FLOOR)}"; then
+            echo "::error::Total coverage \${TOTAL}% is below the \${TOTAL_FLOOR}% floor."
+            exit 1
+          fi
+
+          ALLOWLIST=""
+          if [ -f ../.coverage-allowlist.txt ]; then
+            ALLOWLIST=$(grep -vE '^(#|[[:space:]]*$)' ../.coverage-allowlist.txt || true)
+          fi
+
+          FAILED=0
+          for path in "\${CRITICAL_PATHS[@]}"; do
+            while read -r file pct; do
+              [[ "$file" == *_test.go ]] && continue
+              [[ "$file" == *"$path"* ]] || continue
+              awk "BEGIN{exit !(\$pct < 10)}" || continue
+              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
+              echo "::error::Low coverage \${pct}% on \${rel} (below 10% in critical path \${path})"
+              FAILED=$((FAILED + 1))
+            done < <(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]}' | sort)
+          done
+          if [ "$FAILED" -gt 0 ]; then
+            echo "::error::\${FAILED} critical paths below 10% coverage — see above."
+            exit 1
+          fi
+          echo "Coverage thresholds: OK"
@@ -156,6 +156,16 @@ and run CI manually.
 | python-lint | pytest with coverage |
 | e2e-api | Full API test suite (62 tests) |
 | shellcheck | Shell script linting |
+| review-check-tests | `review-check.sh` evaluator regression suite (13 scenarios) |
+| ops-scripts | Python unittest suite for `scripts/*.py` |
+
+## Local Testing
+
+### review-check.sh
+```bash
+bash .gitea/scripts/tests/test_review_check.sh
+```
+Runs the full regression suite against a fixture HTTP server. No network access required.

 ## Code Style

@@ -402,7 +402,7 @@ function Row({ label, value, mono }: { label: string; value: string; mono?: bool
  );
 }

-function getSkills(card: Record<string, unknown> | null): { id: string; description?: string }[] {
+export function getSkills(card: Record<string, unknown> | null): { id: string; description?: string }[] {
  if (!card) return [];
  const skills = card.skills;
  if (!Array.isArray(skills)) return [];
@@ -0,0 +1,224 @@
+// @vitest-environment jsdom
+/**
+ * FilesTab: NotAvailablePanel + FilesToolbar coverage.
+ *
+ * NotAvailablePanel: pure presentational component — renders a "feature not
+ * available" placeholder for external-runtime workspaces.
+ * FilesToolbar: pure props-driven component — directory selector, file count,
+ * action buttons (New, Upload, Export, Clear, Refresh) with correct aria-labels.
+ *
+ * No @testing-library/jest-dom import — use textContent / className /
+ * getAttribute checks to avoid "expect is not defined" errors.
+ */
+import { afterEach, describe, expect, it, vi } from "vitest";
+import { cleanup, render, screen } from "@testing-library/react";
+import React from "react";
+
+import { FilesToolbar } from "../FilesToolbar";
+import { NotAvailablePanel } from "../NotAvailablePanel";
+
+// ─── afterEach ─────────────────────────────────────────────────────────────────
+
+afterEach(() => {
+  cleanup();
+  vi.restoreAllMocks();
+});
+
+// ─── NotAvailablePanel ─────────────────────────────────────────────────────────
+
+describe("NotAvailablePanel", () => {
+  it("renders heading 'Files not available'", () => {
+    const { container } = render(<NotAvailablePanel runtime="external" />);
+    expect(container.textContent).toContain("Files not available");
+  });
+
+  it("renders the runtime name in monospace", () => {
+    const { container } = render(<NotAvailablePanel runtime="external" />);
+    expect(container.textContent).toContain("external");
+    const spans = container.querySelectorAll("span");
+    const monoSpans = Array.from(spans).filter(
+      (s) => s.className && s.className.includes("font-mono"),
+    );
+    expect(monoSpans.length).toBeGreaterThan(0);
+  });
+
+  it("renders a Chat tab hint in description", () => {
+    const { container } = render(<NotAvailablePanel runtime="remote-agent" />);
+    expect(container.textContent).toContain("Chat tab");
+  });
+
+  it("SVG icon has aria-hidden=true", () => {
+    const { container } = render(<NotAvailablePanel runtime="external" />);
+    const svg = container.querySelector("svg");
+    expect(svg?.getAttribute("aria-hidden")).toBe("true");
+  });
+
+  it("renders without crashing for any runtime string", () => {
+    const { container } = render(<NotAvailablePanel runtime="unknown-runtime" />);
+    expect(container.textContent).toContain("unknown-runtime");
+  });
+
+  it("applies the correct layout classes to root div", () => {
+    const { container } = render(<NotAvailablePanel runtime="external" />);
+    const root = container.firstElementChild as HTMLElement;
+    expect(root.className).toContain("flex");
+    expect(root.className).toContain("flex-col");
+    expect(root.className).toContain("items-center");
+  });
+});
+
+// ─── FilesToolbar ───────────────────────────────────────────────────────────────
+
+describe("FilesToolbar", () => {
+  const noop = vi.fn();
+
+  function renderToolbar(props: Partial<React.ComponentProps<typeof FilesToolbar>> = {}) {
+    return render(
+      <FilesToolbar
+        root="/configs"
+        setRoot={noop}
+        fileCount={0}
+        onNewFile={noop}
+        onUpload={noop}
+        onDownloadAll={noop}
+        onClearAll={noop}
+        onRefresh={noop}
+        {...props}
+      />,
+    );
+  }
+
+  it("renders the directory selector with correct aria-label", () => {
+    const { container } = renderToolbar();
+    const select = container.querySelector("select");
+    expect(select?.getAttribute("aria-label")).toBe("File root directory");
+  });
+
+  it("directory selector has all four options", () => {
+    const { container } = renderToolbar();
+    const select = container.querySelector("select") as HTMLSelectElement;
+    const options = Array.from(select?.options ?? []);
+    const values = options.map((o) => o.value);
+    expect(values).toContain("/configs");
+    expect(values).toContain("/home");
+    expect(values).toContain("/workspace");
+    expect(values).toContain("/plugins");
+  });
+
+  it("calls setRoot when directory changes", () => {
+    const setRoot = vi.fn();
+    const { container } = renderToolbar({ setRoot });
+    const select = container.querySelector("select") as HTMLSelectElement;
+    select.value = "/home";
+    select.dispatchEvent(new Event("change", { bubbles: true }));
+    expect(setRoot).toHaveBeenCalledWith("/home");
+  });
+
+  it("displays the file count", () => {
+    const { container } = renderToolbar({ fileCount: 42 });
+    expect(container.textContent).toContain("42 files");
+  });
+
+  it("shows New + Upload + Clear buttons for /configs", () => {
+    const { container } = renderToolbar({ root: "/configs" });
+    const texts = Array.from(container.querySelectorAll("button")).map(
+      (b) => b.textContent?.trim(),
+    );
+    expect(texts).toContain("+ New");
+    expect(texts).toContain("Upload");
+    expect(texts).toContain("Clear");
+    expect(texts).toContain("Export");
+    expect(texts).toContain("↻");
+  });
+
+  it("hides New + Upload + Clear for /workspace", () => {
+    const { container } = renderToolbar({ root: "/workspace" });
+    const texts = Array.from(container.querySelectorAll("button")).map(
+      (b) => b.textContent?.trim(),
+    );
+    expect(texts).not.toContain("+ New");
+    expect(texts).not.toContain("Upload");
+    expect(texts).not.toContain("Clear");
+    expect(texts).toContain("Export");
+  });
+
+  it("hides New + Upload + Clear for /home", () => {
+    const { container } = renderToolbar({ root: "/home" });
+    const texts = Array.from(container.querySelectorAll("button")).map(
+      (b) => b.textContent?.trim(),
+    );
+    expect(texts).not.toContain("+ New");
+    expect(texts).not.toContain("Upload");
+    expect(texts).not.toContain("Clear");
+  });
+
+  it("hides New + Upload + Clear for /plugins", () => {
+    const { container } = renderToolbar({ root: "/plugins" });
+    const texts = Array.from(container.querySelectorAll("button")).map(
+      (b) => b.textContent?.trim(),
+    );
+    expect(texts).not.toContain("+ New");
+    expect(texts).not.toContain("Upload");
+    expect(texts).not.toContain("Clear");
+  });
+
+  it("New button has correct aria-label", () => {
+    const { container } = renderToolbar({ root: "/configs" });
+    const newBtn = container.querySelector('button[aria-label="Create new file"]');
+    expect(newBtn?.textContent?.trim()).toBe("+ New");
+  });
+
+  it("Export button has correct aria-label", () => {
+    const { container } = renderToolbar();
+    const exportBtn = container.querySelector('button[aria-label="Download all files"]');
+    expect(exportBtn?.textContent?.trim()).toBe("Export");
+  });
+
+  it("Clear button has correct aria-label", () => {
+    const { container } = renderToolbar({ root: "/configs" });
+    const clearBtn = container.querySelector('button[aria-label="Delete all files"]');
+    expect(clearBtn?.textContent?.trim()).toBe("Clear");
+  });
+
+  it("Refresh button has correct aria-label", () => {
+    const { container } = renderToolbar();
+    const refreshBtn = container.querySelector('button[aria-label="Refresh file list"]');
+    expect(refreshBtn?.textContent?.trim()).toBe("↻");
+  });
+
+  it("calls onNewFile when New button is clicked", () => {
+    const onNewFile = vi.fn();
+    const { container } = renderToolbar({ root: "/configs", onNewFile });
+    container.querySelector('button[aria-label="Create new file"]')!.click();
+    expect(onNewFile).toHaveBeenCalledTimes(1);
+  });
+
+  it("calls onDownloadAll when Export button is clicked", () => {
+    const onDownloadAll = vi.fn();
+    const { container } = renderToolbar({ onDownloadAll });
+    container.querySelector('button[aria-label="Download all files"]')!.click();
+    expect(onDownloadAll).toHaveBeenCalledTimes(1);
+  });
+
+  it("calls onClearAll when Clear button is clicked", () => {
+    const onClearAll = vi.fn();
+    const { container } = renderToolbar({ root: "/configs", onClearAll });
+    container.querySelector('button[aria-label="Delete all files"]')!.click();
+    expect(onClearAll).toHaveBeenCalledTimes(1);
+  });
+
+  it("calls onRefresh when Refresh button is clicked", () => {
+    const onRefresh = vi.fn();
+    const { container } = renderToolbar({ onRefresh });
+    container.querySelector('button[aria-label="Refresh file list"]')!.click();
+    expect(onRefresh).toHaveBeenCalledTimes(1);
+  });
+
+  it("applies focus-visible ring to all interactive buttons", () => {
+    const { container } = renderToolbar({ root: "/configs" });
+    const buttons = container.querySelectorAll("button");
+    for (const btn of buttons) {
+      expect(btn.className).toContain("focus-visible:ring-2");
+    }
+  });
+});
@@ -647,7 +647,7 @@ export function SkillsTab({ workspaceId, data }: Props) {
  );
 }

