Merge pull request 'feat(security): RFC#523 3-layer forbidden-env guardrail for tenant workspaces (task #146 )' (#1555 ) from feat/146-forbidden-env-guard into main

Merge pull request 'fix(sop-checklist): widen ack eligibility per RFC#450 Option C (closes internal#442)' (#1554 ) from fix/sop-checklist-widen-ack-internal-442 into main
Merge pull request 'test(e2e): local prod-mimic backend for peer-visibility MCP gate + make e2e-peer-visibility (task #166 )' (#1551 ) from e2e/peer-visibility-local-backend-task166 into main
2026-05-19 01:57:30 +00:00 · 2026-05-19 01:57:08 +00:00 · 2026-05-19 01:57:06 +00:00 · 2026-05-19 01:56:16 +00:00 · 2026-05-19 01:56:12 +00:00 · 2026-05-19 01:56:10 +00:00
147 changed files with 9481 additions and 1646 deletions
@@ -100,11 +100,12 @@ printf 'header = "Authorization: token %s"\n' "$GITEA_TOKEN" > "$CURL_AUTH_FILE"
 # (bash trap 'function' EXIT expands variables at trap-fire time, not def time).
 PR_JSON=$(mktemp)
 REVIEWS_JSON=$(mktemp)
+COMMENTS_JSON=$(mktemp)
 TEAM_PROBE_TMP=$(mktemp)
 NA_STATUSES_TMP=""  # declared here so cleanup() always has the var

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

@@ -206,7 +207,81 @@ CANDIDATES=$(jq -r --arg author "$PR_AUTHOR" --arg head "$PR_HEAD_SHA" "$JQ_FILT
 debug "candidate non-author approvers: $(echo "$CANDIDATES" | tr '\n' ' ')"

 if [ -z "$CANDIDATES" ]; then
-  echo "::error::${TEAM}-review awaiting non-author APPROVE from ${TEAM} team (no candidates yet)"
+  # --- Guardrail (internal#503): explain the most common false
+  # "no candidates" red. Gitea's review event enum is EXACTLY
+  # APPROVED/REQUEST_CHANGES/COMMENT/PENDING. A wrong value ("APPROVE",
+  # lowercase, ...) is silently accepted (HTTP 200) and stored as
+  # state=PENDING. A correctly-started draft review has an EMPTY body;
+  # a NON-empty body + state==PENDING by a non-author == an intended
+  # verdict mis-filed by a wrong event string. Surface it actionably.
+  # This does NOT change the gate result (still fail-closed below) — it
+  # only converts a mystery red into a named, self-fixing error.
+  MISFILED_FILTER='.[]
+    | select(.state == "PENDING")
+    | select(.dismissed != true)
+    | select(.user.login != $author)
+    | select(((.body // "") | gsub("^\\s+|\\s+$";"") | length) > 0)
+    | "\(.id)\t\(.user.login)"'
+  MISFILED=$(jq -r --arg author "$PR_AUTHOR" "$MISFILED_FILTER" "$REVIEWS_JSON" 2>/dev/null || true)
+  if [ -n "$MISFILED" ]; then
+    echo "::error::${TEAM}-review: non-author review(s) were SUBMITTED but stored as PENDING — almost certainly the wrong Gitea review event string (internal#503)."
+    echo "::error::Gitea accepts ONLY the exact enum APPROVED / REQUEST_CHANGES / COMMENT. 'APPROVE' or lowercase is silently (HTTP 200) filed as PENDING and is invisible to this gate."
+    printf '%s\n' "$MISFILED" | while IFS="$(printf '\t')" read -r _rid _rl; do
+      [ -n "${_rid:-}" ] && echo "::error::  review id=${_rid} by '${_rl}': RE-SUBMIT via POST ${API}/repos/${OWNER}/${NAME}/pulls/${PR_NUMBER}/reviews with {\"event\":\"APPROVED\"} (correct enum) — do NOT edit the DB."
+    done
+  fi
+
+  # --- Fallback (internal#348): check issue comments for agent-approval ---
+  # core-qa-agent and core-security-agent approve via issue comments, NOT
+  # the reviews API. The reviews API returns zero entries for comment-only
+  # approvals. This fallback reads PR issue comments and extracts logins that:
+  #   1. Posted a comment matching the agent-prefix pattern for this gate:
+  #        qa      → "[core-qa-agent] APPROVED"
+  #        security → "[core-security-agent] APPROVED"
+  #      OR posted a generic approval keyword (word-anchored, case-insensitive):
+  #        APPROVED / LGTM / ACCEPTED
+  #   2. Are not the PR author
+  #   3. The team-membership probe below is the authoritative filter.
+  AGENT_PATTERN=""
+  case "$TEAM" in
+    qa)       AGENT_PATTERN="\\[core-qa-agent\\]" ;;
+    security) AGENT_PATTERN="\\[core-security-agent\\]" ;;
+  esac
+  HTTP_CODE=$(curl -sS -o "$COMMENTS_JSON" -w '%{http_code}' \
+    -K "$CURL_AUTH_FILE" "${API}/repos/${OWNER}/${NAME}/issues/${PR_NUMBER}/comments")
+  debug "GET /issues/${PR_NUMBER}/comments → HTTP ${HTTP_CODE}"
+  if [ "$HTTP_CODE" = "200" ]; then
+    # JQ expression: select non-author comments that match either the
+    # agent-prefix pattern (case-insensitive) OR a generic approval keyword.
+    JQ_APPROVALS='
+      .[] |
+      select(.user.login != $author) |
+      . as $cmt |
+      if ($agent_pattern | length) > 0 and ($cmt.body // "" | test($agent_pattern; "i")) then
+        $cmt.user.login
+      elif ($cmt.body // "" | test("\\b(APPROVED|LGTM|ACCEPTED)\\b"; "i")) then
+        $cmt.user.login
+      else
+        empty
+      end
+    '
+    CANDIDATES=$(jq -r \
+      --arg author "$PR_AUTHOR" \
+      --arg agent_pattern "$AGENT_PATTERN" \
+      "$JQ_APPROVALS" \
+      "$COMMENTS_JSON" 2>/dev/null | sort -u)
+    debug "comment-based approval candidates: $(echo "$CANDIDATES" | tr '\n' ' ')"
+
+    if [ -n "$CANDIDATES" ]; then
+      echo "::notice::${TEAM}-review: reviews API found no APPROVED reviews; found $(echo "$CANDIDATES" | wc -w | xargs) comment-based approval candidate(s) — verifying team membership..."
+    fi
+  else
+    debug "could not fetch issue comments (HTTP ${HTTP_CODE})"
+  fi
+fi
+
+if [ -z "${CANDIDATES:-}" ]; then
+  echo "::error::${TEAM}-review awaiting non-author APPROVE from ${TEAM} team (no candidates from reviews API or issue comments)"
  exit 1
 fi

@@ -268,6 +268,7 @@ def compute_ack_state(
    items_by_slug: dict[str, dict[str, Any]],
    numeric_aliases: dict[int, str],
    team_membership_probe: "callable[[str, list[str]], list[str]]",
+    high_risk: bool = False,
 ) -> dict[str, dict[str, Any]]:
    """Compute per-item ack state.

@@ -330,11 +331,16 @@ def compute_ack_state(
    for slug, candidates in pending_team_check.items():
        if not candidates:
            continue
-        required = items_by_slug[slug]["required_teams"]
+        # Risk-class-aware required-teams resolution (RFC#450 Option C):
+        # high-risk PRs use `required_teams_high_risk` (when set on the
+        # item); default class uses `required_teams`. The probe closure
+        # is built with the same high_risk flag so the two reads are
+        # always consistent (both sites share `resolve_required_teams`).
+        required = resolve_required_teams(items_by_slug[slug], high_risk)
        approved = team_membership_probe(slug, candidates)  # returns subset
        rejected_not_in_team[slug] = [u for u in candidates if u not in approved]
        ackers_per_slug[slug] = approved
-        # Stash required teams for description rendering.
+        # Stash resolved teams for description rendering.
        items_by_slug[slug]["_required_resolved"] = required

    return {
@@ -765,6 +771,42 @@ def get_tier_mode(pr: dict[str, Any], cfg: dict[str, Any]) -> str:
    return default_mode


+def is_high_risk(pr: dict[str, Any], cfg: dict[str, Any]) -> bool:
+    """Return True when the PR is high-risk per RFC#450 Option C.
+
+    A PR is high-risk when ANY of:
+      - it carries the `tier:high` label (mechanically strictest tier), or
+      - it carries any label listed in cfg.high_risk_labels.
+
+    High-risk PRs use `required_teams_high_risk` (when set on an item)
+    instead of the default `required_teams`. Items without
+    `required_teams_high_risk` are unaffected (the default applies).
+
+    Governance fix for internal#442 — closes the inconsistency between
+    sop-tier-check (tier-aware) and sop-checklist (was tier-blind).
+    """
+    label_set = {(l.get("name") or "") for l in (pr.get("labels") or [])}
+    if "tier:high" in label_set:
+        return True
+    high_risk_labels = set(cfg.get("high_risk_labels") or [])
+    return bool(label_set & high_risk_labels)
+
+
+def resolve_required_teams(item: dict[str, Any], high_risk: bool) -> list[str]:
+    """Pick the active required_teams list for an item.
+
+    When high_risk is True AND the item declares a non-empty
+    `required_teams_high_risk`, return that. Else fall back to
+    `required_teams`. Keeping this in one helper means the gate's
+    decision shape stays single-sited even as items grow.
+    """
+    if high_risk:
+        elevated = item.get("required_teams_high_risk") or []
+        if elevated:
+            return list(elevated)
+    return list(item.get("required_teams") or [])
+
+
 def main(argv: list[str] | None = None) -> int:
    p = argparse.ArgumentParser()
    p.add_argument("--owner", required=True)
@@ -825,6 +867,12 @@ def main(argv: list[str] | None = None) -> int:

    comments = client.get_issue_comments(args.owner, args.repo, args.pr)

+    # High-risk classification (RFC#450 Option C, governance fix for
+    # internal#442). Computed ONCE per PR — used by both the probe
+    # closure and compute_ack_state so the elevation decision is
+    # single-sited.
+    high_risk = is_high_risk(pr, cfg)
+
    # Build team-membership probe closure that caches results per
    # (user, team-id) so a user acking multiple items only triggers
    # one membership lookup per team.
@@ -832,7 +880,7 @@ def main(argv: list[str] | None = None) -> int:

    def probe(slug: str, users: list[str]) -> list[str]:
        item = items_by_slug[slug]
-        team_names: list[str] = item["required_teams"]
+        team_names: list[str] = resolve_required_teams(item, high_risk)
        # Resolve names → ids. NOTE: orgs/{org}/teams/search may not be
        # available — fall back to the list endpoint.
        team_ids: list[int] = []
@@ -877,7 +925,9 @@ def main(argv: list[str] | None = None) -> int:
                    # may still find membership in another team.
        return approved

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

    state, description = render_status(items, ack_state, body_state)
@@ -890,7 +940,10 @@ def main(argv: list[str] | None = None) -> int:
        description = f"[info tier:low] {description}"

    # Diagnostics to job log.
-    print(f"::notice::PR #{args.pr} author={author} head={head_sha[:7]} mode={mode}")
+    print(
+        f"::notice::PR #{args.pr} author={author} head={head_sha[:7]} "
+        f"mode={mode} risk_class={'high' if high_risk else 'default'}"
+    )
    for it in items:
        slug = it["slug"]
        ackers = ack_state[slug]["ackers"]
@@ -17,6 +17,9 @@ Scenarios:
  T8_team_not_member          — team membership → 404 (not a member) → exit 1
  T9_team_403                — team membership → 403 (token not in team) → exit 1
  T14_non_default_base        — open PR targeting staging → script exits 0 (no-op)
+  T15_comments_agent_approval — reviews empty; comments have "[core-qa-agent] APPROVED" → exit 0
+  T16_comments_generic_approval — reviews empty; comments have "APPROVED" by team member → exit 0
+  T17_comments_no_approval   — reviews empty; comments have no approval keywords → exit 1

 Usage:
  FIXTURE_STATE_DIR=/tmp/x python3 _review_check_fixture.py 8080
@@ -97,7 +100,9 @@ class Handler(http.server.BaseHTTPRequestHandler):
        # GET /repos/{owner}/{name}/pulls/{pr_number}/reviews
        m = re.match(r"^/api/v1/repos/([^/]+)/([^/]+)/pulls/(\d+)/reviews$", path)
        if m:
-            if sc in ("T4_reviews_empty", "T5_reviews_only_author"):
+            if sc in ("T4_reviews_empty", "T5_reviews_only_author",
+                      "T15_comments_agent_approval", "T16_comments_generic_approval",
+                      "T17_comments_no_approval"):
                return self._json(200, [])
            if sc == "T6_reviews_dismissed":
                return self._json(200, [{
@@ -116,6 +121,28 @@ class Handler(http.server.BaseHTTPRequestHandler):
                {"state": "APPROVED", "dismissed": False, "user": {"login": "core-devops"}, "commit_id": "abc1234"},
            ])

+        # GET /repos/{owner}/{name}/issues/{pr_number}/comments
+        m = re.match(r"^/api/v1/repos/([^/]+)/([^/]+)/issues/(\d+)/comments$", path)
+        if m:
+            if sc == "T15_comments_agent_approval":
+                return self._json(200, [
+                    {"user": {"login": "core-qa-agent"}, "body": "[core-qa-agent] APPROVED this PR. Good changes.", "id": 1},
+                    {"user": {"login": "alice"}, "body": "I authored this PR", "id": 2},
+                    {"user": {"login": "random-user"}, "body": "Looks okay to me", "id": 3},
+                ])
+            if sc == "T16_comments_generic_approval":
+                return self._json(200, [
+                    {"user": {"login": "core-qa-agent"}, "body": "APPROVED — all acceptance criteria met", "id": 1},
+                    {"user": {"login": "alice"}, "body": "-authored", "id": 2},
+                ])
+            if sc == "T17_comments_no_approval":
+                return self._json(200, [
+                    {"user": {"login": "alice"}, "body": "I authored this PR", "id": 1},
+                    {"user": {"login": "random-user"}, "body": "Looks okay to me", "id": 2},
+                ])
+            # Default scenarios (T1–T9, T14): no comments
+            return self._json(200, [])
+
        # GET /teams/{team_id}/members/{username}
        m = re.match(r"^/api/v1/teams/(\d+)/members/([^/]+)$", path)
        if m:
@@ -127,6 +154,12 @@ class Handler(http.server.BaseHTTPRequestHandler):
            # T7_team_member: member
            return self._empty(204)

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

    def do_POST(self):
@@ -334,6 +334,31 @@ assert_contains "T12 jq: core-devops (non-author APPROVED) in candidates" "core-
 assert_eq "T12 jq: alice (author) NOT in candidates" "" "$(echo "$T12_CANDIDATES" | grep '^alice$' || true)"
 assert_eq "T12 jq: carol (dismissed) NOT in candidates" "" "$(echo "$T12_CANDIDATES" | grep '^carol$' || true)"

+# T15 — comment-based approval via agent prefix pattern → exit 0
+echo
+echo "== T15 comment agent-prefix approval =="
+T15_OUT=$(run_review_check "T15_comments_agent_approval")
+T15_RC=$(cat "$FIX_STATE_DIR/last_rc")
+assert_eq "T15 exit code 0 (agent-comment approval + team member)" "0" "$T15_RC"
+assert_contains "T15 comment fallback notice" "comment-based approval" "$T15_OUT"
+assert_contains "T15 core-qa-agent APPROVED" "APPROVED by core-qa-agent" "$T15_OUT"
+
+# T16 — comment-based approval via generic APPROVED keyword → exit 0
+echo
+echo "== T16 comment generic keyword approval =="
+T16_OUT=$(run_review_check "T16_comments_generic_approval")
+T16_RC=$(cat "$FIX_STATE_DIR/last_rc")
+assert_eq "T16 exit code 0 (generic-approval comment + team member)" "0" "$T16_RC"
+assert_contains "T16 comment fallback notice" "comment-based approval" "$T16_OUT"
+
+# T17 — no approval keywords in comments → exit 1
+echo
+echo "== T17 comments with no approval keywords =="
+T17_OUT=$(run_review_check "T17_comments_no_approval")
+T17_RC=$(cat "$FIX_STATE_DIR/last_rc")
+assert_eq "T17 exit code 1 (no candidates from comments)" "1" "$T17_RC"
+assert_contains "T17 no candidates error" "no candidates from reviews API or issue comments" "$T17_OUT"
+
 echo
 echo "------"
 echo "PASS=$PASS FAIL=$FAIL"
@@ -602,4 +602,216 @@ class TestComputeNaState(unittest.TestCase):
        self.assertEqual(len(na_directives), 1)
        self.assertEqual(na_directives[0][0], "sop-n/a")
        self.assertEqual(na_directives[0][1], "qa-review")
-        self.assertIn("no surface", na_directives[0][2])
+
+
+# ---------------------------------------------------------------------------
+# RFC#450 Option C — risk-classed two-eyes (governance fix for internal#442)
+# ---------------------------------------------------------------------------
+
+
+class TestIsHighRisk(unittest.TestCase):
+    """The high-risk predicate decides which required_teams list applies.
+
+    Predicate: tier:high label OR any label in cfg.high_risk_labels.
+    """
+
+    def setUp(self):
+        self.cfg = sop.load_config(CONFIG_PATH)
+
+    def test_no_labels_is_default_class(self):
+        pr = {"labels": []}
+        self.assertFalse(sop.is_high_risk(pr, self.cfg))
+
+    def test_tier_high_is_high_risk(self):
+        pr = {"labels": [{"name": "tier:high"}]}
+        self.assertTrue(sop.is_high_risk(pr, self.cfg))
+
+    def test_tier_low_is_default_class(self):
+        pr = {"labels": [{"name": "tier:low"}]}
+        self.assertFalse(sop.is_high_risk(pr, self.cfg))
+
+    def test_tier_medium_is_default_class(self):
+        # tier:medium alone is NOT high-risk (Option C — medium routes
+        # to the wider engineers OR-set).
+        pr = {"labels": [{"name": "tier:medium"}]}
+        self.assertFalse(sop.is_high_risk(pr, self.cfg))
+
+    def test_area_security_label_is_high_risk(self):
+        pr = {"labels": [{"name": "tier:medium"}, {"name": "area:security"}]}
+        self.assertTrue(sop.is_high_risk(pr, self.cfg))
+
+    def test_area_schema_label_is_high_risk(self):
+        pr = {"labels": [{"name": "area:schema"}]}
+        self.assertTrue(sop.is_high_risk(pr, self.cfg))
+
+    def test_area_identity_label_is_high_risk(self):
+        pr = {"labels": [{"name": "area:identity"}]}
+        self.assertTrue(sop.is_high_risk(pr, self.cfg))
+
+    def test_area_fleet_image_label_is_high_risk(self):
+        pr = {"labels": [{"name": "area:fleet-image"}]}
+        self.assertTrue(sop.is_high_risk(pr, self.cfg))
+
+    def test_area_gate_meta_label_is_high_risk(self):
+        # Gate-meta = changes to sop-checklist/sop-tier-check itself.
+        pr = {"labels": [{"name": "area:gate-meta"}]}
+        self.assertTrue(sop.is_high_risk(pr, self.cfg))
+
+    def test_unknown_area_label_is_default_class(self):
+        pr = {"labels": [{"name": "area:docs"}]}
+        self.assertFalse(sop.is_high_risk(pr, self.cfg))
+
+
+class TestResolveRequiredTeams(unittest.TestCase):
+    """The team resolver picks the elevated list only for high-risk PRs
+    AND only when the item declares one — items without an elevated
+    list always use the default required_teams."""
+
+    def test_default_class_uses_default_teams(self):
+        item = {"required_teams": ["engineers", "managers", "ceo"], "required_teams_high_risk": ["ceo"]}
+        self.assertEqual(
+            sop.resolve_required_teams(item, high_risk=False),
+            ["engineers", "managers", "ceo"],
+        )
+
+    def test_high_risk_uses_elevated_teams(self):
+        item = {"required_teams": ["engineers", "managers", "ceo"], "required_teams_high_risk": ["ceo"]}
+        self.assertEqual(
+            sop.resolve_required_teams(item, high_risk=True),
+            ["ceo"],
+        )
+
+    def test_high_risk_without_elevated_falls_back_to_default(self):
+        # Items that don't declare required_teams_high_risk (e.g.
+        # comprehensive-testing, staging-smoke) are unaffected by risk-class.
+        item = {"required_teams": ["engineers"]}
+        self.assertEqual(
+            sop.resolve_required_teams(item, high_risk=True),
+            ["engineers"],
+        )
+
+    def test_empty_elevated_list_falls_back_to_default(self):
+        # A defensive case: required_teams_high_risk: [] should not
+        # silently lock out all approvers — fall back to the default
+        # so the gate stays satisfiable. (Tightening should remove the
+        # key, not set it to empty.)
+        item = {"required_teams": ["engineers"], "required_teams_high_risk": []}
+        self.assertEqual(
+            sop.resolve_required_teams(item, high_risk=True),
+            ["engineers"],
+        )
+
+
+class TestRootCauseAckEligibilityWidened(unittest.TestCase):
+    """Closes internal#442: a non-author engineers-team ack now satisfies
+    root-cause / no-backwards-compat for the default class.
+
+    The dead-managers/ceo-persona-token gridlock is the symptom; the
+    root cause is that sop-checklist ignored tier-class. These tests
+    pin the new wider-default behavior so it can't regress silently.
+    """
+
+    def setUp(self):
+        self.items = _items_by_slug()
+        self.aliases = _numeric_aliases()
+
+    @staticmethod
+    def _approve_only(allowed):
+        return lambda slug, users: [u for u in users if u in allowed]
+
+    def test_engineers_ack_satisfies_root_cause_default_class(self):
+        # Bob is in engineers only (not managers, not ceo). Default class.
+        comments = [_comment("bob", "/sop-ack root-cause")]
+        # Probe: bob is approved because root-cause now lists engineers.
+        probe = self._approve_only({"bob"})
+        state = sop.compute_ack_state(
+            comments, "alice", self.items, self.aliases, probe, high_risk=False
+        )
+        self.assertEqual(state["root-cause"]["ackers"], ["bob"])
+
+    def test_engineers_ack_satisfies_no_backwards_compat_default_class(self):
+        comments = [_comment("bob", "/sop-ack no-backwards-compat")]
+        probe = self._approve_only({"bob"})
+        state = sop.compute_ack_state(
+            comments, "alice", self.items, self.aliases, probe, high_risk=False
+        )
+        self.assertEqual(state["no-backwards-compat"]["ackers"], ["bob"])
+
+    def test_engineers_ack_alone_fails_root_cause_when_high_risk(self):
+        # High-risk PR: only ceo can ack. Engineers-only ack must fail.
+        comments = [_comment("bob", "/sop-ack root-cause")]
+        # Probe: bob is in engineers, not ceo. Under high_risk,
+        # required_teams_high_risk=[ceo] → bob is NOT approved.
+        # Probe receives the items + flag indirectly via main(); for
+        # the unit-test path we inject a probe that rejects bob.
+        probe = self._approve_only(set())  # nobody is in ceo
+        state = sop.compute_ack_state(
+            comments, "alice", self.items, self.aliases, probe, high_risk=True
+        )
+        self.assertEqual(state["root-cause"]["ackers"], [])
+        self.assertIn("bob", state["root-cause"]["rejected"]["not_in_team"])
+
+    def test_ceo_ack_satisfies_root_cause_when_high_risk(self):
+        # High-risk PR + ceo-team approver → passes (the senior path).
+        comments = [_comment("hongming", "/sop-ack root-cause")]
+        probe = self._approve_only({"hongming"})
+        state = sop.compute_ack_state(
+            comments, "alice", self.items, self.aliases, probe, high_risk=True
+        )
+        self.assertEqual(state["root-cause"]["ackers"], ["hongming"])
+
+    def test_self_ack_still_forbidden_even_with_widened_eligibility(self):
+        # Author cannot self-ack — widening teams must NOT weaken
+        # the non-author rule.
+        comments = [_comment("alice", "/sop-ack root-cause")]
+        probe = self._approve_only({"alice"})
+        state = sop.compute_ack_state(
+            comments, "alice", self.items, self.aliases, probe, high_risk=False
+        )
+        self.assertEqual(state["root-cause"]["ackers"], [])
+        self.assertIn("alice", state["root-cause"]["rejected"]["self_ack"])
+
+
+class TestHighRiskClassUsesElevatedListInConfig(unittest.TestCase):
+    """End-to-end: the shipped config + RFC#450 predicate must keep
+    root-cause / no-backwards-compat gated on ceo for high-risk PRs."""
+
+    def test_root_cause_high_risk_elevated_to_ceo_only(self):
+        items = _items_by_slug()
+        # tier:high alone makes the PR high-risk → root-cause needs ceo.
+        self.assertEqual(
+            sop.resolve_required_teams(items["root-cause"], high_risk=True),
+            ["ceo"],
+        )
+        # Default class accepts engineers/managers/ceo.
+        self.assertEqual(
+            sorted(sop.resolve_required_teams(items["root-cause"], high_risk=False)),
+            sorted(["engineers", "managers", "ceo"]),
+        )
+
+    def test_no_backwards_compat_high_risk_elevated_to_ceo_only(self):
+        items = _items_by_slug()
+        self.assertEqual(
+            sop.resolve_required_teams(items["no-backwards-compat"], high_risk=True),
+            ["ceo"],
+        )
+        self.assertEqual(
+            sorted(sop.resolve_required_teams(items["no-backwards-compat"], high_risk=False)),
+            sorted(["engineers", "managers", "ceo"]),
+        )
+
+    def test_other_items_unchanged_by_risk_class(self):
+        # Items without required_teams_high_risk are unaffected.
+        items = _items_by_slug()
+        for slug in (
+            "comprehensive-testing",
+            "local-postgres-e2e",
+            "staging-smoke",
+            "five-axis-review",
+            "memory-consulted",
+        ):
+            self.assertEqual(
+                sop.resolve_required_teams(items[slug], high_risk=False),
+                sop.resolve_required_teams(items[slug], high_risk=True),
+                f"item {slug} should not be affected by risk-class",
+            )
@@ -50,6 +50,34 @@ tier_failure_mode:
  "tier:low": soft
 default_mode: hard  # used when no tier:* label is present

+# High-risk class (RFC#450 Option C, governance-fix for internal#442).
+#
+# A PR is "high-risk" when ANY of the listed labels are applied OR when
+# the PR has `tier:high` (mechanically the strictest existing tier).
+# High-risk items use `required_teams_high_risk` (when present on the
+# item); non-high-risk items use the default `required_teams`.
+#
+# This closes the inconsistency that the SOP charter already mandates
+# `tier:high → ceo only` for the sibling `sop-tier-check` gate; the
+# sop-checklist's `root-cause` and `no-backwards-compat` items now
+# follow the same risk-classed two-eyes shape:
+#   - Default class (tier:low/medium, not high-risk): a non-author
+#     engineers/managers/ceo ack satisfies the item — 25+ live
+#     identities, no dependency on a dead/inactive senior persona
+#     token.
+#   - High-risk class (tier:high OR any high_risk_label): still
+#     requires a non-author ceo ack (durable human team).
+#
+# Tightening: add labels to high_risk_labels.
+# Loosening: remove labels.
+high_risk_labels:
+  - "risk:high"
+  - "area:security"
+  - "area:schema"
+  - "area:fleet-image"
+  - "area:identity"
+  - "area:gate-meta"
+
 items:
  - slug: comprehensive-testing
    numeric_alias: 1
@@ -78,11 +106,15 @@ items:
  - slug: root-cause
    numeric_alias: 4
    pr_section_marker: "Root-cause not symptom"
-    required_teams: [managers, ceo]
+    required_teams: [engineers, managers, ceo]
+    required_teams_high_risk: [ceo]
    description: >-
-      One-sentence root-cause statement. Ack from managers tier
-      (team-leads) or ceo. Senior judgment required to attest
-      root-cause-versus-symptom.
+      One-sentence root-cause statement. Default class: non-author
+      engineers/managers/ceo ack suffices (engineers can attest
+      root-cause-vs-symptom for routine fixes). High-risk class
+      (see `high_risk_labels`): non-author ceo ack required —
+      senior judgment for irreversible/security/identity/gate
+      changes. Closes internal#442 + tracks RFC#450.

  - slug: five-axis-review
    numeric_alias: 5
@@ -95,10 +127,14 @@ items:
  - slug: no-backwards-compat
    numeric_alias: 6
    pr_section_marker: "No backwards-compat shim / dead code added"
-    required_teams: [managers, ceo]
+    required_teams: [engineers, managers, ceo]
+    required_teams_high_risk: [ceo]
    description: >-
-      Yes/no + justification if no. Senior ack required because
-      backward-compat shims are how dead-code accretes.
+      Yes/no + justification if no. Default class: non-author
+      engineers/managers/ceo ack suffices. High-risk class
+      (see `high_risk_labels`): non-author ceo ack required —
+      senior judgment for shim-versus-real-fix on irreversible
+      surfaces. Closes internal#442 + tracks RFC#450.

  - slug: memory-consulted
    numeric_alias: 7
@@ -158,8 +158,68 @@ jobs:
            echo "NOTE: No warning in output (may be suppressed by log level)"
          fi

+      - name: Reproduce openclaw failure — pipe held OPEN, no EOF
+        run: |
+          set -euo pipefail
+          echo "=== keep-stdin-open pipe (the real openclaw / Claude Code case) ==="
+          echo ""
+          echo "Before the readline() fix this HANGS: main() did"
+          echo "  stdin.read(65536)  -> on a pipe, blocks until 64KB OR EOF."
+          echo "An MCP client sends one ~150B initialize and keeps stdin"
+          echo "open waiting for the response, so the server never parsed"
+          echo "the request and the client timed out (openclaw: 'MCP error"
+          echo "-32000: Connection closed'). The earlier regular-file /"
+          echo "heredoc-pipe steps PASSED through this bug because a file"
+          echo "(or a closing heredoc) yields EOF immediately."
+          echo ""
+
+          # Drive the server through a real pipe that stays OPEN: write
+          # one initialize, do NOT close stdin, and require a response
+          # within a hard timeout. read(65536) -> no output -> timeout
+          # kills it -> FAIL. readline() -> immediate response -> PASS.
+          python - <<'PYEOF'
+          import json, subprocess, sys, time, select
+
+          proc = subprocess.Popen(
+              [sys.executable, "a2a_mcp_server.py"],
+              stdin=subprocess.PIPE, stdout=subprocess.PIPE,
+              stderr=subprocess.STDOUT,
+              env={**__import__("os").environ},
+          )
+          req = json.dumps({
+              "jsonrpc": "2.0", "id": 1, "method": "initialize",
+              "params": {"protocolVersion": "2024-11-05",
+                         "capabilities": {},
+                         "clientInfo": {"name": "keepopen", "version": "1"}},
+          }) + "\n"
+          proc.stdin.write(req.encode())
+          proc.stdin.flush()
+          # Deliberately DO NOT close proc.stdin — mirror a live MCP client.
+
+          deadline = time.time() + 15
+          line = b""
+          while time.time() < deadline:
+              r, _, _ = select.select([proc.stdout], [], [], 1)
+              if r:
+                  line = proc.stdout.readline()
+                  if line:
+                      break
+          proc.kill()
+
+          if not line:
+              print("FAIL: no response within 15s on an open pipe — "
+                    "stdin.read(65536) regression is back")
+              sys.exit(1)
+          resp = json.loads(line.decode())
+          assert resp.get("id") == 1 and "result" in resp, \
+              f"unexpected response: {line[:200]!r}"
+          assert resp["result"]["serverInfo"]["name"] == "molecule", \
+              f"wrong serverInfo: {line[:200]!r}"
+          print("PASS: server answered initialize on a still-open pipe")
+          PYEOF
+
      - name: Run unit tests for stdio transport
        run: |
          set -euo pipefail
          echo "=== Running stdio transport unit tests ==="
-          python -m pytest tests/test_a2a_mcp_server.py::TestStdioPipeAssertion -v --no-cov
+          python -m pytest tests/test_a2a_mcp_server.py::TestStdioPipeAssertion tests/test_a2a_mcp_server.py::TestStdioKeepOpenPipe -v --no-cov
@@ -538,11 +538,13 @@ jobs:
  all-required:
    # Aggregator sentinel — RFC internal#219 §2 (Phase 4 — closes internal#286).
    #
-    # Single stable required-status name that branch protection points at;
-    # CI churns underneath in `needs:` without any protection edits. Mirrors
-    # the molecule-controlplane Phase 2a impl shipped in CP PR#112 and
-    # referenced by `internal#286` ("Phase 4 is a single small PR... mirrors
-    # CP's existing one").
+    # Emits `CI / all-required (<event>)` where <event> is the workflow trigger
+    # (e.g. `CI / all-required (pull_request)`, `CI / all-required (push)`).
+    # Branch protection MUST be updated to require the event-suffixed name —
+    # requiring `CI / all-required` (bare, no suffix) silently blocks all merges
+    # because Gitea treats absent status contexts as pending (not skipped), and
+    # no workflow emits the bare name. Fixed: BP now requires
+    # `CI / all-required (pull_request)` per issue #1473.
    #
    # Closes the failure mode where status_check_contexts on molecule-core/main
    # only listed `Secret scan` + `sop-tier-check` (the 2 meta-gates), so real
@@ -52,6 +52,30 @@ name: E2E Peer Visibility (literal MCP list_peers)
 #     flip-to-required-ready (mirrors e2e-staging-saas.yml's proven shape;
 #     real EC2-provisioning E2E is push/dispatch/cron only — it is 30+ min
 #     and cannot run per-PR-update).
+#
+# LOCAL BACKEND (added 2026-05-15 — feedback_local_must_mimic_production,
+# feedback_mandatory_local_e2e_before_ship, feedback_local_test_before_
+# staging_e2e)
+# --------------------------------------------------------------------
+# The standing rule is that the local prod-mimic stack runs a MANDATORY
+# local-Postgres E2E BEFORE staging E2E. A staging-only peer-visibility
+# gate caught regressions late + expensively (cold EC2). The
+# `peer-visibility-local` job below runs the SAME byte-identical
+# assertion (tests/e2e/lib/peer_visibility_assert.sh) against the local
+# docker-compose stack — built + booted exactly like e2e-api.yml's
+# proven E2E API Smoke Test job (ephemeral pg/redis ports, go build,
+# background platform-server). It runs on PR + push (local boot is
+# minutes, not the 30+ min cold-EC2 path), so peer-visibility is part of
+# the local gate that fires before the staging E2E.
+#
+# It is its OWN non-required status context `E2E Peer Visibility (local)`
+# — same non-required-by-design decision as the staging job (red until
+# Hermes-401 #162 / OpenClaw-never-online #165 land; flip-to-required
+# tracked at molecule-core#1296). It is an HONEST gate: NO
+# continue-on-error mask (feedback_fix_root_not_symptom). It is kept a
+# distinct context (not folded into e2e-api.yml's required `E2E API
+# Smoke Test`) precisely so a deliberately-RED-today gate cannot wedge
+# the required local-E2E job or any unrelated merge.

 on:
  push:
@@ -65,6 +89,8 @@ on:
      - 'workspace/a2a_mcp_server.py'
      - 'workspace/platform_tools/registry.py'
      - 'tests/e2e/test_peer_visibility_mcp_staging.sh'
+      - 'tests/e2e/test_peer_visibility_mcp_local.sh'
+      - 'tests/e2e/lib/peer_visibility_assert.sh'
      - '.gitea/workflows/e2e-peer-visibility.yml'
  pull_request:
    branches: [main]
@@ -77,6 +103,8 @@ on:
      - 'workspace/a2a_mcp_server.py'
      - 'workspace/platform_tools/registry.py'
      - 'tests/e2e/test_peer_visibility_mcp_staging.sh'
+      - 'tests/e2e/test_peer_visibility_mcp_local.sh'
+      - 'tests/e2e/lib/peer_visibility_assert.sh'
      - '.gitea/workflows/e2e-peer-visibility.yml'
  workflow_dispatch:
  schedule:
@@ -108,16 +136,160 @@ jobs:
    timeout-minutes: 5
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-      - name: Validate driving script
+      - name: Validate driving scripts + shared assertion lib
        run: |
+          bash -n tests/e2e/lib/peer_visibility_assert.sh
+          echo "lib/peer_visibility_assert.sh — bash syntax OK"
          bash -n tests/e2e/test_peer_visibility_mcp_staging.sh
          echo "test_peer_visibility_mcp_staging.sh — bash syntax OK"
-          echo "Real fresh-provision MCP list_peers E2E runs on push to"
+          bash -n tests/e2e/test_peer_visibility_mcp_local.sh
+          echo "test_peer_visibility_mcp_local.sh — bash syntax OK"
+          echo "Staging fresh-provision MCP list_peers E2E runs on push to"
          echo "main / workflow_dispatch / daily cron (30+ min EC2 boot)."
+          echo "The LOCAL backend runs in the peer-visibility-local job"
+          echo "below on this same PR (local docker-compose stack)."