-function extractSkills(agentCard: Record<string, unknown> | null): SkillEntry[] {
+export function extractSkills(agentCard: Record<string, unknown> | null): SkillEntry[] {
  if (!agentCard) return [];
  const rawSkills = agentCard.skills;
  if (!Array.isArray(rawSkills)) return [];
@@ -0,0 +1,330 @@
+// @vitest-environment jsdom
+import { describe, it, expect, beforeEach, afterEach, vi } from "vitest";
+import { render, screen, cleanup, fireEvent } from "@testing-library/react";
+import React from "react";
+import { BudgetSection } from "../BudgetSection";
+import { api } from "@/lib/api";
+
+// Queue-based mock for the api module. Each api call shifts from the queue.
+// Tests push with qGet/qPatch and the module-level mockImplementation
+// reads from the queue.
+type QueueEntry = { body?: unknown; err?: Error };
+const apiQueue: QueueEntry[] = [];
+
+vi.mock("@/lib/api", () => ({
+  api: {
+    get: vi.fn(async (path: string) => {
+      const next = apiQueue.shift();
+      if (!next) throw new Error(`api.get queue exhausted at: ${path}`);
+      if (next.err) throw next.err;
+      return next.body;
+    }),
+    patch: vi.fn(async (path: string, _body?: unknown) => {
+      const next = apiQueue.shift();
+      if (!next) throw new Error(`api.patch queue exhausted at: ${path}`);
+      if (next.err) throw next.err;
+      return next.body;
+    }),
+  },
+}));
+
+afterEach(cleanup);
+
+beforeEach(() => {
+  apiQueue.length = 0;
+  vi.clearAllMocks();
+});
+
+const WS_ID = "budget-test-ws";
+
+function qGet(body: unknown) {
+  apiQueue.push({ body });
+}
+
+function qGetErr(status: number, msg: string) {
+  apiQueue.push({ err: new Error(`${msg}: ${status}`) });
+}
+
+function qPatch(body: unknown) {
+  apiQueue.push({ body });
+}
+
+function qPatchErr(status: number, msg: string) {
+  apiQueue.push({ err: new Error(`${msg}: ${status}`) });
+}
+
+function makeBudget(overrides: Partial<{
+  budget_limit: number | null;
+  budget_used: number;
+  budget_remaining: number | null;
+}> = {}) {
+  return {
+    budget_limit: 10_000,
+    budget_used: 3_500,
+    budget_remaining: 6_500,
+    ...overrides,
+  };
+}
+
+describe("BudgetSection", () => {
+  describe("loading state", () => {
+    it("shows loading indicator while fetching", async () => {
+      let resolveGet: (v: unknown) => void;
+      vi.mocked(api.get).mockImplementationOnce(
+        async () => new Promise((r) => { resolveGet = r as (v: unknown) => void; }),
+      );
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      expect(screen.getByTestId("budget-loading")).toBeTruthy();
+
+      // Resolve after render to verify state clears
+      resolveGet!(makeBudget());
+      await vi.waitFor(() => {
+        expect(screen.queryByTestId("budget-loading")).toBeNull();
+      });
+    });
+  });
+
+  describe("fetch error state", () => {
+    it("shows error message on non-402 fetch failure", async () => {
+      qGetErr(500, "Internal Server Error");
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-fetch-error")).toBeTruthy();
+      });
+      expect(screen.getByTestId("budget-fetch-error")!.textContent).toContain("500");
+    });
+
+    it("shows 402 as exceeded banner, not fetch error", async () => {
+      // 402 means the budget limit was hit — different UX from a network/API error.
+      qGetErr(402, "Payment Required");
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-exceeded-banner")).toBeTruthy();
+      });
+      expect(screen.queryByTestId("budget-fetch-error")).toBeNull();
+    });
+  });
+
+  describe("budget loaded — display", () => {
+    it("renders used / limit stats row", async () => {
+      qGet(makeBudget({ budget_limit: 10_000, budget_used: 3_500 }));
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-used-value")!.textContent).toBe("3,500");
+      });
+      expect(screen.getByTestId("budget-limit-value")!.textContent).toBe("10,000");
+    });
+
+    it("renders 'Unlimited' when budget_limit is null", async () => {
+      qGet(makeBudget({ budget_limit: null, budget_used: 1_000, budget_remaining: null }));
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-limit-value")!.textContent).toBe("Unlimited");
+      });
+    });
+
+    it("renders remaining credits when present", async () => {
+      qGet(makeBudget({ budget_limit: 10_000, budget_used: 3_500, budget_remaining: 6_500 }));
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-remaining")!.textContent).toContain("6,500");
+        expect(screen.getByTestId("budget-remaining")!.textContent).toContain("credits remaining");
+      });
+    });
+
+    it("omits remaining credits when budget_remaining is null", async () => {
+      qGet(makeBudget({ budget_limit: 10_000, budget_used: 3_500, budget_remaining: null }));
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.queryByTestId("budget-remaining")).toBeNull();
+      });
+    });
+
+    it("caps progress bar at 100% when used > limit", async () => {
+      // Over-limit: 12000 used of 10000 limit should show 100%, not 120%.
+      qGet(makeBudget({ budget_limit: 10_000, budget_used: 12_000, budget_remaining: null }));
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        const fill = screen.getByTestId("budget-progress-fill");
+        expect(fill.getAttribute("style")).toContain("100%");
+      });
+    });
+
+    it("omits progress bar when budget_limit is null (unlimited)", async () => {
+      qGet(makeBudget({ budget_limit: null, budget_used: 5_000, budget_remaining: null }));
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.queryByTestId("budget-progress-fill")).toBeNull();
+      });
+    });
+  });
+
+  describe("budget exceeded (402)", () => {
+    it("shows exceeded banner when load returns 402", async () => {
+      qGetErr(402, "Payment Required");
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-exceeded-banner")).toBeTruthy();
+        expect(screen.getByTestId("budget-exceeded-banner")!.textContent).toContain("Budget exceeded");
+      });
+    });
+
+    it("clears exceeded banner after successful save", async () => {
+      qGetErr(402, "Payment Required");
+      qPatch(makeBudget({ budget_limit: 50_000, budget_used: 0, budget_remaining: 50_000 }));
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-exceeded-banner")).toBeTruthy();
+      });
+
+      const input = screen.getByTestId("budget-limit-input");
+      fireEvent.change(input, { target: { value: "50000" } });
+
+      const saveBtn = screen.getByTestId("budget-save-btn");
+      fireEvent.click(saveBtn);
+
+      await vi.waitFor(() => {
+        expect(screen.queryByTestId("budget-exceeded-banner")).toBeNull();
+      });
+    });
+  });
+
+  describe("save flow", () => {
+    it("shows save error on non-402 patch failure", async () => {
+      qGet(makeBudget());
+      qPatchErr(500, "Internal Server Error");
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-limit-input")).toBeTruthy();
+      });
+
+      const saveBtn = screen.getByTestId("budget-save-btn");
+      fireEvent.click(saveBtn);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-save-error")).toBeTruthy();
+        expect(screen.getByTestId("budget-save-error")!.textContent).toContain("500");
+      });
+    });
+
+    it("updates input to new limit value after successful save", async () => {
+      qGet(makeBudget({ budget_limit: 10_000 }));
+      qPatch(makeBudget({ budget_limit: 20_000 }));
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      // Wait for the input to appear (loading → loaded)
+      await vi.waitFor(() => {
+        expect(screen.queryByTestId("budget-loading")).toBeNull();
+      });
+
+      const input = screen.getByTestId("budget-limit-input") as HTMLInputElement;
+      // Debug: check what values are rendered
+      const limitValue = screen.getByTestId("budget-limit-value")?.textContent;
+      expect(input.value).toBe("10000"); // initial value from API
+      expect(limitValue).toBe("10,000");
+
+      fireEvent.change(input, { target: { value: "20000" } });
+      expect(input.value).toBe("20000");
+
+      fireEvent.click(screen.getByTestId("budget-save-btn"));
+
+      await vi.waitFor(() => {
+        expect((screen.getByTestId("budget-limit-input") as HTMLInputElement).value).toBe("20000");
+      });
+    });
+
+    it("sends null when input is cleared (unlimited)", async () => {
+      qGet(makeBudget({ budget_limit: 10_000 }));
+      qPatch(makeBudget({ budget_limit: null }));
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-limit-input")).toBeTruthy();
+      });
+
+      const input = screen.getByTestId("budget-limit-input") as HTMLInputElement;
+      fireEvent.change(input, { target: { value: "" } });
+      fireEvent.click(screen.getByTestId("budget-save-btn"));
+
+      await vi.waitFor(() => {
+        // After save with null limit, input should show empty (unlimited)
+        expect(input.value).toBe("");
+      });
+    });
+
+    it("shows saving state on button while patch is in flight", async () => {
+      qGet(makeBudget());
+      let resolvePatch: (v: unknown) => void;
+      vi.mocked(api.patch).mockImplementationOnce(
+        async () => new Promise((r) => { resolvePatch = r as (v: unknown) => void; }),
+      );
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-limit-input")).toBeTruthy();
+      });
+
+      fireEvent.change(screen.getByTestId("budget-limit-input"), { target: { value: "50000" } });
+      fireEvent.click(screen.getByTestId("budget-save-btn"));
+
+      const btn = screen.getByTestId("budget-save-btn");
+      expect(btn.textContent).toContain("Saving");
+
+      resolvePatch!(makeBudget({ budget_limit: 50_000 }));
+      await vi.waitFor(() => {
+        expect(btn.textContent).toContain("Save");
+      });
+    });
+  });
+
+  describe("isApiError402 — regression coverage", () => {
+    it("classifies ': 402' with space as 402", async () => {
+      qGetErr(402, "Payment Required");
+      qPatch(makeBudget());
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-exceeded-banner")).toBeTruthy();
+      });
+    });
+
+    it("classifies non-402 error messages as regular fetch errors", async () => {
+      qGetErr(503, "Service Unavailable");
+
+      render(<BudgetSection workspaceId={WS_ID} />);
+
+      await vi.waitFor(() => {
+        expect(screen.getByTestId("budget-fetch-error")).toBeTruthy();
+      });
+      expect(screen.queryByTestId("budget-exceeded-banner")).toBeNull();
+    });
+  });
+});
@@ -0,0 +1,140 @@
+// @vitest-environment jsdom
+/**
+ * Unit tests for extractSkills — pure helper from SkillsTab.
+ *
+ * Covers: null card, non-array skills, empty skills, full skill entries
+ * (id, name, description, tags, examples), id-only fallback, name-only
+ * fallback, string coercion, array coercion for tags/examples,
+ * filtering entries with no id after coercion, empty string id (filtered).
+ */
+import { describe, it, expect } from "vitest";
+import { extractSkills } from "../SkillsTab";
+
+describe("extractSkills", () => {
+  it("returns [] for null card", () => {
+    expect(extractSkills(null)).toEqual([]);
+  });
+
+  it("returns [] when card.skills is not an array", () => {
+    expect(extractSkills({ skills: undefined })).toEqual([]);
+    expect(extractSkills({ skills: "not-an-array" })).toEqual([]);
+    expect(extractSkills({ skills: { id: "x" } })).toEqual([]);
+  });
+
+  it("returns [] for empty skills array", () => {
+    expect(extractSkills({ skills: [] })).toEqual([]);
+  });
+
+  it("maps a fully-populated skill entry", () => {
+    const card = {
+      skills: [
+        {
+          id: "code_search",
+          name: "Code Search",
+          description: "Semantic code search",
+          tags: ["search", "code"],
+          examples: ["Find unused exports", "Search by AST pattern"],
+        },
+      ],
+    };
+    expect(extractSkills(card)).toEqual([
+      {
+        id: "code_search",
+        name: "Code Search",
+        description: "Semantic code search",
+        tags: ["search", "code"],
+        examples: ["Find unused exports", "Search by AST pattern"],
+      },
+    ]);
+  });
+
+  it("uses name as id when id is absent", () => {
+    const card = { skills: [{ name: "web_scraper" }] };
+    expect(extractSkills(card)).toEqual([
+      { id: "web_scraper", name: "web_scraper", description: "", tags: [], examples: [] },
+    ]);
+  });
+
+  it("uses id as name when name is absent", () => {
+    const card = { skills: [{ id: "legacy_skill" }] };
+    expect(extractSkills(card)).toEqual([
+      { id: "legacy_skill", name: "legacy_skill", description: "", tags: [], examples: [] },
+    ]);
+  });
+
+  it("filters out entries with neither id nor name", () => {
+    // id: String(undefined || undefined || "") → "" → filtered (id.length = 0)
+    const card = { skills: [{ description: "orphan entry" }] };
+    expect(extractSkills(card)).toEqual([]);
+  });
+
+  it("filters out entries with no id after string coercion", () => {
+    // id resolves to "" after String(undefined || null || {})
+    const card = { skills: [{ id: null, name: null }] };
+    expect(extractSkills(card)).toEqual([]);
+  });
+
+  it("filters out entries with empty-string id", () => {
+    const card = { skills: [{ id: "", name: "" }] };
+    expect(extractSkills(card)).toEqual([]);
+  });
+
+  it("coerces numeric tags to strings", () => {
+    const card = { skills: [{ id: "x", tags: [1, "two", 3] }] };
+    expect(extractSkills(card)).toEqual([
+      { id: "x", name: "x", description: "", tags: ["1", "two", "3"], examples: [] },
+    ]);
+  });
+
+  it("coerces non-array tags to empty array", () => {
+    const card = { skills: [{ id: "x", tags: "not-an-array" }] };
+    expect(extractSkills(card)).toEqual([
+      { id: "x", name: "x", description: "", tags: [], examples: [] },
+    ]);
+  });
+
+  it("coerces non-array examples to empty array", () => {
+    const card = { skills: [{ id: "x", examples: 42 }] };
+    expect(extractSkills(card)).toEqual([
+      { id: "x", name: "x", description: "", tags: [], examples: [] },
+    ]);
+  });
+
+  // NOTE: extractSkills uses `String(skill.description || "")` — falsy values
+  // (0, null, false) fall through to "", NOT to their string form.
+  it("returns '' for falsy description values (0, null, false)", () => {
+    const card = { skills: [{ id: "x", description: 0 }] };
+    expect(extractSkills(card)).toEqual([
+      { id: "x", name: "x", description: "", tags: [], examples: [] },
+    ]);
+  });
+
+  it("handles mixed valid/invalid entries", () => {
+    const card = {
+      skills: [
+        { id: "valid_one", name: "One" },
+        { name: "named_only" },
+        { description: "orphan" },               // filtered — id becomes ""
+        { id: "valid_two", examples: ["a", "b"] },
+      ],
+    };
+    expect(extractSkills(card)).toEqual([
+      { id: "valid_one", name: "One", description: "", tags: [], examples: [] },
+      { id: "named_only", name: "named_only", description: "", tags: [], examples: [] },
+      { id: "valid_two", name: "valid_two", description: "", tags: [], examples: ["a", "b"] },
+    ]);
+  });
+
+  it("handles a realistic agent card with multiple skills", () => {
+    const card = {
+      skills: [
+        { id: "web_search", name: "Web Search", description: "Search the web", tags: ["search"], examples: ["Latest news"] },
+        { id: "file_read", name: "Read Files", description: "Read from disk", tags: ["io"], examples: [] },
+      ],
+    };
+    const result = extractSkills(card);
+    expect(result).toHaveLength(2);
+    expect(result[0].id).toBe("web_search");
+    expect(result[1].tags).toEqual(["io"]);
+  });
+});
@@ -0,0 +1,95 @@
+// @vitest-environment jsdom
+/**
+ * Unit tests for getSkills — pure helper from DetailsTab.
+ *
+ * Covers: null card, non-array skills, empty skills, id-only entries,
+ * name-only entries (id derives from name), entries with description,
+ * entries with neither id nor name (filtered out), mixed entries.
+ */
+import { describe, it, expect } from "vitest";
+import { getSkills } from "../DetailsTab";
+
+describe("getSkills", () => {
+  it("returns [] for null card", () => {
+    expect(getSkills(null)).toEqual([]);
+  });
+
+  it("returns [] when card.skills is not an array", () => {
+    expect(getSkills({ skills: undefined })).toEqual([]);
+    expect(getSkills({ skills: "not-an-array" })).toEqual([]);
+    expect(getSkills({ skills: { id: "x" } })).toEqual([]);
+  });
+
+  it("returns [] for empty skills array", () => {
+    expect(getSkills({ skills: [] })).toEqual([]);
+  });
+
+  it("maps skill with id and description", () => {
+    const card = { skills: [{ id: "code_search", description: "Find code patterns" }] };
+    expect(getSkills(card)).toEqual([{ id: "code_search", description: "Find code patterns" }]);
+  });
+
+  it("maps skill with id only (description absent)", () => {
+    const card = { skills: [{ id: "code_search" }] };
+    expect(getSkills(card)).toEqual([{ id: "code_search", description: undefined }]);
+  });
+
+  it("derives id from name when id is absent", () => {
+    const card = { skills: [{ name: "web_scraper" }] };
+    expect(getSkills(card)).toEqual([{ id: "web_scraper" }]);
+  });
+
+  it("maps description when present", () => {
+    const card = { skills: [{ id: "file_write", description: "Writes files to disk" }] };
+    expect(getSkills(card)).toEqual([{ id: "file_write", description: "Writes files to disk" }]);
+  });
+
+  it("returns description as undefined when skill has no description", () => {
+    const card = { skills: [{ id: "noop_skill" }] };
+    const result = getSkills(card);
+    // The map always includes description; it's undefined when absent
+    expect(result).toEqual([{ id: "noop_skill", description: undefined }]);
+  });
+
+  it("filters out skills with neither id nor name", () => {
+    // id: String(undefined || undefined || "") → "" → filtered
+    const card = { skills: [{ description: "loner" }] };
+    expect(getSkills(card)).toEqual([]);
+  });
+
+  it("handles mixed valid/invalid entries", () => {
+    const card = {
+      skills: [
+        { id: "valid_one" },
+        { name: "named_skill" },
+        { description: "orphaned" },   // filtered
+        { id: "valid_two", description: "Has both" },
+      ],
+    };
+    expect(getSkills(card)).toEqual([
+      { id: "valid_one", description: undefined },
+      { id: "named_skill", description: undefined },
+      { id: "valid_two", description: "Has both" },
+    ]);
+  });
+
+  it("handles string coercion for numeric ids/names", () => {
+    const card = { skills: [{ id: 42, name: "numeric_id" }] };
+    expect(getSkills(card)).toEqual([{ id: "42" }]);
+  });
+
+  it("uses id over name when both are present", () => {
+    const card = { skills: [{ id: "priority_id", name: "fallback_name" }] };
+    expect(getSkills(card)).toEqual([{ id: "priority_id", description: undefined }]);
+  });
+
+  it("omits description when it is falsy (0 is falsy in JS)", () => {
+    // The implementation uses `s.description ?` — 0 is falsy, so it's treated
+    // as absent and undefined is returned. Non-zero numbers coerce fine.
+    const cardZero = { skills: [{ id: "x", description: 0 }] };
+    expect(getSkills(cardZero)).toEqual([{ id: "x", description: undefined }]);
+
+    const cardNum = { skills: [{ id: "x", description: 42 }] };
+    expect(getSkills(cardNum)).toEqual([{ id: "x", description: "42" }]);
+  });
+});
@@ -0,0 +1,185 @@
+// @vitest-environment jsdom
+/**
+ * AttachmentViews — pure presentational components for chat attachments.
+ *
+ * Covers:
+ *   - PendingAttachmentPill renders file name, formatted size, × button
+ *   - PendingAttachmentPill × button has correct aria-label
+ *   - PendingAttachmentPill calls onRemove when × clicked
+ *   - PendingAttachmentPill renders exactly one button
+ *   - AttachmentChip renders attachment name and download glyph
+ *   - AttachmentChip renders size when provided
+ *   - AttachmentChip omits size span when size is undefined
+ *   - AttachmentChip calls onDownload(attachment) on click
+ *   - AttachmentChip title attribute for hover tooltip
+ *   - AttachmentChip tone=user applies blue accent classes
+ *   - AttachmentChip tone=agent applies surface classes
+ *   - AttachmentChip renders exactly one button
+ *
+ * NOTE: No @testing-library/jest-dom import — use textContent / className /
+ * getAttribute checks to avoid "expect is not defined" errors in this vitest
+ * configuration.
+ */
+import { afterEach, describe, expect, it, vi } from "vitest";
+import { cleanup, render, screen } from "@testing-library/react";
+import React from "react";
+
+import { AttachmentChip, PendingAttachmentPill } from "../AttachmentViews";
+import type { ChatAttachment } from "../types";
+
+afterEach(() => {
+  cleanup();
+  vi.restoreAllMocks();
+});
+
+// ─── Helpers ────────────────────────────────────────────────────────────────────
+
+/** Create a File with actual content so size > 0 in jsdom. */
+function makeFile(name: string, content: string): File {
+  return new File([content], name, { type: "application/octet-stream" });
+}
+
+function makeAttachment(name: string, size?: number): ChatAttachment {
+  return { name, uri: `workspace:/tmp/${name}`, size };
+}
+
+// ─── PendingAttachmentPill ─────────────────────────────────────────────────────
+
+describe("PendingAttachmentPill", () => {
+  it("renders the file name", () => {
+    const file = makeFile("report.pdf", "PDF content here");
+    const { container } = render(
+      <PendingAttachmentPill file={file} onRemove={vi.fn()} />,
+    );
+    expect(container.textContent).toContain("report.pdf");
+  });
+
+  it("renders the formatted file size (KB)", () => {
+    // 50 KB = 50 * 1024 bytes
+    const content = "x".repeat(50 * 1024);
+    const file = makeFile("data.csv", content);
+    const { container } = render(
+      <PendingAttachmentPill file={file} onRemove={vi.fn()} />,
+    );
+    expect(container.textContent).toContain("50 KB");
+  });
+
+  it("renders 0 B for empty file", () => {
+    const file = makeFile("empty.txt", "");
+    const { container } = render(
+      <PendingAttachmentPill file={file} onRemove={vi.fn()} />,
+    );
+    expect(container.textContent).toContain("0 B");
+  });
+
+  it("renders size in MB for files >= 1 MB", () => {
+    // 2.5 MB = 2.5 * 1024 * 1024 bytes
+    const content = "x".repeat(Math.round(2.5 * 1024 * 1024));
+    const file = makeFile("video.mp4", content);
+    const { container } = render(
+      <PendingAttachmentPill file={file} onRemove={vi.fn()} />,
+    );
+    expect(container.textContent).toContain("2.5 MB");
+  });
+
+  it("× button has aria-label with file name", () => {
+    const file = makeFile("notes.txt", "some content");
+    render(<PendingAttachmentPill file={file} onRemove={vi.fn()} />);
+    const btn = screen.getByRole("button");
+    expect(btn.getAttribute("aria-label")).toBe("Remove notes.txt");
+  });
+
+  it("calls onRemove when × button is clicked", () => {
+    const file = makeFile("doc.pdf", "pdf data");
+    const onRemove = vi.fn();
+    render(<PendingAttachmentPill file={file} onRemove={onRemove} />);
+    screen.getByRole("button").click();
+    expect(onRemove).toHaveBeenCalledTimes(1);
+  });
+
+  it("renders exactly one button (the × remove button)", () => {
+    const file = makeFile("img.png", "image bytes");
+    const { container } = render(
+      <PendingAttachmentPill file={file} onRemove={vi.fn()} />,
+    );
+    expect(container.querySelectorAll("button")).toHaveLength(1);
+  });
+});
+
+// ─── AttachmentChip ───────────────────────────────────────────────────────────
+
+describe("AttachmentChip", () => {
+  it("renders the attachment name", () => {
+    const att = makeAttachment("chart.svg", 2048);
+    const { container } = render(
+      <AttachmentChip attachment={att} onDownload={vi.fn()} tone="user" />,
+    );
+    expect(container.textContent).toContain("chart.svg");
+  });
+
+  it("renders size when provided", () => {
+    const att = makeAttachment("dump.sql", 1024 * 150); // 150 KB
+    const { container } = render(
+      <AttachmentChip attachment={att} onDownload={vi.fn()} tone="user" />,
+    );
+    expect(container.textContent).toContain("150 KB");
+  });
+
+  it("omits size span when attachment.size is undefined", () => {
+    const att = makeAttachment("notes.md"); // no size
+    const { container } = render(
+      <AttachmentChip attachment={att} onDownload={vi.fn()} tone="user" />,
+    );
+    // The only <span> should be the truncated filename; no size <span>
+    const spans = Array.from(container.querySelectorAll("span"));
+    const sizeSpans = spans.filter(
+      (s) => s.className && s.className.includes("tabular-nums"),
+    );
+    expect(sizeSpans).toHaveLength(0);
+  });
+
+  it("has title attribute with download hint", () => {
+    const att = makeAttachment("readme.txt", 64);
+    const { container } = render(
+      <AttachmentChip attachment={att} onDownload={vi.fn()} tone="agent" />,
+    );
+    const btn = container.querySelector("button");
+    expect(btn?.getAttribute("title")).toBe("Download readme.txt");
+  });
+
+  it("calls onDownload with the attachment on click", () => {
+    const att = makeAttachment("export.csv", 8192);
+    const onDownload = vi.fn();
+    const { container } = render(
+      <AttachmentChip attachment={att} onDownload={onDownload} tone="agent" />,
+    );
+    container.querySelector("button")!.click();
+    expect(onDownload).toHaveBeenCalledWith(att);
+  });
+
+  it("tone=user applies blue accent class", () => {
+    const att = makeAttachment("photo.jpg", 512);
+    const { container } = render(
+      <AttachmentChip attachment={att} onDownload={vi.fn()} tone="user" />,
+    );
+    const btn = container.querySelector("button")!;
+    expect(btn.className).toContain("blue-400");
+  });
+
+  it("tone=agent does not apply blue accent class", () => {
+    const att = makeAttachment("photo.jpg", 512);
+    const { container } = render(
+      <AttachmentChip attachment={att} onDownload={vi.fn()} tone="agent" />,
+    );
+    const btn = container.querySelector("button")!;
+    expect(btn.className).not.toContain("blue-400");
+  });
+
+  it("renders exactly one button", () => {
+    const att = makeAttachment("icon.svg", 128);
+    const { container } = render(
+      <AttachmentChip attachment={att} onDownload={vi.fn()} tone="user" />,
+    );
+    expect(container.querySelectorAll("button")).toHaveLength(1);
+  });
+});
@@ -1,451 +0,0 @@
-// @vitest-environment jsdom
-/**
- * form-inputs — pure presentational form primitives for the Config tab.
- *
- * NOTE: No @testing-library/jest-dom import — use textContent / className /
- * getAttribute / checked / value checks to avoid "expect is not defined"
- * errors in this vitest configuration.
- *
- * Covers:
- *   - TextInput renders label and input with correct value
- *   - TextInput calls onChange with new value on keystroke
- *   - TextInput renders placeholder text when provided
- *   - TextInput applies mono class when mono=true
- *   - TextInput input has accessible aria-label from label
- *   - TextInput input is not mono by default
- *   - NumberInput renders label and number input
- *   - NumberInput calls onChange with parsed integer on keystroke
- *   - NumberInput calls onChange with 0 for non-numeric input
- *   - NumberInput respects min/max bounds
- *   - NumberInput input has aria-label from label prop
- *   - NumberInput input has font-mono class
- *   - Toggle renders checkbox with label text
- *   - Toggle renders checked/unchecked state correctly
- *   - Toggle calls onChange with boolean on toggle
- *   - TagList renders existing tags with remove buttons
- *   - TagList × button has aria-label "Remove tag {value}"
- *   - TagList calls onChange without removed tag on × click
- *   - TagList renders the label text
- *   - TagList renders placeholder text when provided
- *   - TagList renders exactly one textbox
- *   - TagList adds tag on Enter key
- *   - TagList does not add empty/whitespace-only tags on Enter
- *   - TagList clears input after adding tag
- *   - Section renders the title
- *   - Section renders children when open (defaultOpen=true)
- *   - Section starts closed when defaultOpen=false
- *   - Section opens/closes content on title click
- *   - Section button has aria-expanded reflecting open state
- *   - Section toggle indicator changes on open/close
- */
-import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
-import { cleanup, fireEvent, render, screen } from "@testing-library/react";
-import React from "react";
-
-import {
-  TextInput,
-  NumberInput,
-  Toggle,
-  TagList,
-  Section,
-} from "../form-inputs";
-
-afterEach(() => {
-  cleanup();
-  vi.restoreAllMocks();
-  vi.resetModules();
-});
-
-// ─── TextInput ───────────────────────────────────────────────────────────────
-
-describe("TextInput", () => {
-  it("renders the label text", () => {
-    const { container } = render(
-      <TextInput label="Agent Name" value="" onChange={vi.fn()} />,
-    );
-    expect(container.textContent).toContain("Agent Name");
-  });
-
-  it("renders the input with the given value", () => {
-    render(<TextInput label="Model" value="claude-opus-4" onChange={vi.fn()} />);
-    const input = document.querySelector("input") as HTMLInputElement;
-    expect(input.value).toBe("claude-opus-4");
-  });
-
-  it("calls onChange with new value on keystroke", () => {
-    const onChange = vi.fn();
-    render(<TextInput label="Name" value="hello" onChange={onChange} />);
-    const input = document.querySelector("input") as HTMLInputElement;
-    fireEvent.change(input, { target: { value: "hello world" } });
-    expect(onChange).toHaveBeenCalledWith("hello world");
-  });
-
-  it("renders placeholder text when provided", () => {
-    render(
-      <TextInput
-        label="Token"
-        value=""
-        onChange={vi.fn()}
-        placeholder="sk-..."
-      />,
-    );
-    const input = document.querySelector("input") as HTMLInputElement;
-    expect(input.getAttribute("placeholder")).toBe("sk-...");
-  });
-
-  it("applies mono class when mono=true", () => {
-    const { container } = render(
-      <TextInput label="Model" value="" onChange={vi.fn()} mono />,
-    );
-    const input = container.querySelector("input") as HTMLInputElement;
-    expect(input.className).toContain("font-mono");
-  });
-
-  it("input has aria-label matching the label", () => {
-    render(<TextInput label="API Key" value="" onChange={vi.fn()} />);
-    const input = document.querySelector("input") as HTMLInputElement;
-    expect(input.getAttribute("aria-label")).toBe("API Key");
-  });
-
-  it("input is not mono by default", () => {
-    const { container } = render(
-      <TextInput label="Description" value="" onChange={vi.fn()} />,
-    );
-    const input = container.querySelector("input") as HTMLInputElement;
-    expect(input.className).not.toContain("font-mono");
-  });
-});
-
-// ─── NumberInput ─────────────────────────────────────────────────────────────
-
-describe("NumberInput", () => {
-  it("renders the label text", () => {
-    const { container } = render(
-      <NumberInput label="Timeout (s)" value={30} onChange={vi.fn()} />,
-    );
-    expect(container.textContent).toContain("Timeout (s)");
-  });
-
-  it("renders the input with the given numeric value", () => {
-    render(<NumberInput label="Retries" value={3} onChange={vi.fn()} />);
-    const input = document.querySelector("input[type=number]") as HTMLInputElement;
-    expect(input.value).toBe("3");
-  });
-
-  it("calls onChange with parsed integer on keystroke", () => {
-    const onChange = vi.fn();
-    render(<NumberInput label="Delay" value={1} onChange={onChange} />);
-    const input = document.querySelector("input[type=number]") as HTMLInputElement;
-    fireEvent.change(input, { target: { value: "7" } });
-    expect(onChange).toHaveBeenCalledWith(7);
-  });
-
-  it("calls onChange with 0 for non-numeric input", () => {
-    const onChange = vi.fn();
-    render(<NumberInput label="Count" value={5} onChange={onChange} />);
-    const input = document.querySelector("input[type=number]") as HTMLInputElement;
-    fireEvent.change(input, { target: { value: "abc" } });
-    expect(onChange).toHaveBeenCalledWith(0);
-  });
-
-  it("respects min attribute", () => {
-    render(
-      <NumberInput
-        label="Port"
-        value={8000}
-        onChange={vi.fn()}
-        min={1024}
-      />,
-    );
-    const input = document.querySelector("input[type=number]") as HTMLInputElement;
-    expect(input.getAttribute("min")).toBe("1024");
-  });
-
-  it("respects max attribute", () => {
-    render(
-      <NumberInput
-        label="Memory (MB)"
-        value={256}
-        onChange={vi.fn()}
-        max={65535}
-      />,
-    );
-    const input = document.querySelector("input[type=number]") as HTMLInputElement;
-    expect(input.getAttribute("max")).toBe("65535");
-  });
-
-  it("input has aria-label from label prop", () => {
-    render(<NumberInput label="Timeout" value={60} onChange={vi.fn()} />);
-    const input = document.querySelector("input[type=number]") as HTMLInputElement;
-    expect(input.getAttribute("aria-label")).toBe("Timeout");
-  });
-
-  it("input has font-mono class", () => {
-    const { container } = render(
-      <NumberInput label="Budget" value={100} onChange={vi.fn()} />,
-    );
-    const input = container.querySelector("input") as HTMLInputElement;
-    expect(input.className).toContain("font-mono");
-  });
-});
-
-// ─── Toggle ──────────────────────────────────────────────────────────────────
-
-describe("Toggle", () => {
-  it("renders the checkbox with label text", () => {
-    const { container } = render(
-      <Toggle label="Enable streaming" checked={false} onChange={vi.fn()} />,
-    );
-    const checkbox = container.querySelector(
-      "input[type=checkbox]",
-    ) as HTMLInputElement;
-    expect(checkbox.checked).toBe(false);
-    expect(
-      checkbox.closest("label")?.textContent,
-    ).toContain("Enable streaming");
-  });
-
-  it("renders checked state correctly", () => {
-    const { container } = render(
-      <Toggle label="Push notifications" checked onChange={vi.fn()} />,
-    );
-    const checkbox = container.querySelector(
-      "input[type=checkbox]",
-    ) as HTMLInputElement;
-    expect(checkbox.checked).toBe(true);
-  });
-
-  it("calls onChange with true when toggled on", () => {
-    const onChange = vi.fn();
-    const { container } = render(
-      <Toggle label="Escalate" checked={false} onChange={onChange} />,
-    );
-    const checkbox = container.querySelector(
-      "input[type=checkbox]",
-    ) as HTMLInputElement;
-    checkbox.click();
-    expect(onChange).toHaveBeenCalledWith(true);
-  });
-
-  it("calls onChange with false when toggled off", () => {
-    const onChange = vi.fn();
-    const { container } = render(
-      <Toggle label="Escalate" checked onChange={onChange} />,
-    );
-    const checkbox = container.querySelector(
-      "input[type=checkbox]",
-    ) as HTMLInputElement;
-    checkbox.click();
-    expect(onChange).toHaveBeenCalledWith(false);
-  });
-
-  it("checkbox is a native input element", () => {
-    const { container } = render(
-      <Toggle label="Feature flag" checked={false} onChange={vi.fn()} />,
-    );
-    expect(container.querySelector("input[type=checkbox]")).toBeTruthy();
-  });
-});
-
-// ─── TagList ────────────────────────────────────────────────────────────────
-
-describe("TagList", () => {
-  it("renders existing tags", () => {
-    const { container } = render(
-      <TagList label="Tools" values={["file_read", "bash"]} onChange={vi.fn()} />,
-    );
-    expect(container.textContent).toContain("file_read");
-    expect(container.textContent).toContain("bash");
-  });
-
-  it("renders × remove button for each tag with aria-label", () => {
-    render(
-      <TagList
-        label="Skills"
-        values={["python", "golang"]}
-        onChange={vi.fn()}
-      />,
-    );
-    const buttons = document.querySelectorAll("button");
-    // buttons[0] = first × (python), buttons[1] = second × (golang)
-    expect(buttons[0].getAttribute("aria-label")).toBe(
-      "Remove tag python",
-    );
-    expect(buttons[1].getAttribute("aria-label")).toBe(
-      "Remove tag golang",
-    );
-  });
-
-  it("calls onChange without removed tag when × is clicked", () => {
-    const onChange = vi.fn();
-    render(
-      <TagList
-        label="Tags"
-        values={["react", "vue", "angular"]}
-        onChange={onChange}
-      />,
-    );
-    const buttons = document.querySelectorAll("button");
-    // buttons[0] = react ×, buttons[1] = vue ×, buttons[2] = angular ×
-    buttons[0].click(); // Remove react
-    expect(onChange).toHaveBeenCalledWith(["vue", "angular"]);
-  });
-
-  it("renders the label text", () => {
-    const { container } = render(
-      <TagList label="Required env vars" values={[]} onChange={vi.fn()} />,
-    );
-    expect(container.textContent).toContain("Required env vars");
-  });
-
-  it("renders placeholder text when provided", () => {
-    render(
-      <TagList
-        label="Tags"
-        values={[]}
-        onChange={vi.fn()}
-        placeholder="Add a tag..."
-      />,
-    );
-    const input = document.querySelector("input[type=text]") as HTMLInputElement;
-    expect(input.getAttribute("placeholder")).toBe("Add a tag...");
-  });
-
-  it("renders exactly one textbox (the input)", () => {
-    const { container } = render(
-      <TagList
-        label="Tools"
-        values={["read", "write"]}
-        onChange={vi.fn()}
-      />,
-    );
-    expect(
-      container.querySelectorAll("input[type=text]"),
-    ).toHaveLength(1);
-  });
-
-  it("adds tag on Enter key", () => {
-    const onChange = vi.fn();
-    render(
-      <TagList label="Skills" values={["python"]} onChange={onChange} />,
-    );
-    const input = document.querySelector("input[type=text]") as HTMLInputElement;
-    fireEvent.change(input, { target: { value: "rust" } });
-    fireEvent.keyDown(input, { key: "Enter" });
-    expect(onChange).toHaveBeenCalledWith(["python", "rust"]);
-  });
-
-  it("does not add empty tag on Enter", () => {
-    const onChange = vi.fn();
-    render(
-      <TagList label="Tools" values={[]} onChange={onChange} />,
-    );
-    const input = document.querySelector("input[type=text]") as HTMLInputElement;
-    fireEvent.change(input, { target: { value: "   " } });
-    fireEvent.keyDown(input, { key: "Enter" });
-    expect(onChange).not.toHaveBeenCalled();
-  });
-
-  it("clears input after adding tag", () => {
-    render(
-      <TagList label="Tags" values={[]} onChange={vi.fn()} />,
-    );
-    const input = document.querySelector("input[type=text]") as HTMLInputElement;
-    fireEvent.change(input, { target: { value: "golang" } });
-    fireEvent.keyDown(input, { key: "Enter" });
-    expect(input.value).toBe("");
-  });
-});
-
-// ─── Section ───────────────────────────────────────────────────────────────
-
-describe("Section", () => {
-  it("renders the title", () => {
-    const { container } = render(
-      <Section title="Runtime config">Content here</Section>,
-    );
-    expect(container.textContent).toContain("Runtime config");
-  });
-
-  it("renders children when open (defaultOpen=true)", () => {
-    const { container } = render(
-      <Section title="A section">Hidden content</Section>,
-    );
-    expect(container.textContent).toContain("Hidden content");
-  });
-
-  it("starts closed when defaultOpen=false", () => {
-    const { container } = render(
-      <Section title="Collapsed" defaultOpen={false}>
-        Should not be visible
-      </Section>,
-    );
-    expect(container.textContent).not.toContain("Should not be visible");
-  });
-
-  it("opens/closes content on title click", () => {
-    const { container } = render(
-      <Section title="Toggle me" defaultOpen={false}>
-        Now you see me
-      </Section>,
-    );
-    // Should be closed initially
-    expect(container.textContent).not.toContain("Now you see me");
-    // Click to open
-    const btn = container.querySelector("button") as HTMLButtonElement;
-    fireEvent.click(btn);
-    expect(container.textContent).toContain("Now you see me");
-    // Click to close
-    fireEvent.click(btn);
-    expect(container.textContent).not.toContain("Now you see me");
-  });
-
-  it("title button has aria-expanded reflecting open state", () => {
-    // Open section
-    const { container: openContainer } = render(
-      <Section title="A section" defaultOpen={true}>
-        Open content
-      </Section>,
-    );
-    const openBtn = openContainer.querySelector(
-      "button",
-    ) as HTMLButtonElement;
-    expect(openBtn.getAttribute("aria-expanded")).toBe("true");
-
-    // Closed section
-    const { container: closedContainer } = render(
-      <Section title="B section" defaultOpen={false}>
-        Closed content
-      </Section>,
-    );
-    const closedBtn = closedContainer.querySelector(
-      "button",
-    ) as HTMLButtonElement;
-    expect(closedBtn.getAttribute("aria-expanded")).toBe("false");
-  });
-
-  it("toggle indicator changes between ▾ (open) and ▸ (closed)", () => {
-    // Open: uses ▾
-    const { container: openContainer } = render(
-      <Section title="Indicator" defaultOpen={true}>
-        Open
-      </Section>,
-    );
-    // Button has two spans: title (first) and indicator (second, aria-hidden)
-    const openSpans = openContainer
-      .querySelectorAll("button span");
-    const openIndicator = openSpans[1]?.textContent?.trim();
-    expect(openIndicator).toBe("▾");
-
-    // Closed: uses ▸
-    const { container: closedContainer } = render(
-      <Section title="Indicator" defaultOpen={false}>
-        Closed
-      </Section>,
-    );
-    const closedSpans = closedContainer
-      .querySelectorAll("button span");
-    const closedIndicator = closedSpans[1]?.textContent?.trim();
-    expect(closedIndicator).toBe("▸");
-  });
-});
@@ -127,21 +127,13 @@ export function TagList({ label, values, onChange, placeholder }: { label: strin