-  # Real gate: provisions a throwaway org + sibling-per-runtime, drives
-  # the LITERAL list_peers MCP call per runtime, asserts 200 + expected
-  # peer set, then scoped teardown. push(main)/dispatch/cron only.
+  # LOCAL gate: same byte-identical assertion against the local prod-mimic
+  # docker-compose stack — the MANDATORY local-E2E that must run BEFORE
+  # the staging E2E (feedback_mandatory_local_e2e_before_ship,
+  # feedback_local_test_before_staging_e2e). Bootstrap mirrors
+  # e2e-api.yml's proven E2E API Smoke Test job (per-run container names +
+  # ephemeral host ports so concurrent host-network act_runner runs don't
+  # collide; go build; background platform-server). Its OWN non-required
+  # status context `E2E Peer Visibility (local)` — non-required-by-design
+  # exactly like the staging job (red until #162/#165 land;
+  # flip-to-required tracked at molecule-core#1296). HONEST gate, NO
+  # continue-on-error mask (feedback_fix_root_not_symptom). Runs on PR +
+  # push (local boot is minutes, not the 30+ min cold-EC2 path).
+  # bp-required: pending #1296
+  peer-visibility-local:
+    name: E2E Peer Visibility (local)
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    env:
+      # Per-run names + ephemeral ports — same collision-avoidance as
+      # e2e-api.yml (host-network act_runner; feedback_act_runner_*).
+      PG_CONTAINER: pg-e2e-pv-${{ github.run_id }}-${{ github.run_attempt }}
+      REDIS_CONTAINER: redis-e2e-pv-${{ github.run_id }}-${{ github.run_attempt }}
+      # LLM keys so hermes/openclaw can actually boot. The local script
+      # SKIPs (not fails) any runtime whose key is absent, so a partially
+      # keyed CI env still exercises whatever it can.
+      CLAUDE_CODE_OAUTH_TOKEN: ${{ secrets.E2E_CLAUDE_CODE_OAUTH_TOKEN }}
+      E2E_MINIMAX_API_KEY: ${{ secrets.MOLECULE_STAGING_MINIMAX_API_KEY }}
+      E2E_ANTHROPIC_API_KEY: ${{ secrets.MOLECULE_STAGING_ANTHROPIC_API_KEY }}
+      E2E_OPENAI_API_KEY: ${{ secrets.MOLECULE_STAGING_OPENAI_API_KEY }}
+      PV_RUNTIMES: "hermes openclaw claude-code"
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - uses: actions/setup-go@40f1582b2485089dde7abd97c1529aa768e1baff # v5
+        with:
+          go-version: 'stable'
+          cache: true
+          cache-dependency-path: workspace-server/go.sum
+      - name: Pre-pull alpine + ensure provisioner network
+        run: |
+          docker pull alpine:latest >/dev/null
+          docker network create molecule-core-net >/dev/null 2>&1 || true
+          echo "alpine:latest pre-pulled; molecule-core-net ensured."
+      - name: Start Postgres (docker, ephemeral port)
+        run: |
+          docker rm -f "$PG_CONTAINER" 2>/dev/null || true
+          docker run -d --name "$PG_CONTAINER" \
+            -e POSTGRES_USER=dev -e POSTGRES_PASSWORD=dev -e POSTGRES_DB=molecule \
+            -p 0:5432 postgres:16 >/dev/null
+          PG_PORT=$(docker port "$PG_CONTAINER" 5432/tcp | awk -F: '/^0\.0\.0\.0:/ {print $2; exit}')
+          [ -n "$PG_PORT" ] || PG_PORT=$(docker port "$PG_CONTAINER" 5432/tcp | head -1 | awk -F: '{print $NF}')
+          if [ -z "$PG_PORT" ]; then
+            echo "::error::Could not resolve host port for $PG_CONTAINER"
+            docker logs "$PG_CONTAINER" || true; exit 1
+          fi
+          echo "DATABASE_URL=postgres://dev:dev@127.0.0.1:${PG_PORT}/molecule?sslmode=disable" >> "$GITHUB_ENV"
+          for i in $(seq 1 30); do
+            docker exec "$PG_CONTAINER" pg_isready -U dev >/dev/null 2>&1 && { echo "Postgres ready after ${i}s"; exit 0; }
+            sleep 1
+          done
+          echo "::error::Postgres did not become ready in 30s"; docker logs "$PG_CONTAINER" || true; exit 1
+      - name: Start Redis (docker, ephemeral port)
+        run: |
+          docker rm -f "$REDIS_CONTAINER" 2>/dev/null || true
+          docker run -d --name "$REDIS_CONTAINER" -p 0:6379 redis:7 >/dev/null
+          REDIS_PORT=$(docker port "$REDIS_CONTAINER" 6379/tcp | awk -F: '/^0\.0\.0\.0:/ {print $2; exit}')
+          [ -n "$REDIS_PORT" ] || REDIS_PORT=$(docker port "$REDIS_CONTAINER" 6379/tcp | head -1 | awk -F: '{print $NF}')
+          if [ -z "$REDIS_PORT" ]; then
+            echo "::error::Could not resolve host port for $REDIS_CONTAINER"
+            docker logs "$REDIS_CONTAINER" || true; exit 1
+          fi
+          echo "REDIS_URL=redis://127.0.0.1:${REDIS_PORT}" >> "$GITHUB_ENV"
+          for i in $(seq 1 15); do
+            docker exec "$REDIS_CONTAINER" redis-cli ping 2>/dev/null | grep -q PONG && { echo "Redis ready after ${i}s"; exit 0; }
+            sleep 1
+          done
+          echo "::error::Redis did not become ready in 15s"; docker logs "$REDIS_CONTAINER" || true; exit 1
+      - name: Build platform
+        working-directory: workspace-server
+        run: go build -o platform-server ./cmd/server
+      - name: Pick platform port
+        run: |
+          PLATFORM_PORT=$(python3 - <<'PY'
+          import socket
+          with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+              s.bind(("127.0.0.1", 0))
+              print(s.getsockname()[1])
+          PY
+          )
+          echo "PORT=${PLATFORM_PORT}" >> "$GITHUB_ENV"
+          echo "BASE=http://127.0.0.1:${PLATFORM_PORT}" >> "$GITHUB_ENV"
+          echo "Platform host port: ${PLATFORM_PORT}"
+      - name: Kill stale platform-server before start
+        run: |
+          killed=0
+          for pid in $(grep -l "platform-serve" /proc/[0-9]*/comm 2>/dev/null); do
+            kpid="${pid%/comm}"; kpid="${kpid##*/}"
+            cmdline=$(cat "/proc/${kpid}/cmdline" 2>/dev/null | tr '\0' ' ')
+            if echo "$cmdline" | grep -q "platform-server"; then
+              echo "Killing stale platform-server pid ${kpid}"
+              kill "$kpid" 2>/dev/null || true; killed=$((killed + 1))
+            fi
+          done
+          [ "$killed" -gt 0 ] && sleep 2 || true
+          echo "stale-kill done ($killed killed)"
+      - name: Start platform (background)
+        working-directory: workspace-server
+        run: |
+          ./platform-server > platform.log 2>&1 &
+          echo $! > platform.pid
+      - name: Wait for /health
+        run: |
+          for i in $(seq 1 30); do
+            curl -sf "$BASE/health" > /dev/null && { echo "Platform up after ${i}s"; exit 0; }
+            sleep 1
+          done
+          echo "::error::Platform did not become healthy in 30s"
+          cat workspace-server/platform.log || true; exit 1
+      - name: Run LOCAL fresh-provision peer-visibility E2E (literal MCP list_peers)
+        # HONEST gate — NO continue-on-error. Red today (Hermes-401 #162 /
+        # OpenClaw-never-online #165 not yet fixed); green when they land.
+        # Non-required-by-design via its distinct status context until the
+        # molecule-core#1296 flip-to-required.
+        run: bash tests/e2e/test_peer_visibility_mcp_local.sh
+      - name: Dump platform log on failure
+        if: failure()
+        run: cat workspace-server/platform.log || true
+      - name: Stop platform
+        if: always()
+        run: |
+          if [ -f workspace-server/platform.pid ]; then
+            kill "$(cat workspace-server/platform.pid)" 2>/dev/null || true
+          fi
+      - name: Stop service containers
+        if: always()
+        run: |
+          docker rm -f "$PG_CONTAINER" 2>/dev/null || true
+          docker rm -f "$REDIS_CONTAINER" 2>/dev/null || true
+
+  # Real STAGING gate: provisions a throwaway org + sibling-per-runtime,
+  # drives the LITERAL list_peers MCP call per runtime, asserts 200 +
+  # expected peer set, then scoped teardown. push(main)/dispatch/cron only.
  peer-visibility:
    name: E2E Peer Visibility
    runs-on: ubuntu-latest
@@ -52,5 +52,9 @@ jobs:
          # explicitly instead of the combined state avoids false-pause when
          # non-blocking jobs (continue-on-error: true) have failed — those
          # failures pollute combined state but do not gate merges.
+          # NOTE: the event-suffixed context name is intentional — branch protection
+          # MUST require `CI / all-required (pull_request)` (with suffix), NOT the
+          # bare `CI / all-required`. Gitea treats absent contexts as pending, not
+          # skipped; requiring the bare name silently blocks all merges (issue #1473).
          PUSH_REQUIRED_CONTEXTS: CI / all-required (push)
        run: python3 .gitea/scripts/gitea-merge-queue.py
@@ -0,0 +1,168 @@
+name: Lint forbidden tenant-env keys
+
+# RFC#523 Layer 3 (task #146): scan workspace_secrets-writer Go code
+# under workspace-server/ for new code that hardcodes a forbidden
+# operator-scope env var NAME (GITEA_TOKEN, CP_ADMIN_API_TOKEN,
+# RAILWAY_TOKEN, INFISICAL_OPERATOR_TOKEN, MOLECULE_OPERATOR_*, …).
+#
+# Catches the class "a new writer accidentally widens the propagation
+# set" — e.g. a future env-mutator plugin that sets envVars["GITEA_TOKEN"]
+# directly. Today the L1 runtime guard would abort the provision, but
+# this lint surfaces the offending code at PR review time instead of
+# at first provision attempt.
+#
+# Companion layers:
+#   - L1: workspace-server/internal/handlers/workspace_provision_forbidden_env.go
+#         (fail-closed abort at provision time)
+#   - L2: workspace/entrypoint.sh top-of-file env-grep + exit 1
+#
+# Open-source-template-friendly: the deny pattern is generic. A fork
+# can copy this workflow and replace OPERATOR_KEY_PATTERN with its
+# own operator-scope key names.
+#
+# Path-filter discipline:
+#   This workflow runs on every PR (no paths: filter — see
+#   feedback_path_filtered_workflow_cant_be_required). The scan itself
+#   targets workspace_secrets-writer paths via grep -r; it's fast
+#   (sub-second) so unconditional run is fine.
+
+on:
+  pull_request:
+    types: [opened, synchronize, reopened]
+  push:
+    branches: [main, staging]
+
+env:
+  GITHUB_SERVER_URL: https://git.moleculesai.app
+
+jobs:
+  scan:
+    name: Scan workspace_secrets writers for forbidden env keys
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          fetch-depth: 1
+
+      - name: Scan for forbidden operator-scope env key NAMES in writer paths
+        run: |
+          set -euo pipefail
+
+          # Forbidden EXACT-MATCH env var names. Kept in lockstep with
+          # workspace-server/internal/handlers/workspace_provision_forbidden_env.go
+          # forbiddenTenantEnvKeys. The Go-side test
+          # TestIsForbiddenTenantEnvKey_ExactMatches is the source of
+          # truth — if Go-side adds a key, also add it here (and
+          # vice-versa). Drift between the two is the failure mode this
+          # entire 3-layer guardrail is designed to catch.
+          FORBIDDEN_KEYS=(
+            "GITEA_TOKEN" "GITEA_PAT"
+            "GITHUB_TOKEN" "GITHUB_PAT" "GH_TOKEN"
+            "GITLAB_TOKEN" "GL_TOKEN"
+            "BITBUCKET_TOKEN"
+            "CP_ADMIN_API_TOKEN" "CP_ADMIN_TOKEN"
+            "INFISICAL_OPERATOR_TOKEN" "INFISICAL_BOOTSTRAP_TOKEN"
+            "RAILWAY_TOKEN" "RAILWAY_PERSONAL_API_TOKEN"
+            "HETZNER_TOKEN" "HETZNER_API_TOKEN"
+          )
+
+          # Forbidden PREFIX patterns — operator-scope families.
+          FORBIDDEN_PREFIXES=(
+            "MOLECULE_OPERATOR_"
+          )
+
+          # Writer paths: Go source under workspace-server/ that
+          # writes to the env-vars map or to workspace_secrets DB rows.
+          # Tests, the forbidden-env source itself, and the silent-
+          # strip denylist are exempt (they LIST the keys by design).
+          SCAN_ROOT="workspace-server/internal"
+          # Exempt paths fall in two classes:
+          #   1. The deny-set definitions + the silent-strip denylist:
+          #      they LIST the forbidden names by design.
+          #   2. Pre-RFC#523 persona-merge / config-read paths that
+          #      already handle these names correctly (the silent-
+          #      strip downstream + the new L1 fail-closed cover the
+          #      runtime risk; these reads are unchanged).
+          # New code MUST NOT be added to this list without reviewer
+          # signoff and a one-line justification in this diff.
+          EXEMPT_PATHS=(
+            # Class 1 — deny-set definitions
+            "workspace-server/internal/handlers/workspace_provision_forbidden_env.go"
+            "workspace-server/internal/handlers/workspace_provision_forbidden_env_test.go"
+            "workspace-server/internal/provisioner/provisioner.go"
+            "workspace-server/internal/provisioner/provisioner_test.go"
+            # Class 2 — pre-existing persona-fallback / org-helper paths
+            # that set the GITEA_TOKEN fallback lane (stripped downstream
+            # by provisioner.buildContainerEnv per forensic #145). The
+            # new L1 fail-closed runs BEFORE these writers, so any
+            # operator-scope leak via global/workspace_secrets is
+            # already caught. See applyAgentGitHTTPCreds doc-comment.
+            "workspace-server/internal/handlers/agent_git_identity.go"
+            "workspace-server/internal/handlers/org_helpers.go"
+            "workspace-server/internal/handlers/org.go"
+            # Class 2 — CP→platform admin auth (NOT a tenant env write;
+            # this is the control-plane HTTP auth header source).
+            "workspace-server/internal/provisioner/cp_provisioner.go"
+          )
+
+          # Build a single grep -F pattern: every forbidden key wrapped
+          # in quotes (Go string-literal form, which is how env-map
+          # writes appear). e.g. envVars["GITEA_TOKEN"] = ... or
+          # `"GITEA_TOKEN":` in a literal-map declaration.
+          #
+          # We deliberately match the quoted form so a comment that
+          # happens to spell the name without quotes (e.g. "see
+          # GITEA_TOKEN below") doesn't trip the lint.
+          PATTERN=""
+          for k in "${FORBIDDEN_KEYS[@]}"; do
+            PATTERN="${PATTERN}\"${k}\"\n"
+          done
+          for p in "${FORBIDDEN_PREFIXES[@]}"; do
+            # Prefix match needs a regex; switch to grep -E below for
+            # this slice. Kept conceptually here so the deny set lives
+            # in one place; scan is run twice (literal + prefix).
+            true
+          done
+
+          # Build exempt-paths grep filter — `grep -v -f` style.
+          EXEMPT_FILTER=$(mktemp)
+          trap 'rm -f "$EXEMPT_FILTER"' EXIT
+          for p in "${EXEMPT_PATHS[@]}"; do
+            echo "$p" >> "$EXEMPT_FILTER"
+          done
+
+          # --- Exact-match scan ---
+          HITS=""
+          for k in "${FORBIDDEN_KEYS[@]}"; do
+            # Only .go files; skip _test.go for the writer-path scan
+            # since tests legitimately reference the names. The
+            # writer-path lint targets PRODUCTION code only.
+            found=$(grep -rn --include='*.go' --exclude='*_test.go' "\"${k}\"" "$SCAN_ROOT" 2>/dev/null \
+                    | grep -v -F -f "$EXEMPT_FILTER" || true)
+            if [ -n "$found" ]; then
+              HITS="${HITS}${found}\n"
+            fi
+          done
+
+          # --- Prefix scan ---
+          for prefix in "${FORBIDDEN_PREFIXES[@]}"; do
+            found=$(grep -rnE --include='*.go' --exclude='*_test.go' "\"${prefix}[A-Z0-9_]+\"" "$SCAN_ROOT" 2>/dev/null \
+                    | grep -v -F -f "$EXEMPT_FILTER" || true)
+            if [ -n "$found" ]; then
+              HITS="${HITS}${found}\n"
+            fi
+          done
+
+          if [ -n "$HITS" ]; then
+            echo "::error::RFC#523 Layer 3: forbidden operator-scope env var name(s) hardcoded in tenant-workspace writer paths:"
+            printf "$HITS"
+            echo ""
+            echo "These env-var NAMES are on the operator-scope deny list (see"
+            echo "workspace-server/internal/handlers/workspace_provision_forbidden_env.go)."
+            echo "If your code legitimately needs to inject one of these for a"
+            echo "non-tenant code path, add the file to EXEMPT_PATHS in this"
+            echo "workflow with a one-line justification — reviewer signoff required."
+            exit 1
+          fi
+
+          echo "OK No forbidden operator-scope env key names hardcoded in writer paths."
@@ -0,0 +1,88 @@
+name: Lint shellcheck (arm64 pilot)
+
+# Mac-CI dual-track pilot (#233). ADDITIVE / NOT REQUIRED.
+#
+# Validates the arm64 self-hosted lane (no docker.sock, no privileged
+# ops) before any required gate moves onto it. Until a Mac arm64 runner
+# is registered with the `arm64` label, this workflow sits PENDING —
+# that is FINE: `arm64` is NOT in branch_protections required contexts.
+#
+# Pairs with internal#543 (RFC: Mac arm64 multi-arch runner-base).
+# No paths: filter on purpose (feedback_path_filtered_workflow_cant_be_required).
+
+on:
+  pull_request:
+    branches:
+      - main
+      - staging
+  push:
+    branches:
+      - main
+
+permissions:
+  contents: read
+
+jobs:
+  shellcheck-arm64:
+    name: shellcheck-arm64 (pilot)
+    runs-on: [self-hosted, arm64]
+    # NOT a required check; safe to sit pending until Mac runner is up.
+    # If the Mac runner has trouble pulling actions/checkout we fall
+    # back to a plain git clone (see step 'fallback clone').
+    timeout-minutes: 10
+    env:
+      GITHUB_SERVER_URL: https://git.moleculesai.app
+    steps:
+      - name: Identify runner
+        run: |
+          set -eu
+          echo "arch=$(uname -m)"
+          echo "kernel=$(uname -sr)"
+          echo "shell=$BASH_VERSION"
+          # Sanity: must actually be arm64. If amd64 sneaks in here,
+          # fail fast — that means the label routing is wrong.
+          case "$(uname -m)" in
+            aarch64|arm64) echo "arm64 confirmed" ;;
+            *) echo "ERROR: expected arm64, got $(uname -m)"; exit 1 ;;
+          esac
+
+      - name: Checkout
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 1
+
+      - name: Install shellcheck (arm64)
+        run: |
+          set -eu
+          if command -v shellcheck >/dev/null 2>&1; then
+            echo "shellcheck already present: $(shellcheck --version | head -1)"
+          else
+            # Prefer apt if the runner base ships it; else download arm64 binary.
+            if command -v apt-get >/dev/null 2>&1; then
+              sudo apt-get update -qq
+              sudo apt-get install -y --no-install-recommends shellcheck
+            else
+              SC_VER=v0.10.0
+              curl -fsSL "https://github.com/koalaman/shellcheck/releases/download/${SC_VER}/shellcheck-${SC_VER}.linux.aarch64.tar.xz" \
+                | tar -xJf - --strip-components=1
+              sudo mv shellcheck /usr/local/bin/
+            fi
+          fi
+          shellcheck --version | head -2
+
+      - name: Run shellcheck on .gitea/scripts/*.sh
+        run: |
+          set -eu
+          # Only the scripts we control under .gitea/scripts. Pilot
+          # scope is intentionally narrow — broaden in a follow-up
+          # once the lane is proven.
+          mapfile -t TARGETS < <(find .gitea/scripts -maxdepth 2 -type f -name '*.sh' | sort)
+          if [ "${#TARGETS[@]}" -eq 0 ]; then
+            echo "No .sh files found under .gitea/scripts — nothing to check"
+            exit 0
+          fi
+          echo "Checking ${#TARGETS[@]} file(s):"
+          printf '  %s\n' "${TARGETS[@]}"
+          # SC1091 = couldn't follow non-constant source; expected for
+          # CI-time analysis without the full runtime layout.
+          shellcheck --severity=error --exclude=SC1091 "${TARGETS[@]}"
@@ -104,7 +104,7 @@ jobs:
        with:
          python-version: "3.11"

-      - name: Compute next version from PyPI latest
+      - name: Compute next version from PyPI latest and existing tags
        id: bump
        run: |
          set -eu
@@ -112,9 +112,24 @@ jobs:
            | python -c "import sys,json; print(json.load(sys.stdin)['info']['version'])")
          MAJOR=$(echo "$LATEST" | cut -d. -f1)
          MINOR=$(echo "$LATEST" | cut -d. -f2)
-          PATCH=$(echo "$LATEST" | cut -d. -f3)
-          VERSION="${MAJOR}.${MINOR}.$((PATCH+1))"
-          echo "PyPI latest=$LATEST -> next=$VERSION"
+          TAG_LATEST=$(git tag --list "runtime-v${MAJOR}.${MINOR}.*" \
+            | sed -E 's/^runtime-v//' \
+            | grep -E '^[0-9]+\.[0-9]+\.[0-9]+$' \
+            | sort -V \
+            | tail -1 || true)
+          VERSION=$(PYPI_LATEST="$LATEST" TAG_LATEST="$TAG_LATEST" python - <<'PY'
+          import os
+
+          def parse(v):
+              return tuple(int(part) for part in v.split("."))
+
+          pypi = os.environ["PYPI_LATEST"]
+          tag = os.environ.get("TAG_LATEST") or pypi
+          base = max(parse(pypi), parse(tag))
+          print(f"{base[0]}.{base[1]}.{base[2] + 1}")
+          PY
+          )
+          echo "PyPI latest=$LATEST, latest runtime tag=${TAG_LATEST:-none} -> next=$VERSION"
          if ! echo "$VERSION" | grep -qE '^[0-9]+\.[0-9]+\.[0-9]+$'; then
            echo "::error::computed version $VERSION does not match PEP 440 X.Y.Z"
            exit 1
@@ -89,6 +89,7 @@ on:
 permissions:
  contents: read
  pull-requests: read
+  secrets: read

 jobs:
  # bp-exempt: PR review bot signal; required merge state is enforced by CI / all-required.
@@ -30,6 +30,11 @@ jobs:
  scan:
    name: Scan diff for credential-shaped strings
    runs-on: ubuntu-latest
+    # Hard CI gate — must complete or the PR is unmergable. 10-minute ceiling
+    # is generous for a diff-scan against a single SHA. If this times out, the
+    # runner is frozen and holding a slot — the step timeout triggers clean
+    # failure, releasing the runner for the next job.
+    timeout-minutes: 10
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
@@ -133,6 +138,14 @@ jobs:
            [ -z "$f" ] && continue
            [ "$f" = "$SELF_GITHUB" ] && continue
            [ "$f" = "$SELF_GITEA" ] && continue
+            # Test-fixture exclude (internal#425): the secrets-detector's OWN
+            # unit-test corpus deliberately embeds credential-SHAPED example
+            # strings to exercise the detector. Verified 2026-05-18 synthetic
+            # (fabricated ghp_* fixtures, not real). Without this the scanner
+            # self-trips on its own fixtures and fail-closes every deploy.
+            # Same rationale as the SELF_* excludes above; gate NOT weakened
+            # (all other paths still fully scanned).
+            [ "$f" = "workspace-server/internal/secrets/patterns_test.go" ] && continue
            if [ -n "$DIFF_RANGE" ]; then
              ADDED=$(git diff --no-color --unified=0 "$BASE" "$HEAD" -- "$f" 2>/dev/null | grep -E '^\+[^+]' || true)
            else
@@ -16,6 +16,7 @@ on:
 permissions:
  contents: read
  pull-requests: read
+  secrets: read

 jobs:
  # bp-exempt: PR security review bot signal; required merge state is enforced by CI / all-required.
@@ -84,11 +84,8 @@ on:
 permissions:
  contents: read
  pull-requests: read
-  # NOTE: `statuses: write` is the GitHub-Actions name for POST /statuses.
-  # Gitea 1.22.6 may not gate on this permission key (it just checks the
-  # token), but listing it explicitly documents intent for the next
-  # platform-version upgrade.
  statuses: write
+  secrets: read