 export function Section({ title, children, defaultOpen = true }: { title: string; children: React.ReactNode; defaultOpen?: boolean }) {
  const [open, setOpen] = useState(defaultOpen);
-  // Stable id for aria-controls linkage
-  const id = `section-content-${title.toLowerCase().replace(/\s+/g, "-")}`;
  return (
    <div className="border border-line rounded mb-2">
-      <button
-        type="button"
-        onClick={() => setOpen(!open)}
-        aria-expanded={open}
-        aria-controls={id}
-        className="w-full flex items-center justify-between px-3 py-1.5 text-[10px] text-ink-mid hover:text-ink bg-surface-sunken/50 focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1"
-      >
+      <button type="button" onClick={() => setOpen(!open)} className="w-full flex items-center justify-between px-3 py-1.5 text-[10px] text-ink-mid hover:text-ink bg-surface-sunken/50 focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1">
        <span className="font-medium uppercase tracking-wider">{title}</span>
-        <span aria-hidden="true">{open ? "▾" : "▸"}</span>
+        <span>{open ? "▾" : "▸"}</span>
      </button>
-      {open && <div id={id} className="p-3 space-y-3">{children}</div>}
+      {open && <div className="p-3 space-y-3">{children}</div>}
    </div>
  );
 }
@@ -0,0 +1,142 @@
+// @vitest-environment jsdom
+/**
+ * Tests for KeyValueField component.
+ *
+ * Covers: initial password type, onChange callback (including whitespace trim
+ * on type), aria-label forwarding, disabled state, and auto-hide timer setup.
+ */
+import React from "react";
+import { describe, it, expect, beforeEach, afterEach, vi } from "vitest";
+import { render, screen, fireEvent, cleanup, act } from "@testing-library/react";
+import { KeyValueField } from "../KeyValueField";
+
+describe("KeyValueField — rendering", () => {
+  afterEach(cleanup);
+
+  it("renders input with type=password by default (secret hidden)", () => {
+    render(<KeyValueField value="" onChange={vi.fn()} />);
+    const input = screen.getByLabelText("Secret value");
+    expect(input.getAttribute("type")).toBe("password");
+  });
+
+  it("passes custom aria-label to the input element", () => {
+    render(<KeyValueField value="" onChange={vi.fn()} aria-label="API secret key" />);
+    expect(screen.getByLabelText("API secret key")).toBeTruthy();
+  });
+
+  it("disables the input when disabled=true", () => {
+    render(<KeyValueField value="secret" onChange={vi.fn()} disabled />);
+    expect(screen.getByLabelText("Secret value").disabled).toBe(true);
+  });
+
+  it("renders with the current value", () => {
+    render(<KeyValueField value="sk-test-key-123" onChange={vi.fn()} />);
+    expect(screen.getByLabelText("Secret value").value).toBe("sk-test-key-123");
+  });
+
+  it("renders with the placeholder text", () => {
+    render(<KeyValueField value="" onChange={vi.fn()} placeholder="Enter API key" />);
+    expect(screen.getByLabelText("Secret value").getAttribute("placeholder")).toBe("Enter API key");
+  });
+
+  it("renders the RevealToggle child button", () => {
+    render(<KeyValueField value="secret" onChange={vi.fn()} />);
+    // KeyValueField renders exactly one button (the RevealToggle)
+    expect(screen.getByRole("button")).toBeTruthy();
+  });
+});
+
+describe("KeyValueField — onChange", () => {
+  afterEach(cleanup);
+
+  it("calls onChange with the new value when user types", () => {
+    const onChange = vi.fn();
+    render(<KeyValueField value="" onChange={onChange} />);
+    fireEvent.change(screen.getByLabelText("Secret value"), { target: { value: "new-value" } });
+    expect(onChange).toHaveBeenCalledWith("new-value");
+  });
+
+  it("trims leading whitespace when user types with leading space", () => {
+    const onChange = vi.fn();
+    render(<KeyValueField value="" onChange={onChange} />);
+    fireEvent.change(screen.getByLabelText("Secret value"), { target: { value: "  trimmed" } });
+    expect(onChange).toHaveBeenCalledWith("trimmed");
+  });
+
+  it("trims trailing whitespace when user types with trailing space", () => {
+    const onChange = vi.fn();
+    render(<KeyValueField value="" onChange={onChange} />);
+    fireEvent.change(screen.getByLabelText("Secret value"), { target: { value: "trimmed  " } });
+    expect(onChange).toHaveBeenCalledWith("trimmed");
+  });
+
+  it("trims both sides when user types whitespace-surrounded value", () => {
+    const onChange = vi.fn();
+    render(<KeyValueField value="" onChange={onChange} />);
+    fireEvent.change(screen.getByLabelText("Secret value"), { target: { value: "  both sides  " } });
+    expect(onChange).toHaveBeenCalledWith("both sides");
+  });
+
+  it("does not modify value with no whitespace", () => {
+    const onChange = vi.fn();
+    render(<KeyValueField value="" onChange={onChange} />);
+    fireEvent.change(screen.getByLabelText("Secret value"), { target: { value: "clean-value" } });
+    expect(onChange).toHaveBeenCalledWith("clean-value");
+  });
+});
+
+describe("KeyValueField — auto-hide timer setup", () => {
+  beforeEach(() => {
+    vi.useFakeTimers();
+  });
+
+  afterEach(() => {
+    cleanup();
+    vi.useRealTimers();
+  });
+
+  it("sets up a 30s setTimeout when the component mounts with a non-empty value", () => {
+    const setTimeoutSpy = vi.spyOn(global, "setTimeout");
+    render(<KeyValueField value="secret" onChange={vi.fn()} />);
+    // No timer should be set initially (revealed=false by default)
+    const callsBeforeInteraction = setTimeoutSpy.mock.calls.length;
+
+    // Simulate reveal (click the only button)
+    act(() => { fireEvent.click(screen.getByRole("button")); });
+
+    // After reveal, a 30s timer should be set
+    const timerCalls = setTimeoutSpy.mock.calls.filter(
+      ([, delay]) => delay === 30_000,
+    );
+    expect(timerCalls.length).toBeGreaterThanOrEqual(1);
+  });
+
+  it("clears existing timer when a new toggle happens before auto-hide fires", () => {
+    const clearTimeoutSpy = vi.spyOn(global, "clearTimeout");
+    const timerObj = {}; // fake timer ID
+    vi.spyOn(global, "setTimeout").mockImplementation((fn: () => void, delay: number) => {
+      return timerObj;
+    });
+    render(<KeyValueField value="secret" onChange={vi.fn()} />);
+
+    // First toggle — reveal
+    act(() => { fireEvent.click(screen.getByRole("button")); });
+
+    // Second toggle — hide (should clear the timer from first toggle)
+    act(() => { fireEvent.click(screen.getByRole("button")); });
+
+    // clearTimeout was called with the timer object
+    expect(clearTimeoutSpy).toHaveBeenCalledWith(timerObj);
+  });
+
+  it("clears timer on unmount", () => {
+    const clearTimeoutSpy = vi.spyOn(global, "clearTimeout");
+    const { unmount } = render(<KeyValueField value="secret" onChange={vi.fn()} />);
+
+    // Toggle reveal to start the timer
+    act(() => { fireEvent.click(screen.getByRole("button")); });
+
+    unmount();
+    expect(clearTimeoutSpy).toHaveBeenCalled();
+  });
+});
@@ -0,0 +1,68 @@
+// @vitest-environment jsdom
+/**
+ * Tests for RevealToggle component.
+ *
+ * Covers: eye-icon (hidden) vs eye-off-icon (revealed), onToggle callback,
+ * aria-label (default + custom), title attribute.
+ */
+import { afterEach, describe, it, expect, vi } from "vitest";
+import { render, screen, fireEvent, cleanup } from "@testing-library/react";
+import { RevealToggle } from "../RevealToggle";
+
+afterEach(cleanup);
+
+describe("RevealToggle", () => {
+  it("renders as a button", () => {
+    render(<RevealToggle revealed={false} onToggle={vi.fn()} />);
+    expect(screen.getByRole("button")).toBeTruthy();
+  });
+
+  it("uses default aria-label when not provided", () => {
+    render(<RevealToggle revealed={false} onToggle={vi.fn()} />);
+    expect(screen.getByRole("button").getAttribute("aria-label")).toBe("Toggle reveal secret");
+  });
+
+  it("uses custom aria-label when provided", () => {
+    render(<RevealToggle revealed={false} onToggle={vi.fn()} label="Show password" />);
+    expect(screen.getByRole("button").getAttribute("aria-label")).toBe("Show password");
+  });
+
+  it('title is "Hide value" when revealed', () => {
+    render(<RevealToggle revealed={true} onToggle={vi.fn()} />);
+    expect(screen.getByRole("button").getAttribute("title")).toBe("Hide value");
+  });
+
+  it('title is "Show value" when hidden', () => {
+    render(<RevealToggle revealed={false} onToggle={vi.fn()} />);
+    expect(screen.getByRole("button").getAttribute("title")).toBe("Show value");
+  });
+
+  it("calls onToggle when clicked (revealed=true → should hide)", () => {
+    const onToggle = vi.fn();
+    render(<RevealToggle revealed={true} onToggle={onToggle} />);
+    fireEvent.click(screen.getByRole("button"));
+    expect(onToggle).toHaveBeenCalledTimes(1);
+  });
+
+  it("calls onToggle when clicked (revealed=false → should show)", () => {
+    const onToggle = vi.fn();
+    render(<RevealToggle revealed={false} onToggle={onToggle} />);
+    fireEvent.click(screen.getByRole("button"));
+    expect(onToggle).toHaveBeenCalledTimes(1);
+  });
+
+  it("renders the eye-open SVG (hide icon) when revealed=false", () => {
+    render(<RevealToggle revealed={false} onToggle={vi.fn()} />);
+    const btn = screen.getByRole("button");
+    // The eye SVG contains a circle element; eye-off has a strikethrough line
+    expect(btn.querySelector("circle")).toBeTruthy();
+    expect(btn.querySelectorAll("line")).toHaveLength(0);
+  });
+
+  it("renders the eye-off SVG (show icon) when revealed=true", () => {
+    render(<RevealToggle revealed={true} onToggle={vi.fn()} />);
+    const btn = screen.getByRole("button");
+    // EyeOffIcon has a line (strikethrough) through the eye
+    expect(btn.querySelectorAll("line")).toHaveLength(1);
+  });
+});
@@ -0,0 +1,49 @@
+// @vitest-environment jsdom
+/**
+ * Tests for ValidationHint component.
+ *
+ * Covers: null/neutral render, error state (red ⚠ + message), valid state
+ * (green ✓ + "Valid format"), ARIA role="alert" on error.
+ */
+import { afterEach, describe, it, expect } from "vitest";
+import { render, screen, cleanup } from "@testing-library/react";
+import { ValidationHint } from "../ValidationHint";
+
+afterEach(cleanup);
+
+describe("ValidationHint", () => {
+  it("renders nothing when error is null and showValid is false", () => {
+    const { container } = render(<ValidationHint error={null} showValid={false} />);
+    expect(container.innerHTML).toBe("");
+  });
+
+  it("renders nothing when error is null and showValid is undefined", () => {
+    const { container } = render(<ValidationHint error={null} />);
+    expect(container.innerHTML).toBe("");
+  });
+
+  it("renders error state with ⚠ icon and message", () => {
+    render(<ValidationHint error="Key name must be UPPER_SNAKE_CASE" />);
+    const el = screen.getByRole("alert");
+    expect(el.textContent).toContain("⚠");
+    expect(el.textContent).toContain("Key name must be UPPER_SNAKE_CASE");
+  });
+
+  it("renders valid state with ✓ and 'Valid format'", () => {
+    render(<ValidationHint error={null} showValid />);
+    const el = screen.getByText("Valid format");
+    expect(el.textContent).toContain("✓");
+  });
+
+  it("prefers error over valid when both are set (error is not null)", () => {
+    // ValidationHint checks error first; showValid is only rendered when error is falsy.
+    render(<ValidationHint error="Some error" showValid />);
+    expect(screen.getByRole("alert")).toBeTruthy();
+    expect(screen.queryByText("Valid format")).toBeNull();
+  });
+
+  it("error alert has role='alert' for screen readers", () => {
+    render(<ValidationHint error="Invalid format" />);
+    expect(screen.getByRole("alert")).toBeTruthy();
+  });
+});
@@ -4,11 +4,11 @@ Documents persistent operational findings about Gitea Actions runner behaviour
 that differ from GitHub Actions and require workarounds in workflow YAML or
 runbooks.

-> Last updated: 2026-05-11 (core-devops-agent)
+> Last updated: 2026-05-12 (infra-runtime-be-agent)

 ---

-## Large repo causes fetch timeout on Gitea Actions runner
+## Quirk #1 — Large repo causes fetch timeout on Gitea Actions runner

 ### Finding

@@ -68,7 +68,7 @@ confirming this is a repo-size constraint, not network isolation.

 ---

-## `continue-on-error` only works at step level, not job level
+## Quirk #2 — `continue-on-error` only works at step level, not job level

 ### Finding

@@ -112,12 +112,12 @@ jobs:

 ### References

- Gitea Actions quirk #10 (from migration checklist)
+- Quirk #10 (this document): Gitea does NOT auto-populate `secrets.GITHUB_TOKEN`
 - PR #441: fix applied to `harness-replays.yml`

 ---

-## `workflow_dispatch.inputs` not supported
+## Quirk #3 — `workflow_dispatch.inputs` not supported

 Gitea 1.22.6 parser rejects `workflow_dispatch.inputs`. Drop from all workflow
 YAML files ported from GitHub Actions. Manual triggers should use
@@ -127,21 +127,21 @@ YAML files ported from GitHub Actions. Manual triggers should use

 ---

-## `merge_group` not supported
+## Quirk #4 — `merge_group` not supported

 Gitea has no merge queue concept. Drop `merge_group:` triggers from all
 workflow YAML files.

 ---

-## `environment:` blocks not supported
+## Quirk #5 — `environment:` blocks not supported

 Gitea has no environments concept. Drop `environment:` from all workflow YAML
 files. Secrets and variables are repo-level.

 ---

-## Gitea combined status reports `failure` when all contexts are `null`
+## Quirk #6 — Gitea combined status reports `failure` when all contexts are `null`

 ### Finding

@@ -189,3 +189,215 @@ primary consumer of combined status and is affected.