 jobs:
  all-items-acked:
@@ -71,6 +71,7 @@ jobs:
    permissions:
      contents: read
      pull-requests: read
+      secrets: read
    steps:
      - name: Check out base branch (for the script)
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
@@ -1,255 +0,0 @@
-name: canary-verify
-
-# Runs the canary smoke suite against the staging canary tenant fleet
-# after a new :staging-<sha> image lands in ECR. On green, calls the
-# CP redeploy-fleet endpoint to promote :staging-<sha> → :latest so
-# the prod tenant fleet's 5-minute auto-updater picks up the verified
-# digest. On red, :latest stays on the prior known-good digest and
-# prod is untouched.
-#
-# Registry note (2026-05-10): This workflow previously used GHCR
-# (ghcr.io/molecule-ai/platform-tenant) — that registry was retired
-# during the 2026-05-06 Gitea suspension migration when publish-
-# workspace-server-image.yml switched to the operator's ECR org
-# (153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/
-# platform-tenant). The GHCR → ECR migration was never applied to
-# this file, so canary-verify was silently smoke-testing the stale
-# GHCR image while the actual staging/prod tenants ran the ECR image.
-# Result: smoke tests could not catch a broken ECR build. Fix:
-#   - Wait step: reads SHA from running canary /health (tenant-
-#     agnostic, works regardless of registry).
-#   - Promote step: calls CP redeploy-fleet endpoint with target_tag=
-#     staging-<sha>, same mechanism as redeploy-tenants-on-main.yml.
-#     No longer attempts GHCR crane ops.
-#
-# Dependencies:
-#   - publish-workspace-server-image.yml publishes :staging-<sha>
-#     to ECR on staging and main merges.
-#   - Canary tenants are configured to pull :staging-<sha> from ECR
-#     (TENANT_IMAGE env set to the ECR :staging-<sha> tag).
-#   - Repo secrets CANARY_TENANT_URLS / CANARY_ADMIN_TOKENS /
-#     CANARY_CP_SHARED_SECRET are populated.
-
-on:
-  workflow_run:
-    workflows: ["publish-workspace-server-image"]
-    types: [completed]
-  workflow_dispatch:
-
-permissions:
-  contents: read
-  packages: write
-  actions: read
-
-env:
-  # ECR registry (post-2026-05-06 SSOT for tenant images).
-  # publish-workspace-server-image.yml pushes here.
-  IMAGE_NAME: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/platform
-  TENANT_IMAGE_NAME: 153263036946.dkr.ecr.us-east-2.amazonaws.com/molecule-ai/platform-tenant
-  # CP endpoint for redeploy-fleet (used in promote step below).
-  CP_URL: ${{ vars.CP_URL || 'https://staging-api.moleculesai.app' }}
-
-jobs:
-  canary-smoke:
-    # Skip when the upstream workflow failed — no image to test against.
-    if: ${{ github.event.workflow_run.conclusion == 'success' || github.event_name == 'workflow_dispatch' }}
-    runs-on: ubuntu-latest
-    outputs:
-      sha: ${{ steps.compute.outputs.sha }}
-      smoke_ran: ${{ steps.smoke.outputs.ran }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-
-      - name: Compute sha
-        id: compute
-        run: echo "sha=${GITHUB_SHA::7}" >> "$GITHUB_OUTPUT"
-
-      - name: Wait for canary tenants to pick up :staging-<sha>
-        # Poll canary health endpoints every 30s for up to 7 min instead
-        # of a fixed 6-min sleep. Exits as soon as ALL canaries report
-        # the new SHA (~2-3 min typical vs 6 min fixed). Falls back to
-        # proceeding after 7 min even if not all canaries responded —
-        # the smoke suite will catch any that didn't update.
-        #
-        # NOTE: The SHA is read from the running tenant's /health response,
-        # NOT from a registry lookup. This is registry-agnostic and works
-        # regardless of whether the tenant pulls from ECR, GHCR, or any
-        # other registry — the canary is telling us what it's actually
-        # running, which is the ground truth for smoke testing.
-        env:
-          CANARY_TENANT_URLS: ${{ secrets.CANARY_TENANT_URLS }}
-          EXPECTED_SHA: ${{ steps.compute.outputs.sha }}
-        run: |
-          if [ -z "$CANARY_TENANT_URLS" ]; then
-            echo "No canary URLs configured — falling back to 60s wait"
-            sleep 60
-            exit 0
-          fi
-          IFS=',' read -ra URLS <<< "$CANARY_TENANT_URLS"
-          MAX_WAIT=420  # 7 minutes
-          INTERVAL=30
-          ELAPSED=0
-          while [ $ELAPSED -lt $MAX_WAIT ]; do
-            ALL_READY=true
-            for url in "${URLS[@]}"; do
-              HEALTH=$(curl -s --max-time 5 "${url}/health" 2>/dev/null || echo "{}")
-              SHA=$(echo "$HEALTH" | grep -o "\"sha\":\"[^\"]*\"" | head -1 | cut -d'"' -f4)
-              if [ "$SHA" != "$EXPECTED_SHA" ]; then
-                ALL_READY=false
-                break
-              fi
-            done
-            if $ALL_READY; then
-              echo "All canaries running staging-${EXPECTED_SHA} after ${ELAPSED}s"
-              exit 0
-            fi
-            echo "Waiting for canaries... (${ELAPSED}s / ${MAX_WAIT}s)"
-            sleep $INTERVAL
-            ELAPSED=$((ELAPSED + INTERVAL))
-          done
-          echo "Timeout after ${MAX_WAIT}s — proceeding anyway (smoke suite will validate)"
-
-      - name: Run canary smoke suite
-        id: smoke
-        # Graceful-skip when no canary fleet is configured (Phase 2 not yet
-        # stood up — see molecule-controlplane/docs/canary-tenants.md).
-        # Sets `ran=false` on skip so promote-to-latest stays off (we don't
-        # want every main merge auto-promoting without gating). Manual
-        # promote-latest.yml is the release gate while canary is absent.
-        # Once the fleet is real: delete the early-exit branch.
-        env:
-          CANARY_TENANT_URLS: ${{ secrets.CANARY_TENANT_URLS }}
-          CANARY_ADMIN_TOKENS: ${{ secrets.CANARY_ADMIN_TOKENS }}
-          CANARY_CP_BASE_URL: https://staging-api.moleculesai.app
-          CANARY_CP_SHARED_SECRET: ${{ secrets.CANARY_CP_SHARED_SECRET }}
-        run: |
-          set -euo pipefail
-          if [ -z "${CANARY_TENANT_URLS:-}" ] \
-            || [ -z "${CANARY_ADMIN_TOKENS:-}" ] \
-            || [ -z "${CANARY_CP_SHARED_SECRET:-}" ]; then
-            {
-              echo "## ⚠️ canary-verify skipped"
-              echo
-              echo "One or more canary secrets are unset (\`CANARY_TENANT_URLS\`, \`CANARY_ADMIN_TOKENS\`, \`CANARY_CP_SHARED_SECRET\`)."
-              echo "Phase 2 canary fleet has not been stood up yet —"
-              echo "see [canary-tenants.md](https://git.moleculesai.app/molecule-ai/molecule-controlplane/blob/main/docs/canary-tenants.md)."
-              echo
-              echo "**Skipped — promote-to-latest will NOT auto-fire.** Dispatch \`promote-latest.yml\` manually when ready."
-            } >> "$GITHUB_STEP_SUMMARY"
-            echo "ran=false" >> "$GITHUB_OUTPUT"
-            echo "::notice::canary-verify: skipped — no canary fleet configured"
-            exit 0
-          fi
-          bash scripts/canary-smoke.sh
-          echo "ran=true" >> "$GITHUB_OUTPUT"
-
-      - name: Summary on failure
-        if: ${{ failure() }}
-        run: |
-          {
-            echo "## Canary smoke FAILED"
-            echo
-            echo "Canary tenants rejected image \`staging-${{ steps.compute.outputs.sha }}\`."
-            echo ":latest stays pinned to the prior good digest — prod is untouched."
-            echo
-            echo "Fix forward and merge again, or investigate the specific failed"
-            echo "assertions in the canary-smoke step log above."
-          } >> "$GITHUB_STEP_SUMMARY"
-
-  promote-to-latest:
-    # On green, calls the CP redeploy-fleet endpoint with target_tag=
-    # staging-<sha> to promote the verified ECR image. This is the same
-    # mechanism as redeploy-tenants-on-main.yml — no GHCR crane ops.
-    #
-    # Pre-fix history: the old GHCR promote step used `crane tag` against
-    # ghcr.io/molecule-ai/platform-tenant, but publish-workspace-server-
-    # image.yml had already migrated to ECR on 2026-05-07 (commit
-    # 10e510f5). The GHCR tags were never updated, so this step was
-    # silently promoting a stale GHCR image while actual prod tenants
-    # pulled from ECR. Canary smoke tests were GHCR-targeted and could
-    # not catch a broken ECR build.
-    needs: canary-smoke
-    if: ${{ needs.canary-smoke.result == 'success' && needs.canary-smoke.outputs.smoke_ran == 'true' }}
-    runs-on: ubuntu-latest
-    env:
-      SHA: ${{ needs.canary-smoke.outputs.sha }}
-      CP_URL: ${{ vars.CP_URL || 'https://staging-api.moleculesai.app' }}
-      # CP_ADMIN_API_TOKEN gates write access to the redeploy endpoint.
-      # Stored at the repo level so all workflows pick it up automatically.
-      CP_ADMIN_API_TOKEN: ${{ secrets.CP_ADMIN_API_TOKEN }}
-      # canary_slug pin: deploy the verified :staging-<sha> to the canary
-      # first (soak 120s), then fan out to the rest of the fleet.
-      CANARY_SLUG: ${{ vars.CANARY_PROMOTE_SLUG || '' }}
-      SOAK_SECONDS: ${{ vars.CANARY_PROMOTE_SOAK || '120' }}
-      BATCH_SIZE: ${{ vars.CANARY_PROMOTE_BATCH || '3' }}
-    steps:
-      - name: Check CP credentials
-        run: |
-          if [ -z "${CP_ADMIN_API_TOKEN:-}" ]; then
-            echo "::error::CP_ADMIN_API_TOKEN secret is not set — promote step cannot call redeploy-fleet."
-            echo "::error::Set it at: repo Settings → Actions → Variables and Secrets → New Secret."
-            exit 1
-          fi
-
-      - name: Promote verified ECR image to :latest
-        run: |
-          set -euo pipefail
-
-          TARGET_TAG="staging-${SHA}"
-          BODY=$(jq -nc \
-            --arg tag "$TARGET_TAG" \
-            --argjson soak "${SOAK_SECONDS:-120}" \
-            --argjson batch "${BATCH_SIZE:-3}" \
-            --argjson dry false \
-            '{
-              target_tag: $tag,
-              soak_seconds: $soak,
-              batch_size: $batch,
-              dry_run: $dry
-            }')
-
-          if [ -n "${CANARY_SLUG:-}" ]; then
-            BODY=$(jq '. * {canary_slug: $slug}' --arg slug "$CANARY_SLUG" <<<"$BODY")
-          fi
-
-          echo "Calling: POST $CP_URL/cp/admin/tenants/redeploy-fleet"
-          echo "  target_tag: $TARGET_TAG"
-          echo "  body: $BODY"
-
-          HTTP_RESPONSE=$(mktemp)
-          HTTP_CODE_FILE=$(mktemp)
-          set +e
-          curl -sS -o "$HTTP_RESPONSE" -w '%{http_code}' \
-            -m 1200 \
-            -H "Authorization: Bearer $CP_ADMIN_API_TOKEN" \
-            -H "Content-Type: application/json" \
-            -X POST "$CP_URL/cp/admin/tenants/redeploy-fleet" \
-            -d "$BODY" >"$HTTP_CODE_FILE"
-          CURL_EXIT=$?
-          set -e
-
-          HTTP_CODE=$(cat "$HTTP_CODE_FILE" 2>/dev/null || echo "000")
-          [ -z "$HTTP_CODE" ] && HTTP_CODE="000"
-
-          echo "HTTP $HTTP_CODE (curl exit $CURL_EXIT)"
-          cat "$HTTP_RESPONSE" | jq . || cat "$HTTP_RESPONSE"
-
-          if [ "$HTTP_CODE" -ge 400 ]; then
-            echo "::error::CP redeploy-fleet returned HTTP $HTTP_CODE — refusing to proceed."
-            exit 1
-          fi
-
-      - name: Summary
-        run: |
-          {
-            echo "## Canary verified — :latest promoted via CP redeploy-fleet"
-            echo ""
-            echo "- **Target tag:** \`staging-${{ needs.canary-smoke.outputs.sha }}\`"
-            echo "- **Registry:** ECR (\`${TENANT_IMAGE_NAME}\`)"
-            echo "- **Canary slug:** \`${CANARY_SLUG:-<none>}\` (soak ${SOAK_SECONDS}s)"
-            echo "- **Batch size:** ${BATCH_SIZE:-3}"
-            echo ""
-            echo "CP redeploy-fleet is rolling out the verified image across the prod fleet."
-            echo "The fleet's 5-minute health-check loop will pick up the update automatically."
-          } >> "$GITHUB_STEP_SUMMARY"
@@ -1,400 +0,0 @@
-name: redeploy-tenants-on-main
-
-# Auto-refresh prod tenant EC2s after every main merge.
-#
-# Why this workflow exists: publish-workspace-server-image builds and
-# pushes a new platform-tenant :<sha> to ECR on every merge to main,
-# but running tenants pulled their image once at boot and never re-pull.
-# Users see stale code indefinitely.
-#
-# This workflow closes the gap by calling the control-plane admin
-# endpoint that performs a canary-first, batched, health-gated rolling
-# redeploy across every live tenant. Implemented in molecule-ai/
-# molecule-controlplane as POST /cp/admin/tenants/redeploy-fleet
-# (feat/tenant-auto-redeploy, landing alongside this workflow).
-#
-# Registry: ECR (153263036946.dkr.ecr.us-east-2.amazonaws.com/
-# molecule-ai/platform-tenant). GHCR was retired 2026-05-07 during the
-# Gitea suspension migration. The canary-verify.yml promote step now
-# uses the same redeploy-fleet endpoint (fixes the silent-GHCR gap).
-#
-# Runtime ordering:
-#   1. publish-workspace-server-image completes → new :staging-<sha> in ECR.
-#   2. This workflow fires via workflow_run, calls redeploy-fleet with
-#      target_tag=staging-<sha>. No CDN propagation wait needed —
-#      ECR image manifest is consistent immediately after push.
-#   3. Calls redeploy-fleet with canary_slug (if set) and a soak
-#      period. Canary proves the image boots; batches follow.
-#   4. Any failure aborts the rollout and leaves older tenants on the
-#      prior image — safer default than half-and-half state.
-#
-# Rollback path: re-run this workflow with a specific SHA pinned via
-# the workflow_dispatch input. That calls redeploy-fleet with
-# target_tag=<sha>, re-pulling the older image on every tenant.
-
-on:
-  workflow_run:
-    workflows: ['publish-workspace-server-image']
-    types: [completed]
-    branches: [main]
-  workflow_dispatch:
-    inputs:
-      target_tag:
-        # Empty default → auto-trigger and dispatch-without-input both
-        # resolve to `staging-<short_head_sha>` (the digest publish-image
-        # just pushed). Pre-fix this defaulted to 'latest', which only
-        # gets retagged by canary-verify's promote-to-latest job — and
-        # that job soft-skips when CANARY_TENANT_URLS is unset (the
-        # current state, until Phase 2 canary fleet is live). Result:
-        # `:latest` had been pinned to a 4-day-old digest (2026-04-28)
-        # while every main push pushed fresh `staging-<sha>` images;
-        # every prod redeploy pulled the stale `:latest` and the verify
-        # step correctly flagged 3/3 tenants STALE. Pulling the
-        # just-published `staging-<sha>` directly skips the dead retag
-        # path. When canary fleet is real, this workflow should chain
-        # on canary-verify completion (workflow_run from canary-verify),
-        # not publish-image — separate, smaller PR.
-        description: 'Tenant image tag to deploy (e.g. "latest", "staging-a59f1a6c"). Empty = auto staging-<head_sha>.'
-        required: false
-        type: string
-        default: ''
-      canary_slug:
-        description: 'Tenant slug to deploy first + soak (empty = skip canary, fan out immediately).'
-        required: false
-        type: string
-        # Must be an actual prod tenant slug (current: hongming,
-        # chloe-dong, reno-stars). The previous default 'hongmingwang'
-        # didn't match any tenant — CP soft-skipped the missing canary
-        # and the fleet rolled out without the soak gate, defeating the
-        # whole point of canary-first.
-        default: 'hongming'
-      soak_seconds:
-        description: 'Seconds to wait after canary before fanning out.'
-        required: false
-        type: string
-        default: '60'
-      batch_size:
-        description: 'How many tenants SSM redeploys in parallel per batch.'
-        required: false
-        type: string
-        default: '3'
-      dry_run:
-        description: 'Plan only — do not actually redeploy.'
-        required: false
-        type: boolean
-        default: false
-
-permissions:
-  contents: read
-  # No write scopes needed — the workflow hits an external CP endpoint,
-  # not the GitHub API.
-
-# Serialize redeploys so two rapid main pushes' redeploys don't overlap
-# and cause confusing per-tenant SSM state. Without this, GitHub's
-# implicit workflow_run queueing would *probably* serialize them, but
-# the explicit block makes the invariant defensible. Mirrors the
-# concurrency block on redeploy-tenants-on-staging.yml for shape parity.
-#
-# cancel-in-progress: false → aborting a half-rolled-out fleet would
-# leave tenants stuck on whatever image they happened to be on when
-# cancelled. Better to finish the in-flight rollout before starting
-# the next one.
-concurrency:
-  group: redeploy-tenants-on-main
-  cancel-in-progress: false
-
-jobs:
-  redeploy:
-    # Skip the auto-trigger if publish-workspace-server-image didn't
-    # actually succeed. workflow_run fires on any completion state; we
-    # don't want to redeploy against a half-built image.
-    if: |
-      github.event_name == 'workflow_dispatch' ||
-      (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success')
-    runs-on: ubuntu-latest
-    timeout-minutes: 25
-    steps:
-      - name: Note on ECR propagation
-        # ECR image manifests are consistent immediately after push — no
-        # CDN cache to wait for. The old GHCR-based workflow had a 30s
-        # sleep to avoid race conditions; ECR makes that unnecessary.
-        run: echo "ECR image available immediately after push — proceeding."
-
-      - name: Compute target tag
-        id: tag
-        # Resolution order:
-        #   1. Operator-supplied input (workflow_dispatch with explicit
-        #      tag) → used verbatim. Lets ops pin `latest` for emergency
-        #      rollback to last canary-verified digest, or pin a specific
-        #      `staging-<sha>` to roll back to a known-good build.
-        #   2. Default → `staging-<short_head_sha>`. The just-published
-        #      digest. Bypasses the `:latest` retag path that's currently
-        #      dead (canary-verify soft-skips without canary fleet, so
-        #      the only thing retagging `:latest` today is the manual
-        #      promote-latest.yml — last run 2026-04-28). Auto-trigger
-        #      from workflow_run uses workflow_run.head_sha; manual
-        #      dispatch with no input falls through to github.sha.
-        env:
-          INPUT_TAG: ${{ inputs.target_tag }}
-          HEAD_SHA: ${{ github.event.workflow_run.head_sha || github.sha }}
-        run: |
-          set -euo pipefail
-          if [ -n "${INPUT_TAG:-}" ]; then
-            echo "target_tag=$INPUT_TAG" >> "$GITHUB_OUTPUT"
-            echo "Using operator-pinned tag: $INPUT_TAG"
-          else
-            SHORT="${HEAD_SHA:0:7}"
-            echo "target_tag=staging-$SHORT" >> "$GITHUB_OUTPUT"
-            echo "Using auto tag: staging-$SHORT (head_sha=$HEAD_SHA)"
-          fi
-
-      - name: Call CP redeploy-fleet
-        # CP_ADMIN_API_TOKEN must be set as a repo/org secret on
-        # molecule-ai/molecule-core, matching the staging/prod CP's
-        # CP_ADMIN_API_TOKEN env. Stored in Railway, mirrored to this
-        # repo's secrets for CI.
-        env:
-          CP_URL: ${{ vars.CP_URL || 'https://api.moleculesai.app' }}
-          CP_ADMIN_API_TOKEN: ${{ secrets.CP_ADMIN_API_TOKEN }}
-          TARGET_TAG: ${{ steps.tag.outputs.target_tag }}
-          CANARY_SLUG: ${{ inputs.canary_slug || 'hongming' }}
-          SOAK_SECONDS: ${{ inputs.soak_seconds || '60' }}
-          BATCH_SIZE: ${{ inputs.batch_size || '3' }}
-          DRY_RUN: ${{ inputs.dry_run || false }}
-        run: |
-          set -euo pipefail
-
-          if [ -z "${CP_ADMIN_API_TOKEN:-}" ]; then
-            echo "::error::CP_ADMIN_API_TOKEN secret not set — skipping redeploy"
-            echo "::notice::Set CP_ADMIN_API_TOKEN in repo secrets to enable auto-redeploy."
-            exit 1
-          fi
-
-          BODY=$(jq -nc \
-            --arg tag "$TARGET_TAG" \
-            --arg canary "$CANARY_SLUG" \
-            --argjson soak "$SOAK_SECONDS" \
-            --argjson batch "$BATCH_SIZE" \
-            --argjson dry "$DRY_RUN" \
-            '{
-              target_tag: $tag,
-              canary_slug: $canary,
-              soak_seconds: $soak,
-              batch_size: $batch,
-              dry_run: $dry
-            }')
-
-          echo "POST $CP_URL/cp/admin/tenants/redeploy-fleet"
-          echo "  body: $BODY"
-
-          HTTP_RESPONSE=$(mktemp)
-          HTTP_CODE_FILE=$(mktemp)
-          # Route -w into its own tempfile so curl's exit code (e.g. 56
-          # on connection-reset, 22 on --fail-with-body 4xx/5xx) can't
-          # pollute the captured stdout. The previous inline-substitution
-          # shape produced "000000" on connection reset (curl wrote
-          # "000" via -w, then the inline echo-fallback appended another
-          # "000") — caught on the 2026-05-04 redeploy of sha 2b862f6.
-          # set +e/-e keeps the non-zero curl exit from tripping the
-          # outer pipeline. See lint-curl-status-capture.yml for the
-          # CI gate that pins this fix shape.
-          set +e
-          curl -sS -o "$HTTP_RESPONSE" -w '%{http_code}' \
-            -m 1200 \
-            -H "Authorization: Bearer $CP_ADMIN_API_TOKEN" \
-            -H "Content-Type: application/json" \
-            -X POST "$CP_URL/cp/admin/tenants/redeploy-fleet" \
-            -d "$BODY" >"$HTTP_CODE_FILE"
-          set -e
-          # Stderr from curl (e.g. dial errors with -sS) goes to the runner
-          # log so operators can see WHY a connection failed. Stdout is
-          # captured to $HTTP_CODE_FILE because that's where -w writes.
-          HTTP_CODE=$(cat "$HTTP_CODE_FILE" 2>/dev/null || echo "000")
-          [ -z "$HTTP_CODE" ] && HTTP_CODE="000"
-
-          echo "HTTP $HTTP_CODE"
-          cat "$HTTP_RESPONSE" | jq . || cat "$HTTP_RESPONSE"
-
-          # Pretty-print per-tenant results in the job summary so
-          # ops can see which tenants were redeployed without drilling
-          # into the raw response.
-          {
-            echo "## Tenant redeploy fleet"
-            echo ""
-            echo "**Target tag:** \`$TARGET_TAG\`"
-            echo "**Canary:** \`$CANARY_SLUG\` (soak ${SOAK_SECONDS}s)"
-            echo "**Batch size:** $BATCH_SIZE"
-            echo "**Dry run:** $DRY_RUN"
-            echo "**HTTP:** $HTTP_CODE"
-            echo ""
-            echo "### Per-tenant result"
-            echo ""
-            echo '| Slug | Phase | SSM Status | Exit | Healthz | Error |'
-            echo '|------|-------|------------|------|---------|-------|'
-            jq -r '.results[]? | "| \(.slug) | \(.phase) | \(.ssm_status // "-") | \(.ssm_exit_code) | \(.healthz_ok) | \(.error // "-") |"' "$HTTP_RESPONSE" || true
-          } >> "$GITHUB_STEP_SUMMARY"
-
-          if [ "$HTTP_CODE" != "200" ]; then
-            echo "::error::redeploy-fleet returned HTTP $HTTP_CODE"
-            exit 1
-          fi
-          OK=$(jq -r '.ok' "$HTTP_RESPONSE")
-          if [ "$OK" != "true" ]; then
-            echo "::error::redeploy-fleet reported ok=false (see summary for which tenant halted the rollout)"
-            exit 1
-          fi
-          echo "::notice::Tenant fleet redeploy reported ssm_status=Success — verifying actual image roll on each tenant..."
-
-          # Stash the response for the verify step. $RUNNER_TEMP outlasts
-          # the step boundary; $HTTP_RESPONSE doesn't.
-          cp "$HTTP_RESPONSE" "$RUNNER_TEMP/redeploy-response.json"
-
-      - name: Verify each tenant /buildinfo matches published SHA
-        # ROOT FIX FOR #2395.
-        #
-        # `redeploy-fleet`'s `ssm_status=Success` means "the SSM RPC
-        # didn't error" — NOT "the new image is running on the tenant."
-        # `:latest` lives in the local Docker daemon's image cache; if
-        # the SSM document does `docker compose up -d` without an
-        # explicit `docker pull`, the daemon serves the previously-
-        # cached digest and the container restarts on stale code.
-        # 2026-04-30 incident: hongmingwang's tenant reported
-        # ssm_status=Success at 17:00:53Z but kept serving pre-501a42d7
-        # chat_files for 30+ min — the lazy-heal fix never reached the
-        # user despite green deploy + green redeploy.
-        #
-        # This step closes the gap by curling each tenant's /buildinfo
-        # endpoint (added in workspace-server/internal/buildinfo +
-        # /Dockerfile* GIT_SHA build-arg, this PR) and comparing the
-        # returned git_sha to the SHA the workflow expects. Mismatches
-        # fail the workflow, which is what `ok=true` should have
-        # guaranteed all along.
-        #
-        # When the redeploy was triggered by workflow_dispatch with a
-        # specific tag (target_tag != "latest"), the expected SHA may
-        # not equal ${{ github.sha }} — in that case we resolve via
-        # GHCR's manifest. For workflow_run (default :latest) the
-        # workflow_run.head_sha is the SHA that just published.
-        env:
-          EXPECTED_SHA: ${{ github.event.workflow_run.head_sha || github.sha }}
-          TARGET_TAG: ${{ steps.tag.outputs.target_tag }}
-          # Tenant subdomain template — slugs from the response are
-          # appended. Production CP issues `<slug>.moleculesai.app`;
-          # staging CP issues `<slug>.staging.moleculesai.app`. This
-          # workflow runs on main → prod CP → no `staging.` infix.
-          TENANT_DOMAIN: 'moleculesai.app'
-        run: |
-          set -euo pipefail
-
-          EXPECTED_SHORT="${EXPECTED_SHA:0:7}"
-          if [ "$TARGET_TAG" != "latest" ] \
-             && [ "$TARGET_TAG" != "$EXPECTED_SHA" ] \
-             && [ "$TARGET_TAG" != "staging-$EXPECTED_SHORT" ]; then
-            # workflow_dispatch with a pinned tag that isn't the head
-            # SHA — operator is rolling back / pinning. Skip the
-            # verification because we don't have the expected SHA in
-            # this context (would need to crane-inspect the GHCR
-            # manifest, which is a follow-up). Failing-open here is
-            # safe: the operator chose the tag deliberately.
-            #
-            # `staging-<short_head_sha>` IS verified — it's the new
-            # auto-trigger default (see Compute target tag step) and
-            # the digest under that tag SHOULD match EXPECTED_SHA.
-            echo "::notice::target_tag=$TARGET_TAG (operator-pinned) — skipping per-tenant SHA verification."
-            exit 0
-          fi
-
-          RESP="$RUNNER_TEMP/redeploy-response.json"
-          if [ ! -s "$RESP" ]; then
-            echo "::error::redeploy-response.json missing or empty — verify step ran without a response to read"
-            exit 1
-          fi
-
-          # Pull only successfully-redeployed tenants. Any tenant that
-          # halted the rollout already failed the previous step, so we
-          # don't double-count them here.
-          mapfile -t SLUGS < <(jq -r '.results[]? | select(.healthz_ok == true) | .slug' "$RESP")
-          if [ ${#SLUGS[@]} -eq 0 ]; then
-            echo "::warning::No tenants reported healthz_ok — nothing to verify"
-            exit 0
-          fi
-
-          echo "Verifying ${#SLUGS[@]} tenant(s) against EXPECTED_SHA=${EXPECTED_SHA:0:7}..."
-
-          # Two distinct failure modes — STALE (the #2395 bug class, hard-fail)
-          # vs UNREACHABLE (teardown race, soft-warn). See the staging variant's
-          # comment for the full rationale; same logic applies on prod even
-          # though prod has fewer ephemeral tenants — the asymmetry would be a
-          # gratuitous fork.
-          STALE_COUNT=0
-          UNREACHABLE_COUNT=0
-          STALE_LINES=()
-          UNREACHABLE_LINES=()
-          for slug in "${SLUGS[@]}"; do
-            URL="https://${slug}.${TENANT_DOMAIN}/buildinfo"
-            # 30s total: tenant just SSM-restarted, may still be coming
-            # up. Retry-on-empty rather than retry-on-status — we want
-            # to fail fast on "responded with wrong SHA", not "still
-            # warming up".
-            BODY=$(curl -sS --max-time 30 --retry 3 --retry-delay 5 --retry-connrefused "$URL" || true)
-            ACTUAL_SHA=$(echo "$BODY" | jq -r '.git_sha // ""' 2>/dev/null || echo "")
-            if [ -z "$ACTUAL_SHA" ]; then
-              UNREACHABLE_COUNT=$((UNREACHABLE_COUNT + 1))
-              UNREACHABLE_LINES+=("| $slug | (no /buildinfo response) | ${EXPECTED_SHA:0:7} | ⚠ unreachable (likely teardown race) |")
-              continue
-            fi
-            if [ "$ACTUAL_SHA" = "$EXPECTED_SHA" ]; then
-              echo "  $slug: ${ACTUAL_SHA:0:7} ✓"
-            else
-              STALE_COUNT=$((STALE_COUNT + 1))
-              STALE_LINES+=("| $slug | ${ACTUAL_SHA:0:7} | ${EXPECTED_SHA:0:7} | ❌ stale |")
-            fi
-          done
-
-          {
-            echo ""
-            echo "### Per-tenant /buildinfo verification"
-            echo ""
-            echo "Expected SHA: \`${EXPECTED_SHA:0:7}\`"
-            echo ""
-            if [ $STALE_COUNT -gt 0 ]; then
-              echo "**${STALE_COUNT} STALE tenant(s) — these did NOT pick up the new image despite ssm_status=Success:**"
-              echo ""
-              echo "| Slug | Actual /buildinfo SHA | Expected | Status |"
-              echo "|------|----------------------|----------|--------|"
-              for line in "${STALE_LINES[@]}"; do echo "$line"; done
-              echo ""
-            fi
-            if [ $UNREACHABLE_COUNT -gt 0 ]; then
-              echo "**${UNREACHABLE_COUNT} unreachable tenant(s) — likely teardown race (soft-warn, not failing):**"
-              echo ""
-              echo "| Slug | Actual /buildinfo SHA | Expected | Status |"
-              echo "|------|----------------------|----------|--------|"
-              for line in "${UNREACHABLE_LINES[@]}"; do echo "$line"; done
-              echo ""
-            fi
-            if [ $STALE_COUNT -eq 0 ] && [ $UNREACHABLE_COUNT -eq 0 ]; then
-              echo "All ${#SLUGS[@]} tenants returned matching SHA. ✓"
-            fi
-          } >> "$GITHUB_STEP_SUMMARY"
-
-          if [ $UNREACHABLE_COUNT -gt 0 ]; then
-            echo "::warning::$UNREACHABLE_COUNT tenant(s) unreachable post-redeploy. Likely benign teardown race — CP healthz monitor catches real outages."
-          fi
-
-          # Belt-and-suspenders sanity floor: same logic as the staging
-          # variant — see that file's comment for the full rationale.
-          # Floor only applies when fleet >= 4; below that, canary-verify
-          # is the actual gate.
-          TOTAL_VERIFIED=${#SLUGS[@]}
-          if [ $TOTAL_VERIFIED -ge 4 ] && [ $UNREACHABLE_COUNT -gt $((TOTAL_VERIFIED / 2)) ]; then
-            echo "::error::$UNREACHABLE_COUNT of $TOTAL_VERIFIED tenant(s) unreachable — exceeds 50% threshold on a fleet large enough that this signals a real outage, not teardown race."
-            exit 1
-          fi
-
-          if [ $STALE_COUNT -gt 0 ]; then
-            echo "::error::$STALE_COUNT tenant(s) returned a stale SHA. ssm_status=Success was misleading — see job summary."
-            exit 1
-          fi
-
-          echo "::notice::Tenant fleet redeploy complete — all reachable tenants on ${EXPECTED_SHA:0:7} (${UNREACHABLE_COUNT} unreachable, soft-warned)."
@@ -1,362 +0,0 @@
-name: redeploy-tenants-on-staging
-
-# Auto-refresh staging tenant EC2s after every staging-branch merge.
-#
-# Mirror of redeploy-tenants-on-main.yml, with the staging-CP host and
-# the :staging-latest tag. Sister workflow exists for prod (rolls
-# :latest after canary-verify). Both share the same shape — just
-# different CP_URL + target_tag + admin token secret.
-#
-# Why this workflow exists: publish-workspace-server-image now builds
-# on every staging-branch push (PR #2335), pushing
-# platform-tenant:staging-latest to GHCR. Existing tenants pulled
-# their image once at boot and never re-pull, so the new image just
-# sits unused until the tenant is reprovisioned.
-#
-# This workflow closes the gap by calling staging-CP's
-# /cp/admin/tenants/redeploy-fleet, which performs a canary-first,
-# batched, health-gated SSM redeploy across every live staging tenant.
-# Same endpoint shape as prod CP — only the host differs.
-#
-# Runtime ordering:
-#   1. publish-workspace-server-image completes on staging branch →
-#      new :staging-latest in GHCR.
-#   2. This workflow fires via workflow_run, waits 30s for GHCR's CDN
-#      to propagate the new tag.
-#   3. Calls redeploy-fleet with no canary (staging IS canary; we don't
-#      need a sub-canary inside it). Soak still applies to the first
-#      tenant in case of bad-deploy detection.
-#   4. Any failure aborts the rollout and leaves older tenants on the
-#      prior image — safer default than half-and-half state.
-#
-# Rollback path: re-run with workflow_dispatch + target_tag=staging-<sha>
-# of a known-good build.
-
-on:
-  workflow_run:
-    workflows: ['publish-workspace-server-image']
-    types: [completed]
-    branches: [main]
-  workflow_dispatch:
-    inputs:
-      target_tag:
-        description: 'Tenant image tag to deploy (e.g. "staging-latest" or "staging-a59f1a6c"). Defaults to staging-latest when empty.'
-        required: false
-        type: string
-        default: 'staging-latest'
-      canary_slug:
-        description: 'Tenant slug to deploy first + soak (empty = skip canary, fan out immediately). Default empty for staging since staging itself is the canary.'
-        required: false
-        type: string
-        default: ''
-      soak_seconds:
-        description: 'Seconds to wait after canary before fanning out. Only meaningful if canary_slug is set.'
-        required: false
-        type: string
-        default: '60'
-      batch_size:
-        description: 'How many tenants SSM redeploys in parallel per batch.'
-        required: false
-        type: string
-        default: '3'
-      dry_run:
-        description: 'Plan only — do not actually redeploy.'
-        required: false
-        type: boolean
-        default: false
-
-permissions:
-  contents: read
-  # No write scopes needed — the workflow hits an external CP endpoint,
-  # not the GitHub API.
-
-# Serialize per-branch so two rapid staging pushes' redeploys don't
-# overlap and cause confusing per-tenant SSM state. cancel-in-progress
-# is false because aborting a half-rolled-out fleet leaves tenants
-# stuck on whatever image they happened to be on when cancelled.
-concurrency:
-  group: redeploy-tenants-on-staging
-  cancel-in-progress: false
-
-jobs:
-  redeploy:
-    # Skip the auto-trigger if publish-workspace-server-image didn't
-    # actually succeed. workflow_run fires on any completion state; we
-    # don't want to redeploy against a half-built image.
-    if: |
-      github.event_name == 'workflow_dispatch' ||
-      (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success')
-    runs-on: ubuntu-latest
-    timeout-minutes: 25
-    steps:
-      - name: Wait for GHCR tag propagation
-        # GHCR's edge cache takes ~15-30s to consistently serve the new
-        # :staging-latest manifest after the registry accepts the push.
-        # Same rationale as redeploy-tenants-on-main.yml.
-        run: sleep 30
-
-      - name: Call staging-CP redeploy-fleet
-        # CP_STAGING_ADMIN_API_TOKEN must be set as a repo/org secret
-        # on molecule-ai/molecule-core, matching staging-CP's
-        # CP_ADMIN_API_TOKEN env var (visible in Railway controlplane
-        # / staging environment). Stored separately from the prod
-        # CP_ADMIN_API_TOKEN so a leak of one doesn't auth the other.
-        env:
-          CP_URL: ${{ vars.STAGING_CP_URL || 'https://staging-api.moleculesai.app' }}
-          CP_STAGING_ADMIN_API_TOKEN: ${{ secrets.CP_STAGING_ADMIN_API_TOKEN }}
-          TARGET_TAG: ${{ inputs.target_tag || 'staging-latest' }}
-          CANARY_SLUG: ${{ inputs.canary_slug || '' }}
-          SOAK_SECONDS: ${{ inputs.soak_seconds || '60' }}
-          BATCH_SIZE: ${{ inputs.batch_size || '3' }}
-          DRY_RUN: ${{ inputs.dry_run || false }}
-        run: |
-          set -euo pipefail
-
-          # Schedule-vs-dispatch hardening (mirrors sweep-cf-orphans
-          # and sweep-cf-tunnels): hard-fail on auto-trigger when the
-          # secret is missing so a misconfigured-repo doesn't silently
-          # serve stale staging tenants. Soft-skip on operator dispatch.
-          if [ -z "${CP_STAGING_ADMIN_API_TOKEN:-}" ]; then
-            if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
-              echo "::warning::CP_STAGING_ADMIN_API_TOKEN secret not set — skipping redeploy"
-              echo "::warning::Set CP_STAGING_ADMIN_API_TOKEN in repo secrets to enable auto-redeploy."
-              echo "::notice::Pull the value from staging-CP's CP_ADMIN_API_TOKEN env in Railway."
-              exit 0
-            fi
-            echo "::error::staging redeploy cannot run — CP_STAGING_ADMIN_API_TOKEN secret missing"
-            echo "::error::set it at Settings → Secrets and Variables → Actions; pull from staging-CP's CP_ADMIN_API_TOKEN env in Railway."
-            exit 1
-          fi
-
-          BODY=$(jq -nc \
-            --arg tag "$TARGET_TAG" \
-            --arg canary "$CANARY_SLUG" \
-            --argjson soak "$SOAK_SECONDS" \
-            --argjson batch "$BATCH_SIZE" \
-            --argjson dry "$DRY_RUN" \
-            '{
-              target_tag: $tag,
-              canary_slug: $canary,
-              soak_seconds: $soak,
-              batch_size: $batch,
-              dry_run: $dry
-            }')
-
-          echo "POST $CP_URL/cp/admin/tenants/redeploy-fleet"
-          echo "  body: $BODY"
-
-          HTTP_RESPONSE=$(mktemp)
-          HTTP_CODE_FILE=$(mktemp)
-          # Route -w into its own tempfile so curl's exit code (e.g. 56
-          # on connection-reset) can't pollute the captured stdout. The
-          # previous inline-substitution shape produced "000000" on
-          # connection reset — caught on main variant 2026-05-04
-          # redeploying sha 2b862f6. Same fix shape as the synth-E2E
-          # §9c gate (PR #2797). See lint-curl-status-capture.yml for
-          # the CI gate that pins this fix shape.
-          set +e
-          curl -sS -o "$HTTP_RESPONSE" -w '%{http_code}' \
-            -m 1200 \
-            -H "Authorization: Bearer $CP_STAGING_ADMIN_API_TOKEN" \
-            -H "Content-Type: application/json" \
-            -X POST "$CP_URL/cp/admin/tenants/redeploy-fleet" \
-            -d "$BODY" >"$HTTP_CODE_FILE"
-          set -e
-          # Stderr from curl (-sS shows dial errors etc.) goes to the
-          # runner log so operators can see WHY a connection failed.
-          HTTP_CODE=$(cat "$HTTP_CODE_FILE" 2>/dev/null || echo "000")
-          [ -z "$HTTP_CODE" ] && HTTP_CODE="000"
-
-          echo "HTTP $HTTP_CODE"
-          cat "$HTTP_RESPONSE" | jq . || cat "$HTTP_RESPONSE"
-
-          {
-            echo "## Staging tenant redeploy fleet"
-            echo ""
-            echo "**Target tag:** \`$TARGET_TAG\`"
-            echo "**Canary:** \`${CANARY_SLUG:-(none — staging is itself the canary)}\` (soak ${SOAK_SECONDS}s)"
-            echo "**Batch size:** $BATCH_SIZE"
-            echo "**Dry run:** $DRY_RUN"
-            echo "**HTTP:** $HTTP_CODE"
-            echo ""
-            echo "### Per-tenant result"
-            echo ""
-            echo '| Slug | Phase | SSM Status | Exit | Healthz | Error |'
-            echo '|------|-------|------------|------|---------|-------|'
-            jq -r '.results[]? | "| \(.slug) | \(.phase) | \(.ssm_status // "-") | \(.ssm_exit_code) | \(.healthz_ok) | \(.error // "-") |"' "$HTTP_RESPONSE" || true
-          } >> "$GITHUB_STEP_SUMMARY"
-
-          # Distinguish "real fleet failure" from "E2E teardown race".
-          #
-          # CP returns HTTP 500 + ok=false whenever ANY tenant in the
-          # fleet failed SSM or healthz. In practice the recurring source
-          # of these is ephemeral test tenants being torn down by their
-          # parent E2E run mid-redeploy: the EC2 dies → SSM exit=2 or
-          # healthz timeout → CP marks the fleet failed → this workflow
-          # goes red even though every operator-facing tenant rolled fine.
-          #
-          # Ephemeral slug prefixes (kept in sync with sweep-stale-e2e-orgs.yml
-          # — see that file for the source-of-truth list and rationale):
-          #   - e2e-*       — canvas/saas/ext E2E suites
-          #   - rt-e2e-*    — runtime-test harness fixtures (RFC #2251)
-          # Long-lived prefixes that are NOT ephemeral and MUST hard-fail:
-          # demo-prep, dryrun-*, dryrun2-*, plus all human tenant slugs.
-          #
-          # Filter: if HTTP=500/ok=false AND every failed slug matches an
-          # ephemeral prefix, treat as soft-warn and let the verify step
-          # downstream handle unreachable-vs-stale (#2402). Any non-ephemeral
-          # failure or a non-500 HTTP response remains a hard failure.
-          OK=$(jq -r '.ok // "false"' "$HTTP_RESPONSE")
-          FAILED_SLUGS=$(jq -r '
-            .results[]?
-            | select((.healthz_ok != true) or (.ssm_status != "Success"))
-            | .slug' "$HTTP_RESPONSE" 2>/dev/null || true)
-          EPHEMERAL_PREFIX_RE='^(e2e-|rt-e2e-)'
-          NON_EPHEMERAL_FAILED=$(printf '%s\n' "$FAILED_SLUGS" | grep -v '^$' | grep -Ev "$EPHEMERAL_PREFIX_RE" || true)
-
-          if [ "$HTTP_CODE" = "200" ] && [ "$OK" = "true" ]; then
-            : # happy path — fall through to verification
-          elif [ "$HTTP_CODE" = "500" ] && [ -z "$NON_EPHEMERAL_FAILED" ] && [ -n "$FAILED_SLUGS" ]; then
-            COUNT=$(printf '%s\n' "$FAILED_SLUGS" | grep -Ec "$EPHEMERAL_PREFIX_RE" || true)
-            echo "::warning::redeploy-fleet returned HTTP 500 but every failed tenant ($COUNT) is ephemeral (e2e-*/rt-e2e-*) — treating as teardown race, soft-warning."
-            printf '%s\n' "$FAILED_SLUGS" | sed 's/^/::warning::  failed: /'
-          elif [ "$HTTP_CODE" != "200" ]; then
-            echo "::error::redeploy-fleet returned HTTP $HTTP_CODE"
-            if [ -n "$NON_EPHEMERAL_FAILED" ]; then
-              echo "::error::non-ephemeral tenant(s) failed:"
-              printf '%s\n' "$NON_EPHEMERAL_FAILED" | sed 's/^/::error::  /'
-            fi
-            exit 1
-          else
-            # HTTP=200 but ok=false (shouldn't happen with current CP
-            # but keep the gate for completeness).
-            echo "::error::redeploy-fleet reported ok=false (see summary for which tenant halted the rollout)"
-            exit 1
-          fi
-          echo "::notice::Staging tenant fleet redeploy reported ssm_status=Success — verifying actual image roll on each tenant..."
-
-          cp "$HTTP_RESPONSE" "$RUNNER_TEMP/redeploy-response.json"
-
-      - name: Verify each staging tenant /buildinfo matches published SHA
-        # Mirror of the verify step in redeploy-tenants-on-main.yml — see
-        # there for the rationale (#2395 root fix). Staging has the same
-        # ssm_status-success-but-stale-image hazard and benefits from the
-        # same gate. Diff: TENANT_DOMAIN includes the `staging.` infix.
-        env:
-          EXPECTED_SHA: ${{ github.event.workflow_run.head_sha || github.sha }}
-          TARGET_TAG: ${{ inputs.target_tag || 'staging-latest' }}
-          TENANT_DOMAIN: 'staging.moleculesai.app'
-        run: |
-          set -euo pipefail
-
-          # staging-latest is the staging-side moving tag; treat it the
-          # same way main treats `latest`. Operator-pinned SHAs skip
-          # verification (see main variant for why).
-          if [ "$TARGET_TAG" != "staging-latest" ] && [ "$TARGET_TAG" != "latest" ] && [ "$TARGET_TAG" != "$EXPECTED_SHA" ]; then
-            echo "::notice::target_tag=$TARGET_TAG (operator-pinned) — skipping per-tenant SHA verification."
-            exit 0
-          fi
-
-          RESP="$RUNNER_TEMP/redeploy-response.json"
-          if [ ! -s "$RESP" ]; then
-            echo "::error::redeploy-response.json missing or empty"
-            exit 1
-          fi
-
-          mapfile -t SLUGS < <(jq -r '.results[]? | select(.healthz_ok == true) | .slug' "$RESP")
-          if [ ${#SLUGS[@]} -eq 0 ]; then
-            echo "::warning::No staging tenants reported healthz_ok — nothing to verify"
-            exit 0
-          fi
-
-          echo "Verifying ${#SLUGS[@]} staging tenant(s) against EXPECTED_SHA=${EXPECTED_SHA:0:7}..."
-
-          # Two distinct failure modes here:
-          #   STALE_COUNT      — tenant returned a SHA that doesn't match. THIS is
-          #                      the #2395 bug class: tenant up + serving old code.
-          #                      Always hard-fail the workflow.
-          #   UNREACHABLE_COUNT — tenant didn't respond. Almost always a benign
-          #                      teardown race: redeploy-fleet snapshot says
-          #                      healthz_ok=true, then the E2E suite tears the
-          #                      ephemeral tenant down before this step runs (the
-          #                      e2e-* fixtures churn 5-10/hour on staging). Soft-
-          #                      warn so we don't block staging→main on cleanup.
-          #                      Real "tenant up but unreachable" is caught by CP's
-          #                      own healthz monitor + the post-redeploy alert; we
-          #                      don't need to double-count it here.
-          STALE_COUNT=0
-          UNREACHABLE_COUNT=0
-          STALE_LINES=()
-          UNREACHABLE_LINES=()
-          for slug in "${SLUGS[@]}"; do
-            URL="https://${slug}.${TENANT_DOMAIN}/buildinfo"
-            BODY=$(curl -sS --max-time 30 --retry 3 --retry-delay 5 --retry-connrefused "$URL" || true)
-            ACTUAL_SHA=$(echo "$BODY" | jq -r '.git_sha // ""' 2>/dev/null || echo "")
-            if [ -z "$ACTUAL_SHA" ]; then
-              UNREACHABLE_COUNT=$((UNREACHABLE_COUNT + 1))
-              UNREACHABLE_LINES+=("| $slug | (no /buildinfo response) | ${EXPECTED_SHA:0:7} | ⚠ unreachable (likely teardown race) |")
-              continue
-            fi
-            if [ "$ACTUAL_SHA" = "$EXPECTED_SHA" ]; then
-              echo "  $slug: ${ACTUAL_SHA:0:7} ✓"
-            else
-              STALE_COUNT=$((STALE_COUNT + 1))
-              STALE_LINES+=("| $slug | ${ACTUAL_SHA:0:7} | ${EXPECTED_SHA:0:7} | ❌ stale |")
-            fi
-          done
-
-          {
-            echo ""
-            echo "### Per-tenant /buildinfo verification (staging)"
-            echo ""
-            echo "Expected SHA: \`${EXPECTED_SHA:0:7}\`"
-            echo ""
-            if [ $STALE_COUNT -gt 0 ]; then
-              echo "**${STALE_COUNT} STALE tenant(s) — these did NOT pick up the new image despite ssm_status=Success:**"
-              echo ""
-              echo "| Slug | Actual /buildinfo SHA | Expected | Status |"
-              echo "|------|----------------------|----------|--------|"
-              for line in "${STALE_LINES[@]}"; do echo "$line"; done
-              echo ""
-            fi
-            if [ $UNREACHABLE_COUNT -gt 0 ]; then
-              echo "**${UNREACHABLE_COUNT} unreachable tenant(s) — likely E2E teardown race (soft-warn, not failing):**"
-              echo ""
-              echo "| Slug | Actual /buildinfo SHA | Expected | Status |"
-              echo "|------|----------------------|----------|--------|"
-              for line in "${UNREACHABLE_LINES[@]}"; do echo "$line"; done
-              echo ""
-            fi
-            if [ $STALE_COUNT -eq 0 ] && [ $UNREACHABLE_COUNT -eq 0 ]; then
-              echo "All ${#SLUGS[@]} staging tenants returned matching SHA. ✓"
-            fi
-          } >> "$GITHUB_STEP_SUMMARY"
-
-          if [ $UNREACHABLE_COUNT -gt 0 ]; then
-            echo "::warning::$UNREACHABLE_COUNT staging tenant(s) unreachable post-redeploy. Likely benign teardown race — CP healthz monitor catches real outages."
-          fi
-
-          # Belt-and-suspenders sanity floor: if MORE than half the fleet is
-          # unreachable AND the fleet is large enough that "half down" is
-          # statistically meaningful, this is a real outage (e.g. new image
-          # crashes on startup), not a teardown race. Hard-fail.
-          #
-          # Floor only applies when TOTAL_VERIFIED >= 4 — below that, the
-          # canary-verify step is the actual gate for "all tenants down"
-          # detection (it runs against the canary first and aborts the
-          # rollout if the canary fails to come up). Without the >=4 gate,
-          # a 1-tenant fleet (e.g. a single ephemeral e2e-* tenant on a
-          # quiet staging push) would re-flake on the exact teardown-race
-          # condition #2402 fixed: 1 of 1 unreachable = 100% > 50% → fail.
-          TOTAL_VERIFIED=${#SLUGS[@]}
-          if [ $TOTAL_VERIFIED -ge 4 ] && [ $UNREACHABLE_COUNT -gt $((TOTAL_VERIFIED / 2)) ]; then
-            echo "::error::$UNREACHABLE_COUNT of $TOTAL_VERIFIED staging tenant(s) unreachable — exceeds 50% threshold on a fleet large enough that this signals a real outage, not teardown race."
-            exit 1
-          fi
-
-          if [ $STALE_COUNT -gt 0 ]; then
-            echo "::error::$STALE_COUNT staging tenant(s) returned a stale SHA. ssm_status=Success was misleading — see job summary."
-            exit 1
-          fi
-
-          echo "::notice::Staging tenant fleet redeploy complete — all reachable tenants on ${EXPECTED_SHA:0:7} (${UNREACHABLE_COUNT} unreachable, soft-warned)."
@@ -57,24 +57,24 @@ See `CLAUDE.md` for a full list of environment variables and their purposes.