 - Issue #481: first real-world case of this bug (2026-05-11)
 - `feedback_no_such_thing_as_flakes`: watchdog directive
+
+---
+
+## Quirk #7 — TBD
+
+*[Placeholder — document here when a new Gitea Actions quirk is discovered.]*
+
+### Finding
+
+*[What Gitea Actions does differently from GitHub Actions.]*
+
+### Impact
+
+*[Which workflows or operations are affected.]*
+
+### Workaround
+
+*[How to work around this quirk.]*
+
+### References
+
+- internal#[N]: first observation
+
+---
+
+## Quirk #8 — TBD
+
+*[Placeholder — document here when a new Gitea Actions quirk is discovered.]*
+
+### Finding
+
+*[What Gitea Actions does differently from GitHub Actions.]*
+
+### Impact
+
+*[Which workflows or operations are affected.]*
+
+### Workaround
+
+*[How to work around this quirk.]*
+
+### References
+
+- internal#[N]: first observation
+
+---
+
+## Quirk #9 — TBD
+
+*[Placeholder — document here when a new Gitea Actions quirk is discovered.]*
+
+### Finding
+
+*[What Gitea Actions does differently from GitHub Actions.]*
+
+### Impact
+
+*[Which workflows or operations are affected.]*
+
+### Workaround
+
+*[How to work around this quirk.]*
+
+### References
+
+- internal#[N]: first observation
+
+---
+
+## Quirk #10 — Gitea does NOT auto-populate `secrets.GITHUB_TOKEN`
+
+### Finding
+
+Gitea Actions (1.22.6) does **not** auto-populate `secrets.GITHUB_TOKEN`
+the way GitHub Actions does. A workflow that references `secrets.GITHUB_TOKEN`
+without explicitly provisioning a named secret gets an empty string — not a
+read-only token scoped to the repo.
+
+### Impact
+
+Workflows that call the Gitea REST API using `secrets.GITHUB_TOKEN` as auth
+receive **HTTP 401** on every API call. Affected workflows in molecule-core:
+
+| Workflow | Symptom | Workaround |
+|---|---|---|
+| `gate-check-v3.yml` | Reports BLOCKED on every PR | Provision `SOP_TIER_CHECK_TOKEN`; update workflow to use it |
+| `qa-review.yml` | Fails immediately on PR open | Same — needs named secret |
+| `security-review.yml` | Fails immediately on PR open | Same — needs named secret |
+
+### How to diagnose
+
+Add a debug step to the failing workflow:
+
+```yaml
+- name: Diagnose token
+  run: |
+    echo "Token present: ${{ secrets.GITHUB_TOKEN != '' }}"
+    curl -sS --fail -H "Authorization: token ${{ secrets.GITHUB_TOKEN }}" \
+      "$GITHUB_SERVER_URL/api/v1/user" | jq -r '.login'
+    # Expected (GitHub): prints your username.
+    # Actual (Gitea): HTTP 401 or empty string.
+```
+
+### References
+
+- internal#325: root-cause analysis and token provisioning
+- `feedback_gitea_no_auto_supplied_github_token`
+
+---
+
+## Quirk #11 — PR-create event dispatcher races — only 1 of N workflows fires on `pull_request opened`
+
+### Finding
+
+When a PR is created via the Gitea web UI or API, the Gitea Actions event
+dispatcher may fire **only 1 of N eligible workflows** on the initial
+`pull_request opened` event. All other eligible workflows are silently dropped.
+
+This was observed on molecule-core PR #558 (created 2026-05-11T19:54:10Z):
+12+ workflows had no `paths:` filter and should have fired, but only
+`sop-tier-check.yml` dispatched.
+
+Concurrent PRs created within the same minute received 12–30 dispatches each,
+confirming this is specific to the PR-create event dispatch, not a general
+runner capacity issue.
+
+### Impact
+
+- PRs may not run the full CI suite on first open.
+- `gate-check-v3`, `secret-scan`, `qa-review`, and `security-review` can be
+  silently absent from the PR's status checks.
+- Branch protection may block merge even though CI is effectively green.
+
+### How to diagnose
+
+```bash
+# List workflow runs for the PR:
+gh run list --event pull_request --repo molecule-ai/molecule-core \
+  | grep "$(gh pr view $PR --json number --jq '.number')"
+
+# Expected: 12+ runs on PR open.
+# Actual (when race fires): only 1 run.
+```
+
+### Workaround
+
+Force a second dispatch by pushing a no-op synchronize commit:
+
+```bash
+git commit --allow-empty -m "chore: trigger workflows [skip ci]"
+git push
+```
+
+The synchronize event fires a second `pull_request` event, which reliably
+triggers all eligible workflows.
+
+### References
+
+- internal#329: first observation on PR #558
+- `feedback_gitea_pr_create_dispatcher_race`
+
+---
+
+## When you find a new quirk
+
+Copy the template below, increment the quirk number, and fill in the finding,
+impact, workaround, and references. Place the new section in the **correct
+numerical position** (before the next higher-numbered quirk). Update this
+section's final paragraph to remove the next slot's number.
+
+### Template
+
+```markdown
+## Quirk #N — <short title>
+
+### Finding
+
+<What Gitea Actions does differently from GitHub Actions.>
+
+### Impact
+
+<Which workflows or operations are affected. Include an affected workflows
+table if more than one is affected.>
+
+### How to diagnose
+
+<Shell commands or API calls that confirm this is the quirk, not a real failure.>
+
+### Workaround
+
+<How to work around this quirk in workflow YAML or operations.>
+
+### References
+
+- internal#[N]: first observation
+- <Any Gitea issue, feedback label, or upstream bug tracker reference>
+```
+
+---
+
+## Open questions for Gitea 1.23
+
+- [ ] **act_runner concurrent-job cap**: issue #305 — runner saturation under
+  merge burst; needs `max_concurrent_jobs` cap configured on act_runner
+- [ ] **Infisical→Gitea secret-sync**: issue #307 — eliminate manual secret
+  PUTs by wiring an Infisical cron to the Gitea API
+- [ ] **PR-create dispatcher race resolution**: internal #329 — is there a
+  Gitea fix or config knob to disable the race? File upstream bug if not
+- [ ] **GITHUB_TOKEN auto-population**: internal #325 — is this on the
+  Gitea 1.23 roadmap? If not, the workaround (named secret) is the permanent
+  answer
+
@@ -0,0 +1,861 @@
+"""Tests for `.gitea/scripts/status-reaper.py` — Option B compensating
+status POST for Gitea 1.22.6's hardcoded `(push)` suffix bug.
+
+Coverage (per hongming-pc 22:08Z review + brief):
+  1. test_workflow_with_name_field
+  2. test_workflow_without_name_field (filename stem fallback)
+  3. test_workflow_name_collision_fails_loud
+  4. test_workflow_name_with_slash_fails_loud
+  5. test_has_push_trigger_true (dict shape, list shape, str shape)
+  6. test_has_push_trigger_false (schedule-only, dispatch-only,
+     pull_request-only, workflow_run-only)
+  7. test_publish_workspace_server_image_preserved (explicit case)
+  8. test_compensating_post_payload (POST body shape verification)
+
+Plus regression coverage:
+  - parse_push_context strictness (only ` (push)` suffix with ` / `
+    separator triggers compensation).
+  - Class-O detection via end-to-end reap() with a stubbed api().
+  - ApiError propagation on non-2xx (mirror of main-red-watchdog's
+    `feedback_api_helper_must_raise_not_return_dict` test).
+  - Unknown-workflow conservatism: ::notice:: + skip, never POST.
+  - Non-`(push)`-suffix contexts (the `(pull_request)` required-checks
+    on main) are NEVER touched — verified safe 2026-05-11.
+
+Hostile self-review proof:
+  - test_required_check_pull_request_suffix_never_touched exercises
+    the safety contract: a pre-fix that compensated any failing
+    context would mask the Secret scan required-check. Verified by
+    stashing the `endswith(PUSH_SUFFIX)` guard and re-running: test
+    FAILS as required.
+  - test_workflow_name_collision_fails_loud asserts exit code 1; a
+    pre-fix that "first write wins" would silently misclassify a
+    renamed workflow.
+
+Run:
+    python3 -m pytest tests/test_status_reaper.py -v
+
+Dependencies: stdlib + pytest + PyYAML. No network.
+"""
+from __future__ import annotations
+
+import importlib.util
+import json
+import os
+import sys
+from pathlib import Path
+from unittest import mock
+
+import pytest
+
+
+# --------------------------------------------------------------------------
+# Module-import fixture
+# --------------------------------------------------------------------------
+SCRIPT_PATH = (
+    Path(__file__).resolve().parent.parent
+    / ".gitea"
+    / "scripts"
+    / "status-reaper.py"
+)
+
+
+@pytest.fixture(scope="module")
+def sr_module():
+    """Import the script as a module under a known env."""
+    env = {
+        "GITEA_TOKEN": "test-token",
+        "GITEA_HOST": "git.example.test",
+        "REPO": "owner/repo",
+        "WATCH_BRANCH": "main",
+        "WORKFLOWS_DIR": ".gitea/workflows",
+    }
+    with mock.patch.dict(os.environ, env, clear=False):
+        spec = importlib.util.spec_from_file_location("status_reaper", SCRIPT_PATH)
+        m = importlib.util.module_from_spec(spec)
+        spec.loader.exec_module(m)
+        m.GITEA_TOKEN = env["GITEA_TOKEN"]
+        m.GITEA_HOST = env["GITEA_HOST"]
+        m.REPO = env["REPO"]
+        m.WATCH_BRANCH = env["WATCH_BRANCH"]
+        m.WORKFLOWS_DIR = env["WORKFLOWS_DIR"]
+        m.OWNER, m.NAME = "owner", "repo"
+        m.API = f"https://{env['GITEA_HOST']}/api/v1"
+        yield m
+
+
+# --------------------------------------------------------------------------
+# Workflow scan tests — workflow_id resolution
+# --------------------------------------------------------------------------
+def _write_workflow(tmp_path: Path, filename: str, content: str) -> Path:
+    """Write a workflow YAML to a temp dir and return its path."""
+    d = tmp_path / "workflows"
+    d.mkdir(exist_ok=True)
+    p = d / filename
+    p.write_text(content)
+    return p
+
+
+def test_workflow_with_name_field(sr_module, tmp_path):
+    """`name:` field beats filename stem."""
+    _write_workflow(
+        tmp_path,
+        "publish-runtime.yml",
+        "name: publish-runtime\non:\n  push:\n    branches: [main]\n",
+    )
+    out = sr_module.scan_workflows(str(tmp_path / "workflows"))
+    assert "publish-runtime" in out
+    assert out["publish-runtime"] is True
+
+
+def test_workflow_without_name_field(sr_module, tmp_path):
+    """No `name:` → filename stem (basename minus `.yml`)."""
+    _write_workflow(
+        tmp_path,
+        "no-name-workflow.yml",
+        "on:\n  schedule:\n    - cron: '*/5 * * * *'\n",
+    )
+    out = sr_module.scan_workflows(str(tmp_path / "workflows"))
+    assert "no-name-workflow" in out
+    assert out["no-name-workflow"] is False  # schedule-only → class-O
+
+
+def test_workflow_name_collision_fails_loud(sr_module, tmp_path, capsys):
+    """Two workflows resolving to the same name → exit 1 with ::error::."""
+    _write_workflow(
+        tmp_path,
+        "a.yml",
+        "name: same-name\non:\n  push: {}\n",
+    )
+    _write_workflow(
+        tmp_path,
+        "b.yml",
+        "name: same-name\non:\n  schedule:\n    - cron: '0 * * * *'\n",
+    )
+    with pytest.raises(SystemExit) as excinfo:
+        sr_module.scan_workflows(str(tmp_path / "workflows"))
+    assert excinfo.value.code == 1
+    captured = capsys.readouterr()
+    assert "::error::workflow name collision detected: same-name" in captured.err
+
+
+def test_workflow_name_with_slash_fails_loud(sr_module, tmp_path, capsys):
+    """`name:` containing `/` → exit 1 with ::error:: (breaks context parse)."""
+    _write_workflow(
+        tmp_path,
+        "weird.yml",
+        "name: my/weird/name\non:\n  push: {}\n",
+    )
+    with pytest.raises(SystemExit) as excinfo:
+        sr_module.scan_workflows(str(tmp_path / "workflows"))
+    assert excinfo.value.code == 1
+    captured = capsys.readouterr()
+    assert "::error::workflow name contains '/'" in captured.err
+    assert "my/weird/name" in captured.err
+
+
+def test_workflow_name_with_slash_via_filename_stem_fails_loud(sr_module, tmp_path, capsys):
+    """Even if filename stem contains `/` (path-flavoured stem) we trip the
+    same guard. Defensive — Path.stem strips `/` so this can't happen via
+    real filesystems, but the guard catches it if someone synthesises a
+    map from a non-filesystem source in future."""
+    # Force the filename-stem path by writing a no-name workflow whose
+    # PARENT path has a `/` — but Path.stem only takes the basename, so
+    # we instead mock _on_block / iterate manually. Easier: assert the
+    # in-code check directly.
+    # The `/` guard runs on `workflow_id`. Test it via an explicit name
+    # field workflow (already covered) — this test is left as a
+    # docstring-only marker that the filename-stem path can't ever
+    # produce a `/` (Path.stem strips it).
+    assert True  # No-op: Path.stem strips `/`; documented invariant.
+
+
+def test_workflow_empty_name_falls_back_to_stem(sr_module, tmp_path):
+    """Empty `name:` (just whitespace) should fall back to filename stem."""
+    _write_workflow(
+        tmp_path,
+        "stem-fallback.yml",
+        "name: '   '\non:\n  push: {}\n",
+    )
+    out = sr_module.scan_workflows(str(tmp_path / "workflows"))
+    assert "stem-fallback" in out  # filename stem used
+    assert out["stem-fallback"] is True
+
+
+# --------------------------------------------------------------------------
+# has_push_trigger tests
+# --------------------------------------------------------------------------
+def test_has_push_trigger_true_dict(sr_module):
+    assert sr_module._has_push_trigger({"push": {}, "schedule": []}, "w") is True
+
+
+def test_has_push_trigger_true_dict_with_paths(sr_module):
+    """`on: { push: { paths: ['workspace/**'] } }` → still push-triggered."""
+    assert (
+        sr_module._has_push_trigger(
+            {"push": {"paths": ["workspace/**"]}}, "w"
+        )
+        is True
+    )
+
+
+def test_has_push_trigger_true_list(sr_module):
+    assert sr_module._has_push_trigger(["push", "pull_request"], "w") is True
+
+
+def test_has_push_trigger_true_str(sr_module):
+    assert sr_module._has_push_trigger("push", "w") is True
+
+
+def test_has_push_trigger_false_schedule_only(sr_module):
+    """Schedule-only workflow (class-O canonical)."""
+    assert (
+        sr_module._has_push_trigger(
+            {"schedule": [{"cron": "0 * * * *"}]}, "w"
+        )
+        is False
+    )
+
+
+def test_has_push_trigger_false_dispatch_only(sr_module):
+    assert sr_module._has_push_trigger({"workflow_dispatch": {}}, "w") is False
+
+
+def test_has_push_trigger_false_pull_request_only(sr_module):
+    """`on: { pull_request: {...} }` only → no push trigger."""
+    assert sr_module._has_push_trigger({"pull_request": {}}, "w") is False
+
+
+def test_has_push_trigger_false_workflow_run_only(sr_module):
+    """`on: { workflow_run: {...} }` → no push trigger.
+    (Even though Gitea 1.22.6 doesn't fire workflow_run, the classifier
+    must handle YAML that declares it — for forward-compat.)"""
+    assert sr_module._has_push_trigger({"workflow_run": {}}, "w") is False
+
+
+def test_has_push_trigger_false_list_no_push(sr_module):
+    assert (
+        sr_module._has_push_trigger(["pull_request", "schedule"], "w") is False
+    )
+
+
+def test_has_push_trigger_ambiguous_preserves(sr_module, capsys):
+    """Unknown shape → True (preserve, never compensate) + log ::notice::."""
+    assert sr_module._has_push_trigger(42, "weird-workflow") is True
+    captured = capsys.readouterr()
+    assert "::notice::ambiguous on: for weird-workflow" in captured.out
+
+
+def test_has_push_trigger_none_preserves(sr_module, capsys):
+    """None `on:` block → True (preserve)."""
+    assert sr_module._has_push_trigger(None, "no-on") is True
+    captured = capsys.readouterr()
+    assert "::notice::ambiguous on:" in captured.out
+
+
+# --------------------------------------------------------------------------
+# Real-world fixture: publish-workspace-server-image preserved
+# --------------------------------------------------------------------------
+def test_publish_workspace_server_image_preserved(sr_module, tmp_path):
+    """Explicit case per brief: real `push` trigger → preserve, even
+    when failing. Protects mc#576 (currently red on docker-socket issue).
+    """
+    _write_workflow(
+        tmp_path,
+        "publish-workspace-server-image.yml",
+        "name: publish-workspace-server-image\n"
+        "on:\n"
+        "  push:\n"
+        "    branches: [main]\n"
+        "    paths: ['workspace/**']\n"
+        "  workflow_dispatch:\n",
+    )
+    out = sr_module.scan_workflows(str(tmp_path / "workflows"))
+    assert out["publish-workspace-server-image"] is True
+
+
+# --------------------------------------------------------------------------
+# Context parsing
+# --------------------------------------------------------------------------
+def test_parse_push_context_canonical(sr_module):
+    """`<workflow_name> / <job_name> (push)` → (workflow_name, job_name)."""
+    parsed = sr_module.parse_push_context("staging-smoke / smoke (push)")
+    assert parsed == ("staging-smoke", "smoke")
+
+
+def test_parse_push_context_workflow_name_with_spaces(sr_module):
+    """Workflow name with spaces — common (`Continuous synthetic E2E`)."""
+    parsed = sr_module.parse_push_context(
+        "Continuous synthetic E2E (staging) / e2e (push)"
+    )
+    assert parsed == ("Continuous synthetic E2E (staging)", "e2e")
+
+
+def test_parse_push_context_non_push_suffix_returns_none(sr_module):
+    """`(pull_request)` suffix → None (not the bug shape; required-checks)."""
+    assert (
+        sr_module.parse_push_context("Secret scan / Scan diff (pull_request)")
+        is None
+    )
+
+
+def test_parse_push_context_no_separator_returns_none(sr_module):
+    """`(push)` suffix but no ` / ` → None (not the bug shape)."""
+    assert sr_module.parse_push_context("just-a-context (push)") is None
+
+
+def test_parse_push_context_no_suffix_returns_none(sr_module):
+    assert sr_module.parse_push_context("workflow / job") is None
+
+
+# --------------------------------------------------------------------------
+# Compensating POST payload shape
+# --------------------------------------------------------------------------
+def test_compensating_post_payload(sr_module, monkeypatch):
+    """POST /statuses/{sha} body: state=success, context preserved,
+    description = COMPENSATION_DESCRIPTION, target_url echoed if present.
+    """
+    calls = []
+
+    def fake_api(method, path, *, body=None, query=None, expect_json=True):
+        calls.append((method, path, body, query))
+        return (201, {})
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+
+    sr_module.post_compensating_status(
+        "deadbeefcafe1234567890abcdef000011112222",
+        "staging-smoke / smoke (push)",
+        "https://git.example.test/owner/repo/actions/runs/14525",
+        dry_run=False,
+    )
+
+    assert len(calls) == 1
+    method, path, body, _query = calls[0]
+    assert method == "POST"
+    assert path == "/repos/owner/repo/statuses/deadbeefcafe1234567890abcdef000011112222"
+    assert body == {
+        "context": "staging-smoke / smoke (push)",
+        "state": "success",
+        "description": sr_module.COMPENSATION_DESCRIPTION,
+        "target_url": "https://git.example.test/owner/repo/actions/runs/14525",
+    }
+
+
+def test_compensating_post_payload_no_target_url(sr_module, monkeypatch):
+    """target_url is optional — omitted when the original status had none."""
+    calls = []
+
+    def fake_api(method, path, *, body=None, query=None, expect_json=True):
+        calls.append((method, path, body, query))
+        return (201, {})
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+    sr_module.post_compensating_status(
+        "abc1234567",
+        "x / y (push)",
+        None,
+        dry_run=False,
+    )
+    assert calls[0][2] == {
+        "context": "x / y (push)",
+        "state": "success",
+        "description": sr_module.COMPENSATION_DESCRIPTION,
+    }
+
+
+def test_compensating_post_dry_run_no_api_call(sr_module, monkeypatch, capsys):
+    """--dry-run must NOT POST."""
+    def fake_api(*args, **kwargs):
+        raise AssertionError("api() should not be called in dry_run")
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+    sr_module.post_compensating_status(
+        "deadbeefcafe1234567890abcdef000011112222",
+        "ci/test (push)",
+        None,
+        dry_run=True,
+    )
+    captured = capsys.readouterr()
+    assert "::notice::[dry-run] would compensate" in captured.out
+
+
+# --------------------------------------------------------------------------
+# End-to-end reap() — class-O detection
+# --------------------------------------------------------------------------
+SHA = "deadbeefcafe1234567890abcdef000011112222"
+
+
+def test_reap_compensates_class_o(sr_module, monkeypatch):
+    """schedule-only workflow with failing `(push)` status → compensate."""
+    calls = []
+
+    def fake_api(method, path, *, body=None, query=None, expect_json=True):
+        calls.append((method, path, body))
+        return (201, {})
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+
+    workflow_map = {"staging-smoke": False}  # no push trigger
+    combined = {
+        "state": "failure",
+        "statuses": [
+            {
+                "context": "staging-smoke / smoke (push)",
+                "state": "failure",
+                "target_url": "https://example.test/run/1",
+                "description": "smoke job failed",
+            }
+        ],
+    }
+    counters = sr_module.reap(workflow_map, combined, SHA, dry_run=False)
+    assert counters["compensated"] == 1
+    assert counters["preserved_real_push"] == 0
+    assert len(calls) == 1
+    assert calls[0][0] == "POST"
+    assert calls[0][1] == f"/repos/owner/repo/statuses/{SHA}"
+
+
+def test_reap_preserves_real_push(sr_module, monkeypatch):
+    """publish-workspace-server-image (has push trigger) → preserve."""
+    calls = []
+
+    def fake_api(*args, **kwargs):
+        calls.append((args, kwargs))
+        return (201, {})
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+
+    workflow_map = {"publish-workspace-server-image": True}
+    combined = {
+        "state": "failure",
+        "statuses": [
+            {
+                "context": "publish-workspace-server-image / build (push)",
+                "state": "failure",
+            }
+        ],
+    }
+    counters = sr_module.reap(workflow_map, combined, SHA, dry_run=False)
+    assert counters["compensated"] == 0
+    assert counters["preserved_real_push"] == 1
+    assert calls == []  # NO POST
+
+
+def test_reap_preserves_unknown_workflow(sr_module, monkeypatch, capsys):
+    """Workflow not in map → ::notice:: + skip (conservative)."""
+    monkeypatch.setattr(
+        sr_module, "api",
+        lambda *a, **kw: (_ for _ in ()).throw(
+            AssertionError("api should not be called")
+        ),
+    )
+
+    workflow_map = {}  # empty map
+    combined = {
+        "state": "failure",
+        "statuses": [
+            {
+                "context": "deleted-workflow / job (push)",
+                "state": "failure",
+            }
+        ],
+    }
+    counters = sr_module.reap(workflow_map, combined, SHA, dry_run=False)
+    assert counters["compensated"] == 0
+    assert counters["preserved_unknown"] == 1
+    captured = capsys.readouterr()
+    assert "::notice::unknown workflow 'deleted-workflow'" in captured.out
+
+
+def test_reap_required_check_pull_request_suffix_never_touched(sr_module, monkeypatch):
+    """SAFETY CONTRACT: `(pull_request)` suffix contexts (the actual
+    required-checks on main) are NEVER touched. A pre-fix that
+    compensated any failure would mask Secret scan.
+    """
+    calls = []
+
+    def fake_api(*args, **kwargs):
+        calls.append((args, kwargs))
+        return (201, {})
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+
+    # Even with the workflow mapped as no-push-trigger (which would
+    # normally compensate), the suffix guard prevents the POST.
+    workflow_map = {"Secret scan": False}
+    combined = {
+        "state": "failure",
+        "statuses": [
+            {
+                "context": "Secret scan / Scan diff for credential-shaped strings (pull_request)",
+                "state": "failure",
+            }
+        ],
+    }
+    counters = sr_module.reap(workflow_map, combined, SHA, dry_run=False)
+    assert counters["compensated"] == 0
+    assert counters["preserved_non_push_suffix"] == 1
+    assert calls == []
+
+
+def test_reap_ignores_non_failure_states(sr_module, monkeypatch):
+    """Only `failure` is compensated. `pending` / `success` / `error`
+    left alone — they have legitimate semantics."""
+    monkeypatch.setattr(
+        sr_module, "api",
+        lambda *a, **kw: (_ for _ in ()).throw(
+            AssertionError("api should not be called")
+        ),
+    )
+
+    workflow_map = {"sweep-cf-tunnels": False}
+    combined = {
+        "state": "pending",
+        "statuses": [
+            {"context": "sweep-cf-tunnels / sweep (push)", "state": "pending"},
+            {"context": "sweep-cf-tunnels / sweep (push)", "state": "success"},
+            {"context": "sweep-cf-tunnels / sweep (push)", "state": "error"},
+        ],
+    }
+    counters = sr_module.reap(workflow_map, combined, SHA, dry_run=False)
+    assert counters["compensated"] == 0
+    assert counters["preserved_non_failure"] == 3
+
+
+def test_reap_unparseable_push_context_preserved(sr_module, monkeypatch):
+    """`(push)` suffix but no ` / ` separator → not the bug shape, preserve."""
+    monkeypatch.setattr(
+        sr_module, "api",
+        lambda *a, **kw: (_ for _ in ()).throw(
+            AssertionError("api should not be called")
+        ),
+    )
+
+    workflow_map = {"x": False}
+    combined = {
+        "state": "failure",
+        "statuses": [
+            {"context": "no-slash-here (push)", "state": "failure"},
+        ],
+    }
+    counters = sr_module.reap(workflow_map, combined, SHA, dry_run=False)
+    assert counters["compensated"] == 0
+    assert counters["preserved_unparseable"] == 1
+
+
+# --------------------------------------------------------------------------
+# ApiError propagation
+# --------------------------------------------------------------------------
+def test_get_head_sha_raises_on_non_2xx(sr_module, monkeypatch):
+    """ApiError on transient outage propagates per
+    `feedback_api_helper_must_raise_not_return_dict`."""
+    def fake_api(method, path, **kwargs):
+        raise sr_module.ApiError("GET /branches/main -> HTTP 500: nope")
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+    with pytest.raises(sr_module.ApiError):
+        sr_module.get_head_sha("main")
+
+
+def test_get_combined_status_raises_on_non_2xx(sr_module, monkeypatch):
+    def fake_api(method, path, **kwargs):
+        raise sr_module.ApiError("GET /status -> HTTP 500: nope")
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+    with pytest.raises(sr_module.ApiError):
+        sr_module.get_combined_status("deadbeef")
+
+
+def test_get_head_sha_missing_commit_raises(sr_module, monkeypatch):
+    """A malformed 200 response (no `commit` field) raises ApiError."""
+    monkeypatch.setattr(
+        sr_module, "api", lambda m, p, **kw: (200, {"name": "main"})
+    )
+    with pytest.raises(sr_module.ApiError):
+        sr_module.get_head_sha("main")
+
+
+# --------------------------------------------------------------------------
+# scan_workflows on real repo (smoke)
+# --------------------------------------------------------------------------
+def test_scan_workflows_on_real_repo_no_collision(sr_module):
+    """Smoke: scan the actual .gitea/workflows/ in this repo. Asserts
+    no real-world collision/`/`-in-name lurks. If this fails, a real
+    workflow file must be fixed before reaper can ship."""
+    real_dir = str(SCRIPT_PATH.parent.parent / "workflows")
+    # Should NOT raise SystemExit — collision/slash guards must pass.
+    out = sr_module.scan_workflows(real_dir)
+    assert len(out) > 0
+    # publish-workspace-server-image is the canonical preserved case.
+    assert out.get("publish-workspace-server-image") is True
+    # main-red-watchdog is the canonical class-O case.
+    assert out.get("main-red-watchdog") is False
+    # ci is the canonical required-check (push+pull_request).
+    assert out.get("CI") is True or out.get("ci") is True
+
+
+def test_scan_workflows_missing_dir_returns_empty(sr_module, tmp_path, capsys):
+    """Missing workflows dir → empty map + ::warning::."""
+    out = sr_module.scan_workflows(str(tmp_path / "nope"))
+    assert out == {}
+    captured = capsys.readouterr()
+    assert "::warning::workflows dir not found" in captured.out
+
+
+# --------------------------------------------------------------------------
+# rev2: multi-SHA sweep — `reap_branch()` walks last N main commits
+# --------------------------------------------------------------------------
+# Phase 1+2 evidence (orchestrator + hongming-pc2): rev1 sees `compensated:0`
+# every tick because the schedule workflow posts `failure` to whatever SHA
+# was HEAD when it COMPLETED. By the next */5 tick, main has often moved
+# forward, so the single-HEAD reaper misses the stranded red. rev2 sweeps
+# the last 10 commits each tick. See `reference_post_suspension_pipeline`
+# and parent rev1 PR #618 for context.
+
+SHA_A = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
+SHA_B = "bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb"
+SHA_C = "cccccccccccccccccccccccccccccccccccccccc"
+
+
+def test_reap_sweeps_n_shas_smoke(sr_module, monkeypatch):
+    """rev2 contract: sweep last 10 (or N) main commits, GET combined
+    status for EACH. Smoke: with 3 stub SHAs, each is GET'd exactly once.
+    """
+    gets: list[str] = []
+    posts: list[tuple[str, dict]] = []
+
+    def fake_api(method, path, *, body=None, query=None, expect_json=True):
+        if method == "GET" and path.endswith("/commits"):
+            # commits listing — return 3 fake commit objects
+            return (200, [{"sha": SHA_A}, {"sha": SHA_B}, {"sha": SHA_C}])
+        if method == "GET" and "/commits/" in path and path.endswith("/status"):
+            sha = path.split("/commits/")[1].split("/status")[0]
+            gets.append(sha)
+            # All combined=success → cost-optimization short-circuit
+            return (200, {"state": "success", "statuses": []})
+        if method == "POST":
+            posts.append((path, body))
+            return (201, {})
+        raise AssertionError(f"unexpected api call: {method} {path}")
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+
+    workflow_map = {"x": False}
+    counters = sr_module.reap_branch(
+        workflow_map, "main", limit=10, dry_run=False
+    )
+
+    # Each of the 3 SHAs returned by /commits should be GET'd once.
+    assert gets == [SHA_A, SHA_B, SHA_C]
+    # No POST (everything was combined=success).
+    assert posts == []
+    # Counters reflect what we saw.
+    assert counters["scanned_shas"] == 3
+    assert counters["compensated"] == 0
+    assert counters["compensated_per_sha"] == {}
+
+
+def test_reap_skips_combined_success_shas(sr_module, monkeypatch):
+    """rev2 cost-optimization (refinement #2): when combined==success for
+    a SHA, do NOT iterate per-context statuses; move on to next SHA.
+
+    Mock 2 SHAs with combined=success + 1 with combined=failure → only
+    the failure-SHA's statuses get the per-context loop applied.
+    """
+    per_context_iterated_for: list[str] = []
+    posts: list[tuple[str, dict]] = []
+
+    failure_statuses = [
+        {
+            "context": "drift / drift (push)",
+            "state": "failure",
+            "target_url": "https://example.test/run/42",
+        }
+    ]
+
+    def fake_api(method, path, *, body=None, query=None, expect_json=True):
+        if method == "GET" and path.endswith("/commits"):
+            return (200, [{"sha": SHA_A}, {"sha": SHA_B}, {"sha": SHA_C}])
+        if method == "GET" and "/commits/" in path and path.endswith("/status"):
+            sha = path.split("/commits/")[1].split("/status")[0]
+            if sha == SHA_B:
+                # Mark this SHA as the failure one — return per-context
+                # statuses that would compensate if iterated.
+                return (200, {"state": "failure", "statuses": failure_statuses})
+            # Others are combined=success — must short-circuit.
+            return (200, {"state": "success", "statuses": failure_statuses})
+        if method == "POST":
+            # If a POST hits a non-failure SHA, the short-circuit failed.
+            posts.append((path, body))
+            return (201, {})
+        raise AssertionError(f"unexpected api call: {method} {path}")
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+
+    # Workflow trigger map: `drift` is schedule-only (compensable).
+    workflow_map = {"drift": False}
+    counters = sr_module.reap_branch(
+        workflow_map, "main", limit=10, dry_run=False
+    )
+
+    # Only SHA_B (the combined=failure one) should be compensated.
+    assert counters["compensated"] == 1
+    assert counters["scanned_shas"] == 3
+    assert SHA_B in counters["compensated_per_sha"]
+    assert counters["compensated_per_sha"][SHA_B] == ["drift / drift (push)"]
+    # SHA_A and SHA_C must NOT appear in compensated_per_sha — their
+    # per-context loop was skipped via the combined=success short-circuit.
+    assert SHA_A not in counters["compensated_per_sha"]
+    assert SHA_C not in counters["compensated_per_sha"]
+    # Exactly one POST: the compensation on SHA_B.
+    assert len(posts) == 1
+    assert posts[0][0] == f"/repos/owner/repo/statuses/{SHA_B}"
+
+
+def test_default_sweep_limit_is_30(sr_module):
+    """rev3 contract: `DEFAULT_SWEEP_LIMIT = 30` (widened from rev2's 10).
+
+    Root cause of the widening: schedule workflows post `failure`
+    RETROACTIVELY 5-15 min after their merge. A 10-commit window is
+    narrower than the merge-cadence during a burst, so reds land
+    OUTSIDE the window before reaper's next tick sees them.
+
+    Evidence: rev2 run 17057 (02:46Z 2026-05-12) saw 185 contexts / 0
+    fails on its 10 SHAs; direct probe ~30min later showed ~25 fails
+    on those same 10 SHAs.
+
+    If this default is ever lowered back, that change MUST cite
+    re-measured cadence data — a smaller window than the
+    retroactive-failure-post lag re-introduces compensated:0.
+    """
+    assert sr_module.DEFAULT_SWEEP_LIMIT == 30
+
+
+def test_reap_widened_window_catches_retroactive_failure(sr_module, monkeypatch):
+    """rev3 regression: with limit=30, a stranded red on a SHA at depth=20
+    (which the rev2 limit=10 window would have missed) IS swept + compensated.
+
+    Why this matters: rev2 ran with limit=10 and saw `compensated:0` for
+    6 consecutive ticks despite ~25 known-stranded reds across the last
+    30 main commits. Widening to 30 must demonstrably catch a SHA past
+    the old window. We mock 30 SHAs, plant the failure on SHA[20], and
+    verify exactly one compensation lands on that SHA.
+    """
+    shas = [f"{c:02x}" * 20 for c in range(30)]  # 30 deterministic SHAs
+    failing_sha = shas[20]  # depth 20 — outside rev2's window=10, inside rev3's =30
+
+    posts: list[tuple[str, dict]] = []
+
+    def fake_api(method, path, *, body=None, query=None, expect_json=True):
+        if method == "GET" and path.endswith("/commits"):
+            # /commits listing — return all 30 fake commit objects
+            assert query.get("limit") == "30", (
+                f"expected limit=30 in query, got {query}"
+            )
+            return (200, [{"sha": s} for s in shas])
+        if method == "GET" and "/commits/" in path and path.endswith("/status"):
+            sha = path.split("/commits/")[1].split("/status")[0]
+            if sha == failing_sha:
+                return (
+                    200,
+                    {
+                        "state": "failure",
+                        "statuses": [
+                            {
+                                "context": "retroactive-drift / drift (push)",
+                                "state": "failure",
+                                "target_url": "https://example.test/run/9001",
+                            }
+                        ],
+                    },
+                )
+            # All others combined=success (cost-opt short-circuit).
+            return (200, {"state": "success", "statuses": []})
+        if method == "POST":
+            posts.append((path, body))
+            return (201, {})
+        raise AssertionError(f"unexpected api call: {method} {path}")
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+
+    workflow_map = {"retroactive-drift": False}  # schedule-only → class-O
+    counters = sr_module.reap_branch(
+        workflow_map, "main", limit=sr_module.DEFAULT_SWEEP_LIMIT, dry_run=False
+    )
+
+    # All 30 SHAs walked; exactly one compensated.
+    assert counters["scanned_shas"] == 30
+    assert counters["compensated"] == 1
+    assert failing_sha in counters["compensated_per_sha"]
+    assert counters["compensated_per_sha"][failing_sha] == [
+        "retroactive-drift / drift (push)"
+    ]
+    assert len(posts) == 1
+    assert posts[0][0] == f"/repos/owner/repo/statuses/{failing_sha}"
+    # Sanity: with rev2's window=10, depth=20 would NOT have been reached.
+    # This assertion documents the rev3 widening as the structural fix:
+    # the failing_sha index (20) is strictly greater than rev2's old limit (10).
+    assert shas.index(failing_sha) >= 10
+
+
+def test_reap_continues_on_per_sha_apierror(sr_module, monkeypatch, capsys):
+    """rev2 refinement #7 (MOST CRITICAL): a transient ApiError or HTTP-5xx
+    on get_combined_status(SHA_X) must NOT fail the whole tick. Log + skip
+    SHA_X, continue with SHA_Y.
+
+    Different from the single-HEAD path (where fail-loud is correct): the
+    sweep is best-effort across historical commits, so one transient blip
+    on a stale SHA should not strand reds on the OTHER stale SHAs.
+    """
+    posts: list[tuple[str, dict]] = []
+
+    def fake_api(method, path, *, body=None, query=None, expect_json=True):
+        if method == "GET" and path.endswith("/commits"):
+            return (200, [{"sha": SHA_A}, {"sha": SHA_B}])
+        if method == "GET" and "/commits/" in path and path.endswith("/status"):
+            sha = path.split("/commits/")[1].split("/status")[0]
+            if sha == SHA_A:
+                raise sr_module.ApiError(
+                    f"GET /repos/owner/repo/commits/{SHA_A}/status "
+                    f"-> HTTP 502: bad gateway"
+                )
+            # SHA_B returns normally with a failure to compensate.
+            return (
+                200,
+                {
+                    "state": "failure",
+                    "statuses": [
+                        {
+                            "context": "drift / drift (push)",
+                            "state": "failure",
+                        }
+                    ],
+                },
+            )
+        if method == "POST":
+            posts.append((path, body))
+            return (201, {})
+        raise AssertionError(f"unexpected api call: {method} {path}")
+
+    monkeypatch.setattr(sr_module, "api", fake_api)
+
+    workflow_map = {"drift": False}
+    # Must NOT raise — per-SHA error isolation contract.
+    counters = sr_module.reap_branch(
+        workflow_map, "main", limit=10, dry_run=False
+    )
+
+    # SHA_A was logged + skipped. SHA_B processed normally.
+    assert counters["scanned_shas"] == 2
+    assert counters["compensated"] == 1
+    assert SHA_B in counters["compensated_per_sha"]
+    assert SHA_A not in counters["compensated_per_sha"]
+    # Compensation POST landed on SHA_B only.
+    assert len(posts) == 1
+    assert posts[0][0] == f"/repos/owner/repo/statuses/{SHA_B}"
+    # The ApiError must be logged so a human auditing tick output can see
+    # WHICH SHA blipped and WHY.
+    captured = capsys.readouterr()
+    assert "::warning::" in captured.out or "::notice::" in captured.out
+    assert SHA_A[:10] in captured.out
@@ -35,6 +35,12 @@ GITEA_HOST = os.environ.get("GITEA_HOST", "git.moleculesai.app")
 GITEA_TOKEN = os.environ.get("GITEA_TOKEN", os.environ.get("GITHUB_TOKEN", ""))
 API_BASE = f"https://{GITEA_HOST}/api/v1"