 This repo is scoped to **code** (canvas, workspace, workspace-server, related
 infra). Public content (blog posts, marketing copy, OG images, SEO briefs,
-DevRel demos) lives in [`Molecule-AI/docs`](https://git.moleculesai.app/molecule-ai/docs).
+DevRel demos) lives in [`molecule-ai/docs`](https://git.moleculesai.app/molecule-ai/docs).
 The `Block forbidden paths` CI gate fails any PR that writes to `marketing/`
-or other removed paths — open against `Molecule-AI/docs` instead.
+or other removed paths — open against `molecule-ai/docs` instead.

 | Content type | Target |
 |---|---|
-| Blog posts | `Molecule-AI/docs` → `content/blog/<YYYY-MM-DD-slug>/` |
-| Doc pages | `Molecule-AI/docs` → `content/docs/` |
-| Marketing copy / PMM positioning | `Molecule-AI/docs` → `marketing/` |
-| OG images, visual assets | `Molecule-AI/docs` → `app/` or `marketing/` |
-| SEO briefs | `Molecule-AI/docs` → `marketing/` |
-| DevRel demos (runnable code) | Standalone repo under `Molecule-AI/`, OR embedded in `Molecule-AI/docs` |
+| Blog posts | `molecule-ai/docs` → `content/blog/<YYYY-MM-DD-slug>/` |
+| Doc pages | `molecule-ai/docs` → `content/docs/` |
+| Marketing copy / PMM positioning | `molecule-ai/docs` → `marketing/` |
+| OG images, visual assets | `molecule-ai/docs` → `app/` or `marketing/` |
+| SEO briefs | `molecule-ai/docs` → `marketing/` |
+| DevRel demos (runnable code) | Standalone repo under `molecule-ai/`, OR embedded in `molecule-ai/docs` |
 | Launch checklists, internal tracking | GitHub Issues — **not** committed files |
 | Engineering docs (`docs/adr/`, `docs/architecture/`, `docs/incidents/`) | This repo (internal, not published) |
 | Live product pages (e.g. `canvas/src/app/pricing/page.tsx`) | This repo (these are app code, not marketing copy) |

 If a PR fails the `Block forbidden paths` check, the contents belong in
-`Molecule-AI/docs`. No CI drag, no Canvas E2E, content lands in minutes.
+`molecule-ai/docs`. No CI drag, no Canvas E2E, content lands in minutes.

 ## Development Workflow

@@ -190,9 +190,9 @@ Runs the full regression suite against a fixture HTTP server. No network access
 Code in this repo lands in molecule-core. Some related runtime artifacts
 live in their own repos:

- [`Molecule-AI/molecule-ai-workspace-runtime`](https://git.moleculesai.app/molecule-ai/molecule-ai-workspace-runtime) — Python adapter SDK (`molecule_runtime`) that runs inside containerized Molecule workspaces. Bridges Claude Code SDK / hermes / langgraph / etc. → A2A queue.
- [`Molecule-AI/molecule-sdk-python`](https://git.moleculesai.app/molecule-ai/molecule-sdk-python) — `A2AServer` + `RemoteAgentClient` for external agents that register over the public `/registry/register` flow.
- [`Molecule-AI/molecule-mcp-claude-channel`](https://git.moleculesai.app/molecule-ai/molecule-mcp-claude-channel) — Claude Code channel plugin. Bridges A2A traffic into a running Claude Code session via MCP `notifications/claude/channel`. Polling-based (no tunnel required); install with `claude --channels plugin:molecule@Molecule-AI/molecule-mcp-claude-channel`.
+- [`molecule-ai/molecule-ai-workspace-runtime`](https://git.moleculesai.app/molecule-ai/molecule-ai-workspace-runtime) — Python adapter SDK (`molecule_runtime`) that runs inside containerized Molecule workspaces. Bridges Claude Code SDK / hermes / langgraph / etc. → A2A queue.
+- [`molecule-ai/molecule-sdk-python`](https://git.moleculesai.app/molecule-ai/molecule-sdk-python) — `A2AServer` + `RemoteAgentClient` for external agents that register over the public `/registry/register` flow.
+- [`molecule-ai/molecule-mcp-claude-channel`](https://git.moleculesai.app/molecule-ai/molecule-mcp-claude-channel) — Claude Code channel plugin. Bridges A2A traffic into a running Claude Code session via MCP `notifications/claude/channel`. Polling-based (no tunnel required); install inside Claude Code via `/plugin marketplace add https://git.moleculesai.app/molecule-ai/molecule-mcp-claude-channel.git` → `/plugin install molecule@molecule-channel`, then launch with `claude --dangerously-load-development-channels --channels plugin:molecule@molecule-channel`.

 When extending the **A2A surface** in molecule-core (`workspace-server/internal/handlers/a2a_proxy.go` etc.), consider whether the change has a downstream impact on the runtime SDK or the channel plugin — they're versioned independently but share the wire shape.

@@ -4,10 +4,10 @@
 # use this Makefile; CI calls docker compose / go test directly so the
 # Makefile can evolve without breaking the build.

-.PHONY: help dev up down logs build test
+.PHONY: help dev up down logs build test e2e-peer-visibility

 help: ## Show this help.
-	@grep -E '^[a-zA-Z_-]+:.*?## ' $(MAKEFILE_LIST) | awk 'BEGIN {FS = ":.*?## "}; {printf "\033[36m%-12s\033[0m %s\n", $$1, $$2}'
+	@grep -E '^[a-zA-Z0-9_-]+:.*?## ' $(MAKEFILE_LIST) | awk 'BEGIN {FS = ":.*?## "}; {printf "\033[36m%-22s\033[0m %s\n", $$1, $$2}'

 dev: ## Start the full stack with air hot-reload for the platform service.
 	docker compose -f docker-compose.yml -f docker-compose.dev.yml up
@@ -26,3 +26,13 @@ build: ## Force a fresh build of the platform image (no cache).

 test: ## Run Go unit tests in workspace-server/.
 	cd workspace-server && go test -race ./...
+
+# ─── Local prod-mimic E2E gates ────────────────────────────────────────
+# Run the LITERAL peer-visibility MCP list_peers gate against the
+# already-running local stack (`make up` or `make dev`). Same byte-
+# identical assertion as the staging gate — only provisioning differs.
+# Skips any runtime whose provider key is absent (partially-keyed env
+# is fine). See tests/e2e/test_peer_visibility_mcp_local.sh for the
+# env contract (CLAUDE_CODE_OAUTH_TOKEN / E2E_MINIMAX_API_KEY / etc).
+e2e-peer-visibility: ## Run the LOCAL peer-visibility MCP gate vs the running stack (needs `make up` first).
+	bash tests/e2e/test_peer_visibility_mcp_local.sh
@@ -238,7 +238,7 @@ The result is not just “an agent that learns.” It is **an organization that
 - subscribe to one or more workspaces; peer messages surface as conversation turns; replies route back through Molecule's A2A
 - no tunnel, no public endpoint — the plugin self-registers each watched workspace as `delivery_mode=poll` and long-polls `/activity?since_id=…`
 - multi-tenant friendly: one plugin install can watch workspaces across multiple Molecule tenants (`MOLECULE_PLATFORM_URLS` per-workspace)
- install via the standard marketplace flow: `/plugin marketplace add Molecule-AI/molecule-mcp-claude-channel` → `/plugin install molecule-channel@molecule-mcp-claude-channel`
+- install via the standard marketplace flow: `/plugin marketplace add https://git.moleculesai.app/molecule-ai/molecule-mcp-claude-channel.git` → `/plugin install molecule@molecule-channel`, then launch with `claude --dangerously-load-development-channels --channels plugin:molecule@molecule-channel`

 ## Built For Teams That Need More Than A Demo

@@ -237,7 +237,7 @@ Molecule AI 并不是要替代下面这些 framework，而是把它们纳入更
 - 订阅一个或多个 workspace；peer 的消息会以 user-turn 出现，回复会经 Molecule A2A 路由出去
 - 无需公网隧道、无需公开端点 —— 插件启动时自动把每个 watched workspace 注册成 `delivery_mode=poll`，长轮询 `/activity?since_id=…`
 - 多租户友好：单次安装即可同时 watch 跨多个 Molecule 租户的 workspace（`MOLECULE_PLATFORM_URLS` 按 workspace 配置）
- 通过标准 marketplace 流程安装：`/plugin marketplace add Molecule-AI/molecule-mcp-claude-channel` → `/plugin install molecule-channel@molecule-mcp-claude-channel`
+- 通过标准 marketplace 流程安装：`/plugin marketplace add https://git.moleculesai.app/molecule-ai/molecule-mcp-claude-channel.git` → `/plugin install molecule@molecule-channel`，然后用 `claude --dangerously-load-development-channels --channels plugin:molecule@molecule-channel` 启动

 ## 适合什么团队

@@ -0,0 +1,113 @@
+import { describe, it, expect, vi } from "vitest";
+
+// Marketing-launch SEO (mc#1486). These tests pin the public crawler
+// contract: anything that flips public marketing routes to disallow,
+// drops the sitemap from robots.txt, or removes the OG image
+// reference from root metadata should fail loudly here.
+
+// next/font and the rest of the layout's runtime tree are not
+// vitest-compatible (next/font expects the Next.js compiler swc
+// transform). We import layout.tsx only for its exported `metadata`
+// constant — mock the font module to a constructor-returning stub.
+vi.mock("next/font/google", () => ({
+  Inter: () => ({ variable: "--font-inter" }),
+  JetBrains_Mono: () => ({ variable: "--font-jetbrains" }),
+}));
+
+import robots from "../robots";
+import sitemap from "../sitemap";
+import { metadata } from "../layout";
+
+describe("robots.ts", () => {
+  it("allows public marketing routes and blocks authed/app routes", () => {
+    const r = robots();
+    expect(r.rules).toBeDefined();
+    const rule = Array.isArray(r.rules) ? r.rules[0] : r.rules!;
+    expect(rule.userAgent).toBe("*");
+    const allow = Array.isArray(rule.allow) ? rule.allow : [rule.allow];
+    expect(allow).toEqual(expect.arrayContaining(["/", "/pricing", "/blog"]));
+    const disallow = Array.isArray(rule.disallow)
+      ? rule.disallow
+      : [rule.disallow];
+    expect(disallow).toEqual(
+      expect.arrayContaining(["/api/", "/orgs", "/cp/"]),
+    );
+  });
+
+  it("declares the sitemap URL", () => {
+    const r = robots();
+    expect(r.sitemap).toMatch(/\/sitemap\.xml$/);
+  });
+
+  it("declares a canonical host", () => {
+    const r = robots();
+    expect(r.host).toMatch(/^https:\/\//);
+  });
+});
+
+describe("sitemap.ts", () => {
+  it("includes apex, pricing, and the live blog post", () => {
+    const entries = sitemap();
+    const urls = entries.map((e) => e.url);
+    expect(urls.some((u) => u.endsWith("/"))).toBe(true);
+    expect(urls.some((u) => u.endsWith("/pricing"))).toBe(true);
+    expect(
+      urls.some((u) => u.includes("/blog/2026-04-20-chrome-devtools-mcp")),
+    ).toBe(true);
+  });
+
+  it("does NOT include authed/app routes", () => {
+    const entries = sitemap();
+    const urls = entries.map((e) => e.url);
+    expect(urls.some((u) => u.includes("/orgs"))).toBe(false);
+    expect(urls.some((u) => u.includes("/api/"))).toBe(false);
+  });
+
+  it("sets a non-zero priority and a valid changeFrequency on every entry", () => {
+    const valid = new Set([
+      "always",
+      "hourly",
+      "daily",
+      "weekly",
+      "monthly",
+      "yearly",
+      "never",
+    ]);
+    for (const e of sitemap()) {
+      expect(e.priority).toBeGreaterThan(0);
+      expect(valid.has(String(e.changeFrequency))).toBe(true);
+    }
+  });
+});
+
+describe("root layout metadata", () => {
+  it("sets a templated title + non-empty description", () => {
+    const t = metadata.title as { default: string; template: string };
+    expect(t.default).toMatch(/Molecule AI/);
+    expect(t.template).toMatch(/%s/);
+    expect((metadata.description ?? "").length).toBeGreaterThan(50);
+  });
+
+  it("declares OG + Twitter text fields (image comes from opengraph-image.tsx)", () => {
+    const og = metadata.openGraph;
+    expect(og).toBeDefined();
+    expect((og as { title: string }).title).toMatch(/Molecule AI/);
+    expect((og as { description: string }).description.length).toBeGreaterThan(
+      50,
+    );
+    const tw = metadata.twitter;
+    expect(tw).toBeDefined();
+    // Next.js typings narrow twitter.card to a union — assert via cast.
+    expect((tw as { card: string }).card).toBe("summary_large_image");
+  });
+
+  it("sets a canonical alternate", () => {
+    expect(metadata.alternates?.canonical).toBe("/");
+  });
+
+  it("enables indexing at the metadata level (robots.ts owns per-route)", () => {
+    const r = metadata.robots as { index: boolean; follow: boolean };
+    expect(r.index).toBe(true);
+    expect(r.follow).toBe(true);
+  });
+});
@@ -27,9 +27,78 @@ import {
  themeBootScript,
 } from "@/lib/theme-cookie";

+// Marketing-launch SEO (mc#1486). Canonical apex is app.moleculesai.app —
+// tenant subdomains (<slug>.moleculesai.app) reuse the same Next.js build
+// but are gated behind auth (AuthGate redirects anonymous → /cp/auth/login)
+// and are de-indexed in robots.ts. The metadata here applies to the
+// public marketing surface served from the apex host.
+//
+// Override per-route by exporting a page-level `metadata`/`generateMetadata`
+// — Next.js merges page metadata over layout metadata using
+// `title.template` for "<page> | Molecule AI" composition.
+const SITE_URL =
+  process.env.NEXT_PUBLIC_SITE_URL ?? "https://app.moleculesai.app";
+
 export const metadata: Metadata = {
-  title: "Molecule AI",
-  description: "AI Org Chart Canvas",
+  metadataBase: new URL(SITE_URL),
+  title: {
+    default: "Molecule AI — the AI org chart canvas",
+    template: "%s | Molecule AI",
+  },
+  description:
+    "Molecule AI is an org-chart canvas for AI agent teams. Wire Claude Code, Codex, Hermes, and OpenClaw agents into a governed multi-agent workspace with credit metering, audit, and one-click runtime provisioning.",
+  applicationName: "Molecule AI",
+  keywords: [
+    "AI agents",
+    "multi-agent",
+    "agent orchestration",
+    "AI org chart",
+    "Claude Code",
+    "Codex",
+    "MCP",
+    "agent governance",
+    "A2A",
+    "agent runtime",
+  ],
+  authors: [{ name: "Molecule AI" }],
+  creator: "Molecule AI",
+  publisher: "Molecule AI",
+  alternates: { canonical: "/" },
+  // OG + Twitter images come from the file-convention sibling
+  // `opengraph-image.tsx` — Next.js auto-attaches them to og:image
+  // and twitter:image when present at the segment root. We keep the
+  // text fields here so they win over per-page metadata when a page
+  // doesn't override them. `images: []` as the structural fallback
+  // for hosts that won't follow the file convention; the real URL
+  // is injected by Next.js at build time from opengraph-image.tsx.
+  openGraph: {
+    type: "website",
+    siteName: "Molecule AI",
+    url: SITE_URL,
+    title: "Molecule AI — the AI org chart canvas",
+    description:
+      "Wire Claude Code, Codex, Hermes, and OpenClaw agents into a governed multi-agent workspace. Credit metering, audit, and one-click runtime provisioning.",
+    locale: "en_US",
+  },
+  twitter: {
+    card: "summary_large_image",
+    title: "Molecule AI — the AI org chart canvas",
+    description:
+      "Wire Claude Code, Codex, Hermes, and OpenClaw agents into a governed multi-agent workspace.",
+  },
+  icons: {
+    icon: "/molecule-icon.png",
+    apple: "/molecule-icon.png",
+  },
+  // robots.ts owns the per-route allow/disallow contract; this is the
+  // header-level fallback for routes the crawler reaches before
+  // robots.txt resolves. Default = index public marketing routes;
+  // app/auth/api/orgs are noindex'd by robots.ts.
+  robots: {
+    index: true,
+    follow: true,
+    googleBot: { index: true, follow: true, "max-image-preview": "large" },
+  },
 };

 export default async function RootLayout({
@@ -94,6 +163,75 @@ export default async function RootLayout({
          nonce={nonce}
          dangerouslySetInnerHTML={{ __html: themeBootScript }}
        />
+        {/*
+         * JSON-LD structured data (mc#1486). Two graph nodes:
+         *
+         *   - Organization: surfaces the brand to Google Knowledge
+         *     Graph + Bing entity index. URL+logo+sameAs are the
+         *     minimum recommended set for new brands without a
+         *     Wikipedia page.
+         *
+         *   - WebSite: enables the sitelinks search box and tells
+         *     crawlers the canonical site URL when the same content
+         *     is reachable via multiple subdomains (apex + tenant).
+         *
+         * Type-application/ld+json runs synchronously without
+         * executing JS, so 'strict-dynamic' isn't required — we still
+         * carry the nonce because production CSP's default-src 'self'
+         * applies to any <script> element. The "type" attribute is
+         * what keeps the browser from running the body as JS, but
+         * CSP nonces are gated on the element not the type, so we
+         * include the nonce too.
+         */}
+        <script
+          type="application/ld+json"
+          nonce={nonce}
+          dangerouslySetInnerHTML={{
+            __html: JSON.stringify({
+              "@context": "https://schema.org",
+              "@graph": [
+                {
+                  "@type": "Organization",
+                  "@id": `${SITE_URL}#organization`,
+                  name: "Molecule AI",
+                  url: SITE_URL,
+                  logo: `${SITE_URL}/molecule-icon.png`,
+                  sameAs: [
+                    "https://github.com/molecule-ai",
+                    "https://x.com/moleculeai",
+                  ],
+                },
+                {
+                  "@type": "WebSite",
+                  "@id": `${SITE_URL}#website`,
+                  url: SITE_URL,
+                  name: "Molecule AI",
+                  publisher: { "@id": `${SITE_URL}#organization` },
+                  inLanguage: "en-US",
+                },
+                {
+                  "@type": "SoftwareApplication",
+                  "@id": `${SITE_URL}#software`,
+                  name: "Molecule AI",
+                  applicationCategory: "DeveloperApplication",
+                  operatingSystem: "Web",
+                  description:
+                    "Org-chart canvas for AI agent teams with credit metering, audit, and one-click runtime provisioning.",
+                  url: SITE_URL,
+                  offers: {
+                    "@type": "AggregateOffer",
+                    priceCurrency: "USD",
+                    lowPrice: "0",
+                    highPrice: "99",
+                    offerCount: "3",
+                    url: `${SITE_URL}/pricing`,
+                  },
+                  publisher: { "@id": `${SITE_URL}#organization` },
+                },
+              ],
+            }),
+          }}
+        />
      </head>
      <body className={`bg-surface text-ink ${interFont.variable} ${monoFont.variable}`}>
        <ThemeProvider initialTheme={theme}>
@@ -0,0 +1,82 @@
+import { ImageResponse } from "next/og";
+
+// Marketing-launch SEO (mc#1486). Next.js App-Router file-system OG
+// convention: served as `/opengraph-image` and auto-attached as
+// `og:image` + `twitter:image`. Dynamic (not a static PNG in /public)
+// so we can iterate the brand mark + tagline pre-launch without
+// churning a binary blob in git history.
+export const runtime = "edge";
+
+export const alt = "Molecule AI — the AI org chart canvas";
+export const size = { width: 1200, height: 630 };
+export const contentType = "image/png";
+
+export default function OG() {
+  return new ImageResponse(
+    (
+      <div
+        style={{
+          width: "100%",
+          height: "100%",
+          display: "flex",
+          flexDirection: "column",
+          alignItems: "flex-start",
+          justifyContent: "center",
+          padding: "80px",
+          background:
+            "linear-gradient(135deg, #0a0a0a 0%, #1a1a2e 60%, #16213e 100%)",
+          color: "#ffffff",
+          fontFamily: "system-ui, -apple-system, sans-serif",
+        }}
+      >
+        <div
+          style={{
+            fontSize: 28,
+            color: "#a3a3c2",
+            letterSpacing: "0.18em",
+            textTransform: "uppercase",
+            marginBottom: 24,
+          }}
+        >
+          Molecule AI
+        </div>
+        <div
+          style={{
+            fontSize: 76,
+            fontWeight: 700,
+            lineHeight: 1.05,
+            letterSpacing: "-0.02em",
+            maxWidth: 980,
+          }}
+        >
+          The AI org chart canvas
+        </div>
+        <div
+          style={{
+            fontSize: 32,
+            color: "#c8c8d8",
+            marginTop: 32,
+            lineHeight: 1.3,
+            maxWidth: 980,
+          }}
+        >
+          Wire Claude Code, Codex, Hermes, and OpenClaw agents into a governed
+          multi-agent workspace.
+        </div>
+        <div
+          style={{
+            position: "absolute",
+            right: 80,
+            bottom: 80,
+            fontSize: 22,
+            color: "#7a7a96",
+            display: "flex",
+          }}
+        >
+          moleculesai.app
+        </div>
+      </div>
+    ),
+    { ...size },
+  );
+}
@@ -103,7 +103,7 @@ export default function Home() {
                setHydrationError(null);
                window.location.reload();
              }}
-              className="px-4 py-2 bg-accent-strong hover:bg-accent text-white rounded-md text-sm"
+              className="px-4 py-2 bg-accent-strong hover:bg-accent text-white rounded-md text-sm focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1"
            >
              Retry
            </button>
@@ -115,7 +115,9 @@ export default function Home() {

  return (
    <>
-      <Canvas />
+      <main aria-label="Agent canvas">
+        <Canvas />
+      </main>
      <Legend />
      <CommunicationOverlay />
      {hydrationError && (
@@ -134,7 +136,7 @@ export default function Home() {
              setHydrationError(null);
              window.location.reload();
            }}
-            className="px-4 py-2 bg-accent-strong hover:bg-accent text-white rounded-md text-sm"
+            className="px-4 py-2 bg-accent-strong hover:bg-accent text-white rounded-md text-sm focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1"
          >
            Retry
          </button>
@@ -176,7 +178,7 @@ brew services start redis`}</pre>
      </p>
      <button
        onClick={() => window.location.reload()}
-        className="px-4 py-2 bg-accent-strong hover:bg-accent text-white rounded-md text-sm mt-2"
+        className="px-4 py-2 bg-accent-strong hover:bg-accent text-white rounded-md text-sm mt-2 focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1"
      >
        Reload
      </button>
@@ -0,0 +1,45 @@
+import type { MetadataRoute } from "next";
+
+// Marketing-launch SEO (mc#1486). Next.js App-Router robots convention:
+// this file is served as `/robots.txt` at build time and is the single
+// source of truth for crawler allow/disallow.
+//
+// Contract:
+//   - Public marketing routes (/, /pricing, /blog/*) are crawlable.
+//   - Authed/app routes (/orgs, /api/*) are noindex'd. They render
+//     useful content only after a session round-trip, so a crawler hit
+//     just wastes our crawl budget and exposes endpoint shapes.
+//   - Tenant subdomains (<slug>.moleculesai.app) share this build but
+//     are blocked at the host level by the canvas middleware sending
+//     an `X-Robots-Tag: noindex` header — robots.txt is per-host and
+//     this file's `host` field claims the apex as canonical.
+//
+// Note: `sitemap` is published via the sibling `sitemap.ts` route; we
+// reference it explicitly here so crawlers don't have to guess.
+const SITE_URL =
+  process.env.NEXT_PUBLIC_SITE_URL ?? "https://app.moleculesai.app";
+
+export default function robots(): MetadataRoute.Robots {
+  return {
+    rules: [
+      {
+        userAgent: "*",
+        allow: ["/", "/pricing", "/blog"],
+        // Authed app surface + API + transient checkout returns. The
+        // /orgs route boots the org-selector behind AuthGate; even
+        // though SSR returns markup, that markup is a login wall when
+        // hit by an unauthenticated crawler, so indexing it dilutes
+        // brand searches with a "Please sign in" snippet.
+        disallow: [
+          "/orgs",
+          "/orgs/",
+          "/api/",
+          "/cp/",
+          "/checkout/",
+        ],
+      },
+    ],
+    sitemap: `${SITE_URL}/sitemap.xml`,
+    host: SITE_URL,
+  };
+}
@@ -0,0 +1,42 @@
+import type { MetadataRoute } from "next";
+
+// Marketing-launch SEO (mc#1486). App-Router sitemap convention: this
+// file is served as `/sitemap.xml` and enumerates the public marketing
+// surface for search crawlers + AI training pipelines.
+//
+// Scope deliberately narrow:
+//   - Apex landing, pricing, and the (currently single) blog post.
+//   - Authed app routes are excluded — they're disallowed in robots.ts
+//     and would appear as "Please sign in" wall to a crawler.
+//
+// `lastModified` uses a build-time timestamp rather than per-route
+// fs.stat so the same value applies regardless of where the build
+// runs (Vercel/Railway/local). When we add CMS-backed blog content,
+// swap to a per-entry timestamp from the source-of-truth metadata.
+const SITE_URL =
+  process.env.NEXT_PUBLIC_SITE_URL ?? "https://app.moleculesai.app";
+
+const BUILD_DATE = new Date();
+
+export default function sitemap(): MetadataRoute.Sitemap {
+  return [
+    {
+      url: `${SITE_URL}/`,
+      lastModified: BUILD_DATE,
+      changeFrequency: "weekly",
+      priority: 1.0,
+    },
+    {
+      url: `${SITE_URL}/pricing`,
+      lastModified: BUILD_DATE,
+      changeFrequency: "weekly",
+      priority: 0.9,
+    },
+    {
+      url: `${SITE_URL}/blog/2026-04-20-chrome-devtools-mcp`,
+      lastModified: new Date("2026-04-20"),
+      changeFrequency: "monthly",
+      priority: 0.6,
+    },
+  ];
+}
@@ -132,7 +132,7 @@ export function AuditTrailPanel({ workspaceId }: Props) {

  if (loading) {
    return (
-      <div className="flex items-center justify-center h-32">
+      <div role="status" aria-live="polite" className="flex items-center justify-center h-32">
        <span className="text-xs text-ink-mid">Loading audit trail…</span>
      </div>
    );
@@ -133,13 +133,13 @@ export function ConversationTraceModal({ open, workspaceId: _workspaceId, onClos
            {/* Timeline */}
            <div className="flex-1 overflow-y-auto px-5 py-4">
              {loading && (
-                <div className="text-xs text-ink-mid text-center py-8">
+                <div role="status" aria-live="polite" className="text-xs text-ink-mid text-center py-8">
                  Loading trace from all workspaces...
                </div>
              )}

              {!loading && entries.length === 0 && (
-                <div className="text-xs text-ink-mid text-center py-8">
+                <div role="status" aria-live="polite" className="text-xs text-ink-mid text-center py-8">
                  No activity found
                </div>
              )}
@@ -105,7 +105,7 @@ export function EmptyState() {

        {/* Template grid */}
        {loading ? (
-          <div className="flex items-center justify-center gap-2 text-xs text-ink-mid py-4">
+          <div role="status" aria-live="polite" className="flex items-center justify-center gap-2 text-xs text-ink-mid py-4">
            <Spinner />
            Loading templates...
          </div>
@@ -15,7 +15,7 @@
 //     ($AGENT_URL). They ARE NOT filled in server-side because the
 //     server doesn't know where the operator's agent will live.