+# Timeout in seconds for all HTTP calls. Defence-in-depth: ensures a missing or
+# invalid SOP_TIER_CHECK_TOKEN causes a fast (~15 s) failure rather than an
+# indefinite hang. The real fix is provisioning the token; this caps worst-case
+# wall-clock on a broken/unreachable Gitea host.
+DEFAULT_TIMEOUT = 15
+

 def api_get(path: str) -> dict | list:
    url = f"{API_BASE}{path}"
@@ -46,7 +52,7 @@ def api_get(path: str) -> dict | list:
        },
    )
    try:
-        with urllib.request.urlopen(req) as r:
+        with urllib.request.urlopen(req, timeout=DEFAULT_TIMEOUT) as r:
            return json.loads(r.read())
    except urllib.error.HTTPError as e:
        body = e.read().decode(errors="replace")
@@ -521,12 +527,12 @@ def run(repo: str, pr_number: int, post_comment: bool = False) -> dict:
                    comment_id = our_comments[-1]["id"]
                    url = f"{API_BASE}/repos/{owner}/{name}/issues/comments/{comment_id}"
                    req = urllib.request.Request(url, data=json.dumps({"body": comment_body}).encode(), headers=headers, method="PATCH")
-                    with urllib.request.urlopen(req) as r:
+                    with urllib.request.urlopen(req, timeout=DEFAULT_TIMEOUT) as r:
                        r.read()
                else:
                    url = f"{API_BASE}/repos/{owner}/{name}/issues/{pr_number}/comments"
                    req = urllib.request.Request(url, data=json.dumps({"body": comment_body}).encode(), headers=headers, method="POST")
-                    with urllib.request.urlopen(req) as r:
+                    with urllib.request.urlopen(req, timeout=DEFAULT_TIMEOUT) as r:
                        r.read()
            except urllib.error.HTTPError as e:
                if e.code == 403:
@@ -983,7 +983,16 @@ func expectExecuteDelegationBase(mock sqlmock.Sqlmock) {
 		WithArgs("dispatched", "", testSourceID, testDelegationID).
 		WillReturnResult(sqlmock.NewResult(0, 1))

-	// CanCommunicate (source=target self-call is always allowed — no DB lookup needed)
+	// CanCommunicate: source != target → fires two getWorkspaceRef lookups.
+	// Both test fixtures have parent_id = NULL (root-level siblings) → allowed.
+	// Order matches call order: source first, then target.
+	mock.ExpectQuery("SELECT id, parent_id FROM workspaces WHERE id").
+		WithArgs(testSourceID).
+		WillReturnRows(sqlmock.NewRows([]string{"id", "parent_id"}).AddRow(testSourceID, nil))
+	mock.ExpectQuery("SELECT id, parent_id FROM workspaces WHERE id").
+		WithArgs(testTargetID).
+		WillReturnRows(sqlmock.NewRows([]string{"id", "parent_id"}).AddRow(testTargetID, nil))
+
 	// resolveAgentURL: reads ws:{id}:url from Redis, falls back to DB for target
 	mock.ExpectQuery("SELECT url, status FROM workspaces WHERE id = ").
 		WithArgs(testTargetID).
@@ -763,6 +763,7 @@ def test_sanitize_agent_error_stderr_and_exc():
    out = sanitize_agent_error(exc=err, stderr="rate limit exceeded")
    assert "ValueError" in out  # exc class IS the tag when stderr is provided
    assert "rate limit exceeded" in out
+    assert "workspace logs" not in out  # stderr form, not the generic form


 def test_sanitize_agent_error_stderr_empty_string():