-import { useCallback, useState } from "react";
+import { useCallback, useRef, useState } from "react";
 import * as Dialog from "@radix-ui/react-dialog";

 type Tab = "python" | "curl" | "claude" | "mcp" | "hermes" | "codex" | "openclaw" | "kimi" | "fields";
@@ -84,6 +84,33 @@ export function ExternalConnectModal({ info, onClose }: Props) {
    : "python";
  const [tab, setTab] = useState<Tab>(initialTab);
  const [copiedKey, setCopiedKey] = useState<string | null>(null);
+  const tabRefs = useRef<Map<Tab, HTMLButtonElement | null>>(new Map());
+
+  const handleTabKeyDown = useCallback(
+    (e: React.KeyboardEvent<HTMLButtonElement>, current: Tab, tabs: Tab[]) => {
+      const idx = tabs.indexOf(current);
+      if (e.key === "ArrowRight" || e.key === "ArrowDown") {
+        e.preventDefault();
+        const next = tabs[(idx + 1) % tabs.length];
+        setTab(next);
+        tabRefs.current.get(next)?.focus();
+      } else if (e.key === "ArrowLeft" || e.key === "ArrowUp") {
+        e.preventDefault();
+        const prev = tabs[(idx - 1 + tabs.length) % tabs.length];
+        setTab(prev);
+        tabRefs.current.get(prev)?.focus();
+      } else if (e.key === "Home") {
+        e.preventDefault();
+        setTab(tabs[0]);
+        tabRefs.current.get(tabs[0])?.focus();
+      } else if (e.key === "End") {
+        e.preventDefault();
+        setTab(tabs[tabs.length - 1]);
+        tabRefs.current.get(tabs[tabs.length - 1])?.focus();
+      }
+    },
+    [],
+  );

  const copy = useCallback(async (value: string, key: string) => {
    try {
@@ -160,6 +187,19 @@ export function ExternalConnectModal({ info, onClose }: Props) {
    `MOLECULE_WORKSPACE_TOKEN=${info.auth_token}`,
  );

+  // Build the tab list once so both the tab bar and keyboard handler
+  // share the same ordered array. Computed here (after all filled* vars)
+  // so TypeScript's block-scoping analysis can reach them.
+  const tabList: Tab[] = [];
+  if (filledUniversalMcp) tabList.push("mcp");
+  tabList.push("python");
+  if (filledChannel) tabList.push("claude");
+  if (filledHermes) tabList.push("hermes");
+  if (filledCodex) tabList.push("codex");
+  if (filledOpenClaw) tabList.push("openclaw");
+  if (filledKimi) tabList.push("kimi");
+  tabList.push("curl", "fields");
+
  return (
    <Dialog.Root open onOpenChange={(o) => !o && onClose()}>
      <Dialog.Portal>
@@ -180,34 +220,18 @@ export function ExternalConnectModal({ info, onClose }: Props) {
            aria-label="Connection snippet format"
            className="mt-4 flex gap-1 border-b border-line"
          >
-            {(() => {
-              // Build the tab order dynamically. Claude Code first
-              // (when offered) since it's the simplest setup; Python
-              // SDK second (full register+heartbeat+inbound); Universal
-              // MCP third (any MCP-aware runtime, outbound-only); curl
-              // for one-shot register; Fields for raw values.
-              // Tab order: Universal MCP first (default, runtime-
-              // agnostic primitives), then runtime-specific channel/
-              // SDK tabs, then curl + Fields. Each runtime tab only
-              // appears when the platform supplies the snippet — no
-              // dead "tab missing snippet" UX.
-              const tabs: Tab[] = [];
-              if (filledUniversalMcp) tabs.push("mcp");
-              tabs.push("python");
-              if (filledChannel) tabs.push("claude");
-              if (filledHermes) tabs.push("hermes");
-              if (filledCodex) tabs.push("codex");
-              if (filledOpenClaw) tabs.push("openclaw");
-              if (filledKimi) tabs.push("kimi");
-              tabs.push("curl", "fields");
-              return tabs;
-            })().map((t) => (
+            {tabList.map((t) => (
              <button
                key={t}
                type="button"
                role="tab"
+                id={`tab-${t}`}
                aria-selected={tab === t}
+                aria-controls={`panel-${t}`}
+                tabIndex={tab === t ? 0 : -1}
+                ref={(el) => { tabRefs.current.set(t, el); }}
                onClick={() => setTab(t)}
+                onKeyDown={(e) => handleTabKeyDown(e, t, tabList)}
                className={`px-3 py-2 text-sm border-b-2 -mb-px transition-colors focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1 focus-visible:ring-offset-surface ${
                  tab === t
                    ? "border-accent text-ink"
@@ -235,18 +259,39 @@ export function ExternalConnectModal({ info, onClose }: Props) {
            ))}
          </div>

-          {/* Snippet area */}
-          <div className="mt-3">
-            {tab === "claude" && filledChannel && (
-              <SnippetBlock
-                value={filledChannel}
-                label="Claude Code channel — polls workspace's A2A; no tunnel needed"
-                copyKey="claude"
-                copied={copiedKey === "claude"}
-                onCopy={() => copy(filledChannel, "claude")}
-              />
-            )}
-            {tab === "python" && (
+          {/* Snippet area — all panels always in the DOM so aria-controls
+              targets are stable. Hidden panels use aria-hidden so screen
+              readers skip them; active panel uses role=tabpanel with
+              aria-labelledby pointing to the tab button. */}
+          <div className="mt-3" data-testid="snippet-panels">
+            {/* Claude Code tab */}
+            <div
+              id="panel-claude"
+              data-testid="panel-claude"
+              role="tabpanel"
+              aria-labelledby="tab-claude"
+              hidden={tab !== "claude" || !filledChannel}
+              className={tab === "claude" && filledChannel ? "" : "hidden"}
+            >
+              {filledChannel && (
+                <SnippetBlock
+                  value={filledChannel}
+                  label="Claude Code channel — polls workspace's A2A; no tunnel needed"
+                  copyKey="claude"
+                  copied={copiedKey === "claude"}
+                  onCopy={() => copy(filledChannel, "claude")}
+                />
+              )}
+            </div>
+            {/* Python SDK tab */}
+            <div
+              id="panel-python"
+              data-testid="panel-python"
+              role="tabpanel"
+              aria-labelledby="tab-python"
+              hidden={tab !== "python"}
+              className={tab === "python" ? "" : "hidden"}
+            >
              <SnippetBlock
                value={filledPython}
                label="Python SDK — includes heartbeat loop (push-mode, needs public URL)"
@@ -254,8 +299,16 @@ export function ExternalConnectModal({ info, onClose }: Props) {
                copied={copiedKey === "python"}
                onCopy={() => copy(filledPython, "python")}
              />
-            )}
-            {tab === "curl" && (
+            </div>
+            {/* curl tab */}
+            <div
+              id="panel-curl"
+              data-testid="panel-curl"
+              role="tabpanel"
+              aria-labelledby="tab-curl"
+              hidden={tab !== "curl"}
+              className={tab === "curl" ? "" : "hidden"}
+            >
              <SnippetBlock
                value={filledCurl}
                label="curl — one-shot register only (no heartbeat)"
@@ -263,53 +316,111 @@ export function ExternalConnectModal({ info, onClose }: Props) {
                copied={copiedKey === "curl"}
                onCopy={() => copy(filledCurl, "curl")}
              />
-            )}
-            {tab === "mcp" && filledUniversalMcp && (
-              <SnippetBlock
-                value={filledUniversalMcp}
-                label="Universal MCP — standalone register + heartbeat + tools for any MCP-aware runtime (Claude Code, hermes, codex). Pair with Python or Claude Code tab if you need inbound A2A delivery."
-                copyKey="mcp"
-                copied={copiedKey === "mcp"}
-                onCopy={() => copy(filledUniversalMcp, "mcp")}
-              />
-            )}
-            {tab === "hermes" && filledHermes && (
-              <SnippetBlock
-                value={filledHermes}
-                label="Hermes channel — bridges this workspace's A2A traffic into your hermes-agent session as platform messages (push parity with Claude Code). Long-poll based; no tunnel needed."
-                copyKey="hermes"
-                copied={copiedKey === "hermes"}
-                onCopy={() => copy(filledHermes, "hermes")}
-              />
-            )}
-            {tab === "codex" && filledCodex && (
-              <SnippetBlock
-                value={filledCodex}
-                label="Codex MCP config — wires the molecule MCP server into ~/.codex/config.toml. Outbound tools today; inbound A2A push needs the Python SDK tab paired in (codex's MCP runtime doesn't route arbitrary notifications/* yet)."
-                copyKey="codex"
-                copied={copiedKey === "codex"}
-                onCopy={() => copy(filledCodex, "codex")}
-              />
-            )}
-            {tab === "openclaw" && filledOpenClaw && (
-              <SnippetBlock
-                value={filledOpenClaw}
-                label="OpenClaw MCP config — wires the molecule MCP server via openclaw mcp set + starts the gateway on loopback. Outbound tools today; inbound A2A push on an external openclaw needs the Python SDK tab paired in (a sessions.steer bridge daemon is future work)."
-                copyKey="openclaw"
-                copied={copiedKey === "openclaw"}
-                onCopy={() => copy(filledOpenClaw, "openclaw")}
-              />
-            )}
-            {tab === "kimi" && filledKimi && (
-              <SnippetBlock
-                value={filledKimi}
-                label="Kimi CLI — self-contained Python bridge. Registers, heartbeats, polls for canvas messages, and echoes replies back. NAT-safe (no public URL). Run in a background terminal or via launchd."
-                copyKey="kimi"
-                copied={copiedKey === "kimi"}
-                onCopy={() => copy(filledKimi, "kimi")}
-              />
-            )}
-            {tab === "fields" && (
+            </div>
+            {/* Universal MCP tab */}
+            <div
+              id="panel-mcp"
+              data-testid="panel-mcp"
+              role="tabpanel"
+              aria-labelledby="tab-mcp"
+              hidden={tab !== "mcp" || !filledUniversalMcp}
+              className={tab === "mcp" && filledUniversalMcp ? "" : "hidden"}
+            >
+              {filledUniversalMcp && (
+                <SnippetBlock
+                  value={filledUniversalMcp}
+                  label="Universal MCP — standalone register + heartbeat + tools for any MCP-aware runtime (Claude Code, hermes, codex). Pair with Python or Claude Code tab if you need inbound A2A delivery."
+                  copyKey="mcp"
+                  copied={copiedKey === "mcp"}
+                  onCopy={() => copy(filledUniversalMcp, "mcp")}
+                />
+              )}
+            </div>
+            {/* Hermes tab */}
+            <div
+              id="panel-hermes"
+              data-testid="panel-hermes"
+              role="tabpanel"
+              aria-labelledby="tab-hermes"
+              hidden={tab !== "hermes" || !filledHermes}
+              className={tab === "hermes" && filledHermes ? "" : "hidden"}
+            >
+              {filledHermes && (
+                <SnippetBlock
+                  value={filledHermes}
+                  label="Hermes channel — bridges this workspace's A2A traffic into your hermes-agent session as platform messages (push parity with Claude Code). Long-poll based; no tunnel needed."
+                  copyKey="hermes"
+                  copied={copiedKey === "hermes"}
+                  onCopy={() => copy(filledHermes, "hermes")}
+                />
+              )}
+            </div>
+            {/* Codex tab */}
+            <div
+              id="panel-codex"
+              data-testid="panel-codex"
+              role="tabpanel"
+              aria-labelledby="tab-codex"
+              hidden={tab !== "codex" || !filledCodex}
+              className={tab === "codex" && filledCodex ? "" : "hidden"}
+            >
+              {filledCodex && (
+                <SnippetBlock
+                  value={filledCodex}
+                  label="Codex MCP config — wires the molecule MCP server into ~/.codex/config.toml. Outbound tools today; inbound A2A push needs the Python SDK tab paired in (codex's MCP runtime doesn't route arbitrary notifications/* yet)."
+                  copyKey="codex"
+                  copied={copiedKey === "codex"}
+                  onCopy={() => copy(filledCodex, "codex")}
+                />
+              )}
+            </div>
+            {/* OpenClaw tab */}
+            <div
+              id="panel-openclaw"
+              data-testid="panel-openclaw"
+              role="tabpanel"
+              aria-labelledby="tab-openclaw"
+              hidden={tab !== "openclaw" || !filledOpenClaw}
+              className={tab === "openclaw" && filledOpenClaw ? "" : "hidden"}
+            >
+              {filledOpenClaw && (
+                <SnippetBlock
+                  value={filledOpenClaw}
+                  label="OpenClaw MCP config — wires the molecule MCP server via openclaw mcp set + starts the gateway on loopback. Outbound tools today; inbound A2A push on an external openclaw needs the Python SDK tab paired in (a sessions.steer bridge daemon is future work)."
+                  copyKey="openclaw"
+                  copied={copiedKey === "openclaw"}
+                  onCopy={() => copy(filledOpenClaw, "openclaw")}
+                />
+              )}
+            </div>
+            {/* Kimi tab */}
+            <div
+              id="panel-kimi"
+              data-testid="panel-kimi"
+              role="tabpanel"
+              aria-labelledby="tab-kimi"
+              hidden={tab !== "kimi" || !filledKimi}
+              className={tab === "kimi" && filledKimi ? "" : "hidden"}
+            >
+              {filledKimi && (
+                <SnippetBlock
+                  value={filledKimi}
+                  label="Kimi CLI — self-contained Python bridge. Registers, heartbeats, polls for canvas messages, and echoes replies back. NAT-safe (no public URL). Run in a background terminal or via launchd."
+                  copyKey="kimi"
+                  copied={copiedKey === "kimi"}
+                  onCopy={() => copy(filledKimi, "kimi")}
+                />
+              )}
+            </div>
+            {/* Fields tab */}
+            <div
+              id="panel-fields"
+              data-testid="panel-fields"
+              role="tabpanel"
+              aria-labelledby="tab-fields"
+              hidden={tab !== "fields"}
+              className={tab === "fields" ? "" : "hidden"}
+            >
              <div className="space-y-2">
                <Field label="workspace_id" value={info.workspace_id} onCopy={() => copy(info.workspace_id, "wsid")} copied={copiedKey === "wsid"} />
                <Field label="platform_url" value={info.platform_url} onCopy={() => copy(info.platform_url, "url")} copied={copiedKey === "url"} />
@@ -323,7 +434,7 @@ export function ExternalConnectModal({ info, onClose }: Props) {
                <Field label="registry_endpoint" value={info.registry_endpoint} onCopy={() => copy(info.registry_endpoint, "reg")} copied={copiedKey === "reg"} />
                <Field label="heartbeat_endpoint" value={info.heartbeat_endpoint} onCopy={() => copy(info.heartbeat_endpoint, "hb")} copied={copiedKey === "hb"} />
              </div>
-            )}
+            </div>
          </div>

          <div className="mt-5 flex justify-end gap-2">
@@ -440,6 +440,7 @@ function ProviderPickerModal({
                      onChange={(e) => updateEntry(index, { value: e.target.value.trimStart() })}
                      placeholder={entry.key.includes("API_KEY") ? "sk-..." : "Enter value"}
                      type="password"
+                      aria-label={`Value for ${entry.key}`}
                      ref={index === 0 ? firstInputRef : undefined}
                      onKeyDown={(e) => {
                        if (e.key === "Enter" && entry.value.trim()) {
@@ -459,7 +460,7 @@ function ProviderPickerModal({
                )}

                {entry.error && (
-                  <div className="mt-1.5 text-[10px] text-bad">{entry.error}</div>
+                  <div role="alert" aria-live="assertive" className="mt-1.5 text-[10px] text-bad">{entry.error}</div>
                )}
              </div>
            ))}
@@ -694,6 +695,7 @@ function AllKeysModal({
                    onChange={(e) => updateEntry(index, { value: e.target.value.trimStart() })}
                    placeholder={entry.key.includes("API_KEY") ? "sk-..." : "Enter value"}
                    type="password"
+                    aria-label={`Value for ${entry.key}`}
                    autoFocus={index === 0}
                    onKeyDown={(e) => {
                      if (e.key === "Enter" && entry.value.trim()) {
@@ -718,7 +720,7 @@ function AllKeysModal({
          ))}

          {globalError && (
-            <div className="px-3 py-2 bg-red-950/40 border border-red-800/50 rounded-lg text-[11px] text-bad">
+            <div role="alert" aria-live="assertive" className="px-3 py-2 bg-red-950/40 border border-red-800/50 rounded-lg text-[11px] text-bad">
              {globalError}
            </div>
          )}
@@ -71,7 +71,7 @@ export function WorkspaceUsage({ workspaceId }: WorkspaceUsageProps) {
            <SkeletonRow />
          </>
        ) : error ? (
-          <p className="text-xs text-bad" data-testid="usage-error">
+          <p role="alert" aria-live="assertive" className="text-xs text-bad" data-testid="usage-error">
            {error}
          </p>
        ) : metrics ? (
@@ -131,7 +131,9 @@ describe("ExternalConnectModal — tab switching", () => {
  it("switches to the Python SDK tab and shows the snippet with stamped token", () => {
    renderAndFlush(defaultInfo);
    fireEvent.click(screen.getByRole("tab", { name: /python sdk/i }));
-    const preEl = document.querySelector("pre");
+    // Query within the python panel so we get the right pre (not the first in DOM).
+    const pythonPanel = document.querySelector("[data-testid='panel-python']");
+    const preEl = pythonPanel?.querySelector("pre");
    expect(preEl?.textContent).toContain("AUTH_TOKEN");
    // The placeholder is replaced with the real auth token
    expect(preEl?.textContent).toContain("secret-auth-token-abc");
@@ -140,7 +142,9 @@ describe("ExternalConnectModal — tab switching", () => {
  it("switches to the curl tab and shows the snippet with stamped token", () => {
    renderAndFlush(defaultInfo);
    fireEvent.click(screen.getByRole("tab", { name: /curl/i }));
-    const preEl = document.querySelector("pre");
+    // Query within the curl panel so we get the right pre (not the first in DOM).
+    const curlPanel = document.querySelector("[data-testid='panel-curl']");
+    const preEl = curlPanel?.querySelector("pre");
    expect(preEl?.textContent).toContain("curl");
    expect(preEl?.textContent).toContain("secret-auth-token-abc");
  });
@@ -148,9 +152,11 @@ describe("ExternalConnectModal — tab switching", () => {
  it("switches to the Fields tab and shows raw values", () => {
    renderAndFlush(defaultInfo);
    fireEvent.click(screen.getByRole("tab", { name: /fields/i }));
-    expect(screen.getByText("ws-123")).toBeTruthy();
-    expect(screen.getByText("https://app.example.com")).toBeTruthy();
-    expect(screen.getByText("secret-auth-token-abc")).toBeTruthy();
+    // Query within the fields panel for specific values.
+    const fieldsPanel = document.querySelector("[data-testid='panel-fields']");
+    expect(fieldsPanel?.textContent).toContain("ws-123");
+    expect(fieldsPanel?.textContent).toContain("https://app.example.com");
+    expect(fieldsPanel?.textContent).toContain("secret-auth-token-abc");
  });

  it("hides the Hermes tab when hermes_channel_snippet is absent", () => {
@@ -168,7 +174,8 @@ describe("ExternalConnectModal — snippet token stamping", () => {
  it("stamps the real auth_token into the Python snippet instead of the placeholder", () => {
    renderAndFlush(defaultInfo);
    fireEvent.click(screen.getByRole("tab", { name: /python sdk/i }));
-    const preEl = document.querySelector("pre");
+    const pythonPanel = document.querySelector("[data-testid='panel-python']");
+    const preEl = pythonPanel?.querySelector("pre");
    expect(preEl?.textContent).not.toContain("<paste from create response>");
    expect(preEl?.textContent).toContain("secret-auth-token-abc");
  });
@@ -176,7 +183,8 @@ describe("ExternalConnectModal — snippet token stamping", () => {
  it("stamps the real auth_token into the curl snippet", () => {
    renderAndFlush(defaultInfo);
    fireEvent.click(screen.getByRole("tab", { name: /curl/i }));
-    const preEl = document.querySelector("pre");
+    const curlPanel = document.querySelector("[data-testid='panel-curl']");
+    const preEl = curlPanel?.querySelector("pre");
    // curl template uses WORKSPACE_AUTH_TOKEN placeholder, not the generic one
    expect(preEl?.textContent).toContain("secret-auth-token-abc");
  });
@@ -184,7 +192,8 @@ describe("ExternalConnectModal — snippet token stamping", () => {
  it("stamps the real auth_token into the Universal MCP snippet", () => {
    renderAndFlush(defaultInfo);
    // Default tab is Universal MCP
-    const preEl = document.querySelector("pre");
+    const mcpPanel = document.querySelector("[data-testid='panel-mcp']");
+    const preEl = mcpPanel?.querySelector("pre");
    expect(preEl?.textContent).toContain("secret-auth-token-abc");
    expect(preEl?.textContent).not.toContain("<paste from create response>");
  });
@@ -193,8 +202,10 @@ describe("ExternalConnectModal — snippet token stamping", () => {
 describe("ExternalConnectModal — copy functionality", () => {
  it("calls navigator.clipboard.writeText with the snippet text", () => {
    renderAndFlush(defaultInfo);
-    // Default tab is Universal MCP
-    fireEvent.click(screen.getByRole("button", { name: /^copy$/i }));
+    // Default tab is Universal MCP — query the copy button within the mcp panel.
+    const mcpPanel = document.querySelector("[data-testid='panel-mcp']");
+    const copyBtn = mcpPanel?.querySelector("button");
+    if (copyBtn) fireEvent.click(copyBtn);
    expect(clipboardWriteText).toHaveBeenCalledWith(
      expect.stringContaining("secret-auth-token-abc"),
    );
@@ -227,7 +238,8 @@ describe("ExternalConnectModal — missing optional fields", () => {
    };
    renderAndFlush(minimalInfo);
    fireEvent.click(screen.getByRole("tab", { name: /fields/i }));
-    expect(screen.getByText("(missing)")).toBeTruthy();
+    const fieldsPanel = document.querySelector("[data-testid='panel-fields']");
+    expect(fieldsPanel?.textContent).toContain("(missing)");
  });

  it("hides the Hermes tab when hermes_channel_snippet is absent", () => {
@@ -11,13 +11,21 @@ import { render, screen, fireEvent, cleanup, act } from "@testing-library/react"
 import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
 import { TestConnectionButton } from "../ui/TestConnectionButton";
 import type { SecretGroup } from "@/types/secrets";
-import { validateSecret } from "@/lib/api/secrets";
+import { validateSecret, ApiError } from "@/lib/api/secrets";

 // ─── Mock validateSecret ──────────────────────────────────────────────────────
 // vi.mock is hoisted, so validateSecret (imported above) refers to the mocked
 // namespace value once vi.mock runs. Use vi.mocked() to access it in tests.
 vi.mock("@/lib/api/secrets", () => ({
  validateSecret: vi.fn(),
+  ApiError: class ApiError extends Error {
+    status: number;
+    constructor(status: number, message: string) {
+      super(message);
+      this.name = "ApiError";
+      this.status = status;
+    }
+  },
 }));

 // SecretGroup is a string literal type: 'github' | 'anthropic' | 'openrouter' | 'custom'
@@ -102,7 +110,7 @@ describe("TestConnectionButton — state machine", () => {
    expect(screen.getByText("Permission denied")).toBeTruthy();
  });

-  it("shows generic error message on unexpected exception", async () => {
+  it("shows a connectivity message on a genuine network exception", async () => {
    vi.mocked(validateSecret).mockRejectedValue(new Error("timeout"));
    render(<TestConnectionButton provider={toGroup("anthropic")} secretValue="sk-..." />);

@@ -110,8 +118,23 @@ describe("TestConnectionButton — state machine", () => {
    await act(async () => { /* flush */ });

    expect(screen.getByRole("alert")).toBeTruthy();
-    // The error detail is hardcoded to "Connection timed out. Service may be down."
-    expect(document.body.querySelector('[role="alert"]')?.textContent).toMatch(/timed out/i);
+    // A real thrown network error → honest connectivity message (not a
+    // fabricated "service down"); see internal#492.
+    expect(document.body.querySelector('[role="alert"]')?.textContent).toMatch(
+      /could not reach the validation service/i,
+    );
+  });
+
+  it("does not claim a timeout when the validate endpoint 404s (internal#492)", async () => {
+    vi.mocked(validateSecret).mockRejectedValue(new ApiError(404, "Not Found"));
+    render(<TestConnectionButton provider={toGroup("anthropic")} secretValue="sk-..." />);
+
+    fireEvent.click(screen.getByRole("button"));
+    await act(async () => { /* flush */ });
+
+    const alert = document.body.querySelector('[role="alert"]')?.textContent ?? "";
+    expect(alert).not.toMatch(/timed out/i);
+    expect(alert).toMatch(/not available/i);
  });
 });

@@ -223,6 +223,7 @@ export function MobileCanvas({
            textTransform: "uppercase",
            fontWeight: 600,
          }}
+          className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
        >
          Reset
        </button>
@@ -2,8 +2,11 @@

 // 04 · Chat — message thread + composer + sub-tabs.
 // Wired to the same /workspaces/:id/a2a (method message/send) endpoint
-// that the desktop ChatTab uses, but with a slimmer surface: no
-// attachments, no A2A topology overlay, no conversation tracing.
+// that the desktop ChatTab uses. Render parity with desktop ChatTab is
+// achieved by reusing its renderers rather than forking a reduced
+// mobile path: the Agent Comms sub-tab mounts the same AgentCommsPanel,
+// and message attachments route through the same AttachmentPreview
+// dispatch the desktop My-Chat bubble uses (#231/#232).

 import { useEffect, useMemo, useRef, useState } from "react";
 import ReactMarkdown from "react-markdown";
@@ -16,6 +19,9 @@ import {
  useChatSend,
  useChatSocket,
 } from "@/components/tabs/chat/hooks";
+import { AgentCommsPanel } from "@/components/tabs/chat/AgentCommsPanel";
+import { AttachmentPreview } from "@/components/tabs/chat/AttachmentPreview";
+import { downloadChatFile } from "@/components/tabs/chat/uploads";

 import { toMobileAgent } from "./components";
 import { MOBILE_FONT_MONO, MOBILE_FONT_SANS, usePalette } from "./palette";
@@ -304,6 +310,17 @@ export function MobileChat({
  const removePendingFile = (index: number) =>
    setPendingFiles((prev) => prev.filter((_, i) => i !== index));

+  // Route attachment downloads through the same authenticated helper
+  // the desktop ChatTab uses (downloadChatFile) so platform-scheme
+  // URIs get a real Blob with auth headers instead of about:blank.
+  const downloadAttachment = (att: ChatAttachment) => {
+    downloadChatFile(agentId, att).catch(() => {
+      // AttachmentPreview's own error affordance covers the in-bubble
+      // failure state; matches ChatTab's behaviour of not double-
+      // reporting a download failure.
+    });
+  };
+
  const send = async () => {
    const text = draft.trim();
    if ((!text && pendingFiles.length === 0) || sending || !reachable) return;
@@ -339,6 +356,7 @@ export function MobileChat({
            type="button"
            onClick={onBack}
            aria-label="Back"
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            style={{
              width: 36,
              height: 36,
@@ -385,6 +403,7 @@ export function MobileChat({
          <button
            type="button"
            aria-label="More"
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            style={{
              width: 36,
              height: 36,
@@ -415,6 +434,7 @@ export function MobileChat({
                key={t.id}
                type="button"
                onClick={() => setTab(t.id)}
+                className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
                style={{
                  padding: "4px 0 8px",
                  border: "none",
@@ -433,7 +453,19 @@ export function MobileChat({
        </div>
      </div>

+      {/* Agent Comms — reuse the desktop AgentCommsPanel verbatim so
+          mobile renders the identical peer/A2A + delegation feed
+          (history GET + live socket events) instead of a placeholder
+          (#231). The panel owns its own scroll/load/error/empty
+          states, matching ChatTab's agent-comms tabpanel. */}
+      {tab === "a2a" && (
+        <div style={{ flex: 1, minHeight: 0, overflow: "hidden" }}>
+          <AgentCommsPanel workspaceId={agentId} />
+        </div>
+      )}
+
      {/* Messages */}
+      {tab === "my" && (
      <div
        ref={scrollRef}
        style={{
@@ -445,20 +477,8 @@ export function MobileChat({
          gap: 8,
        }}
      >
-        {tab === "a2a" && (
-          <div
-            style={{
-              padding: "20px 4px",
-              textAlign: "center",
-              color: p.text3,
-              fontSize: 13,
-            }}
-          >
-            Agent Comms — peer-to-peer A2A traffic surfaces in the Comms tab.
-          </div>
-        )}
        {tab === "my" && historyLoading && (
-          <div style={{ padding: "20px 4px", textAlign: "center", color: p.text3, fontSize: 13 }}>
+          <div role="status" aria-live="polite" style={{ padding: "20px 4px", textAlign: "center", color: p.text3, fontSize: 13 }}>
            Loading chat history…
          </div>
        )}
@@ -478,6 +498,8 @@ export function MobileChat({
              onClick={() => {
                loadInitial();
              }}
+              aria-label="Retry loading chat history"
+              className="focus:outline-none focus-visible:ring-2 focus-visible:ring-red-400"
              style={{
                padding: "6px 14px",
                borderRadius: 14,
@@ -493,7 +515,7 @@ export function MobileChat({
          </div>
        )}
        {tab === "my" && !historyLoading && !historyError && messages.length === 0 && (
-          <div style={{ padding: "20px 4px", textAlign: "center", color: p.text3, fontSize: 13 }}>
+          <div role="status" aria-live="polite" style={{ padding: "20px 4px", textAlign: "center", color: p.text3, fontSize: 13 }}>
            Send a message to start chatting.
          </div>
        )}
@@ -521,9 +543,31 @@ export function MobileChat({
                    overflowWrap: "anywhere",
                  }}
                >
-                  <MarkdownBubble dark={dark} accent={p.accent}>
-                    {m.content}
-                  </MarkdownBubble>
+                  {m.content && (
+                    <MarkdownBubble dark={dark} accent={p.accent}>
+                      {m.content}
+                    </MarkdownBubble>
+                  )}
+                  {m.attachments && m.attachments.length > 0 && (
+                    <div
+                      style={{
+                        display: "flex",
+                        flexWrap: "wrap",
+                        gap: 4,
+                        marginTop: m.content ? 6 : 0,
+                      }}
+                    >
+                      {m.attachments.map((att, i) => (
+                        <AttachmentPreview
+                          key={`${m.id}-${i}`}
+                          workspaceId={agentId}
+                          attachment={att}
+                          onDownload={downloadAttachment}
+                          tone={mine ? "user" : "agent"}
+                        />
+                      ))}
+                    </div>
+                  )}
                  <div
                    style={{
                      fontSize: 10,
@@ -554,7 +598,13 @@ export function MobileChat({
          </div>
        )}
      </div>
+      )}

+      {/* Footer ID + composer belong to My Chat only. The Agent Comms
+          tab is a read-only peer/A2A feed (parity with desktop
+          ChatTab, where the agent-comms tabpanel has no composer). */}
+      {tab === "my" && (
+      <>
      {/* Footer ID */}
      <div
        style={{
@@ -619,6 +669,7 @@ export function MobileChat({
                  type="button"
                  onClick={() => removePendingFile(i)}
                  aria-label={`Remove ${f.name}`}
+                  className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
                  style={{
                    border: "none",
                    background: "transparent",
@@ -659,6 +710,7 @@ export function MobileChat({
            onClick={() => fileInputRef.current?.click()}
            disabled={!reachable || sending || uploading}
            aria-label="Attach"
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            style={{
              width: 32,
              height: 32,
@@ -680,6 +732,7 @@ export function MobileChat({
            ref={composerRef}
            value={draft}
            onChange={(e) => setDraft(e.target.value)}
+            aria-label="Message"
            onKeyDown={(e) => {
              // Enter sends; Shift+Enter inserts a newline. Skip when the
              // IME is composing — pressing Enter to commit a Chinese/
@@ -703,7 +756,12 @@ export function MobileChat({
              border: "none",
              outline: "none",
              background: "transparent",
-              fontSize: 14.5,
+              // iOS Safari/PWA zooms the viewport when a focused textarea
+              // has a computed font-size below 16px. 14.5 triggers that
+              // focus-zoom; the page looks broken until the user pinches
+              // back (#224, same class as desktop #1434 / sibling #225).
+              // 16px is the minimum that keeps focus from zooming.
+              fontSize: 16,
              lineHeight: 1.4,
              color: p.text,
              padding: "6px 0",
@@ -719,12 +777,13 @@ export function MobileChat({
            onClick={send}
            disabled={(!draft.trim() && pendingFiles.length === 0) || !reachable || sending || uploading}
            aria-label="Send"
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            style={{
              width: 36,
              height: 36,
              borderRadius: 999,
              border: "none",
-              cursor: (draft.trim() || pendingFiles.length > 0) && !sending && !uploading ? "pointer" : "not-allowed",
+              cursor: (draft.trim() || pendingFiles.length === 0) && !sending && !uploading ? "pointer" : "not-allowed",
              flexShrink: 0,
              background:
                (draft.trim() || pendingFiles.length > 0) && reachable && !sending && !uploading
@@ -746,6 +805,8 @@ export function MobileChat({
          </button>
        </div>
      </div>
+      </>
+      )}
    </div>
  );
 }
@@ -231,6 +231,7 @@ export function MobileComms({ dark }: { dark: boolean }) {
                fontSize: 13,
                fontWeight: 500,
              }}
+              className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            >
              {o.label}
              <span
@@ -251,11 +252,11 @@ export function MobileComms({ dark }: { dark: boolean }) {

      <div style={{ padding: "0 14px", display: "flex", flexDirection: "column", gap: 8 }}>
        {loading && items.length === 0 ? (
-          <div style={{ padding: "30px 4px", textAlign: "center", color: p.text3, fontSize: 13 }}>
+          <div role="status" aria-live="polite" style={{ padding: "30px 4px", textAlign: "center", color: p.text3, fontSize: 13 }}>
            Loading recent comms…
          </div>
        ) : filtered.length === 0 ? (
-          <div style={{ padding: "30px 4px", textAlign: "center", color: p.text3, fontSize: 13 }}>
+          <div role="status" aria-live="polite" style={{ padding: "30px 4px", textAlign: "center", color: p.text3, fontSize: 13 }}>
            No A2A traffic yet.
          </div>
        ) : (
@@ -83,11 +83,12 @@ export function MobileDetail({
            type="button"
            onClick={onBack}
            aria-label="Back"
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            style={iconButtonStyle(p, dark)}
          >
            {Icons.back({ size: 18 })}
          </button>
-          <button type="button" aria-label="More" style={iconButtonStyle(p, dark)}>
+          <button type="button" aria-label="More" className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900" style={iconButtonStyle(p, dark)}>
            {Icons.more({ size: 18 })}
          </button>
        </div>
@@ -183,6 +184,7 @@ export function MobileDetail({
              key={t.id}
              type="button"
              onClick={() => setTab(t.id)}
+              className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
              style={{
                padding: "8px 14px",
                borderRadius: 999,
@@ -215,6 +217,7 @@ export function MobileDetail({
          type="button"
          onClick={onChat}
          data-testid="mobile-chat-cta"
+          className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
          style={{
            width: "100%",
            height: 52,
@@ -416,6 +419,8 @@ function DetailActivity({ workspaceId, dark }: { workspaceId: string; dark: bool
  if (items === null) {
    return (
      <div
+        role="status"
+        aria-live="polite"
        style={{
          background: p.surface,
          borderRadius: 16,
@@ -200,6 +200,7 @@ export function MobileHome({
          justifyContent: "center",
          boxShadow: "0 8px 24px rgba(40,30,20,0.25), 0 2px 6px rgba(40,30,20,0.15)",
        }}
+        className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
      >
        {Icons.plus({ size: 22 })}
      </button>
@@ -92,6 +92,7 @@ export function MobileMe({
                    border: on ? `2px solid ${p.text}` : "2px solid transparent",
                    boxShadow: on ? `0 0 0 2px ${p.bg} inset` : "none",
                  }}
+                  className="focus:outline-none focus-visible:ring-2 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
                />
              );
            })}
@@ -184,6 +185,7 @@ function SegmentedRow({
              fontSize: 13,
              fontWeight: 600,
            }}
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
          >
            {o.label}
          </button>
@@ -148,6 +148,7 @@ export function MobileSpawn({ dark, onClose }: { dark: boolean; onClose: () => v
            type="button"
            onClick={onClose}
            aria-label="Close"
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            style={{
              width: 32,
              height: 32,
@@ -170,6 +171,8 @@ export function MobileSpawn({ dark, onClose }: { dark: boolean; onClose: () => v
        <div style={{ padding: "0 14px" }}>
          {loadingTemplates ? (
            <div
+              role="status"
+              aria-live="polite"
              style={{
                padding: "24px 8px",
                textAlign: "center",
@@ -214,6 +217,8 @@ export function MobileSpawn({ dark, onClose }: { dark: boolean; onClose: () => v
                      setTplId(t.id);
                      setTier(tCode);
                    }}
+                    aria-label={`Select template: ${t.name} (tier ${t.tier})`}
+                    className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
                    style={{
                      background: on
                        ? dark
@@ -302,6 +307,7 @@ export function MobileSpawn({ dark, onClose }: { dark: boolean; onClose: () => v
          <input
            value={name}
            onChange={(e) => setName(e.target.value)}
+            aria-label="Agent name"
            placeholder={tplId
              ? (templates.find((t) => t.id === tplId)?.name ?? "agent-name")
              : "agent-name"}
@@ -312,7 +318,12 @@ export function MobileSpawn({ dark, onClose }: { dark: boolean; onClose: () => v
              border: `0.5px solid ${p.border}`,
              borderRadius: 12,
              fontFamily: MOBILE_FONT_MONO,
-              fontSize: 13.5,
+              // iOS Safari/PWA zooms the viewport when a focused input has
+              // a computed font-size below 16px; the layout jumps and the
+              // page looks broken until the user pinches back (#224 / #225,
+              // same class as desktop #1434). 16px is the minimum that
+              // suppresses that focus-zoom.
+              fontSize: 16,
              color: p.text,
              outline: "none",
              boxSizing: "border-box",
@@ -330,6 +341,8 @@ export function MobileSpawn({ dark, onClose }: { dark: boolean; onClose: () => v
                key={t}
                type="button"
                onClick={() => setTier(t)}
+                aria-label={`Select tier ${t}: ${TIER_LABEL[t]}`}
+                className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
                style={{
                  flex: 1,
                  padding: "10px 8px",
@@ -377,6 +390,8 @@ export function MobileSpawn({ dark, onClose }: { dark: boolean; onClose: () => v
            type="button"
            onClick={handleSpawn}
            disabled={busy || !tplId || templates.length === 0}
+            aria-label="Spawn agent"
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            style={{
              width: "100%",
              height: 52,
@@ -21,6 +21,14 @@ import { MobileChat } from "../MobileChat";
 vi.mock("@/lib/api");
 import { api } from "@/lib/api";

+// AgentCommsPanel (mounted by the Agent Comms sub-tab, #231) subscribes
+// to the global socket via useSocketEvent. Stub it to a no-op so the
+// panel mounts without the real ReconnectingSocket — the parity tests
+// only assert the panel renders (vs the old static placeholder).
+vi.mock("@/hooks/useSocketEvent", () => ({
+  useSocketEvent: vi.fn(),
+}));
+
 // ─── Mock store ───────────────────────────────────────────────────────────────

 const mockAgentId = "ws-chat-test";
@@ -155,6 +163,12 @@ beforeEach(() => {
  mockOnBack.mockClear();
  mockStoreState.nodes = [];
  mockStoreState.agentMessages = {};
+  // jsdom doesn't implement scrollIntoView. The Agent Comms tab now
+  // mounts AgentCommsPanel (#231), which scrolls its feed to bottom on
+  // arrival; a no-op stub keeps the panel from throwing under jsdom
+  // (same stub AgentCommsPanel's own render test installs).
+  Element.prototype.scrollIntoView =
+    vi.fn() as unknown as Element["scrollIntoView"];
  // Set up spies on the real api methods. Tests override these per-call.
  const getSpy = vi.spyOn(api, "get");
  const postSpy = vi.spyOn(api, "post");
@@ -249,6 +263,20 @@ describe("MobileChat — composer", () => {
    const sendBtn = container.querySelector('[aria-label="Send"]') as HTMLButtonElement;
    expect(sendBtn.disabled).toBe(true);
  });
+
+  // Regression #224: the composer textarea must render with font-size
+  // ≥ 16px. iOS Safari and PWAs auto-zoom the viewport when a focused
+  // input has a computed font-size below 16px — the layout jumps and
+  // the page looks broken until the user pinches back. Same class as
+  // desktop #1434 / sibling MobileSpawn #225.
+  it("composer textarea renders at font-size 16px or greater (iOS focus-zoom regression #224)", () => {
+    const { container } = renderChat(mockAgentId);
+    const textarea = container.querySelector("textarea") as HTMLTextAreaElement;
+    expect(textarea).toBeTruthy();
+    const fs = Number.parseFloat(textarea.style.fontSize);
+    expect(Number.isFinite(fs)).toBe(true);
+    expect(fs).toBeGreaterThanOrEqual(16);
+  });
 });

 // ─── Tabs ─────────────────────────────────────────────────────────────────────
@@ -474,3 +502,146 @@ describe("MobileChat — chat history", () => {
    expect(getSpy).toHaveBeenCalledTimes(2);
  });
 });
+
+// ─── #232 · Attachment render parity with desktop ChatTab ────────────────────
+//
+// Regression for the CTO-reported mobile bug: MobileChat used to render
+// only m.content (no attachment surface), so files sent/received in a
+// conversation were invisible on mobile while desktop showed them. The
+// fix routes m.attachments through the same AttachmentPreview the
+// desktop ChatTab bubble uses.
+
+describe("MobileChat — attachment render parity (#232)", () => {
+  beforeEach(() => {
+    mockStoreState.nodes = [onlineNode];
+  });
+
+  it("renders an attachment from a history message via AttachmentPreview", async () => {
+    const getSpy = vi.spyOn(api, "get");
+    // useChatHistory reads { messages, reached_end }.
+    getSpy.mockResolvedValueOnce({
+      messages: [
+        {
+          id: "m-att-1",
+          role: "agent",
+          content: "Here is the report",
+          attachments: [
+            {
+              name: "report.csv",
+              uri: "workspace://out/report.csv",
+              mimeType: "text/csv",
+              size: 2048,
+            },
+          ],
+          timestamp: new Date().toISOString(),
+        },
+      ],
+      reached_end: true,
+    });
+
+    let rr: ReturnType<typeof renderChat>;
+    await act(async () => {
+      rr = renderChat(mockAgentId);
+    });
+    const { container } = rr!;
+
+    // A non-image attachment renders the AttachmentChip download button
+    // with title="Download <name>" — same component the desktop bubble
+    // dispatches through AttachmentPreview.
+    await waitFor(() => {
+      const chip = container.querySelector('[title="Download report.csv"]');
+      expect(chip).toBeTruthy();
+    });
+    expect(container.textContent ?? "").toContain("report.csv");
+  });
+});
+
+// ─── #231 · Agent Comms (A2A/peer) render parity with desktop ChatTab ────────
+//
+// Regression for the CTO-reported mobile bug: the Agent Comms sub-tab
+// rendered a static placeholder string ("peer-to-peer A2A traffic
+// surfaces in the Comms tab") instead of the real feed. The fix mounts
+// the same AgentCommsPanel the desktop ChatTab agent-comms tabpanel
+// uses, so peer/A2A + delegation activity is visible on mobile.
+
+describe("MobileChat — Agent Comms render parity (#231)", () => {
+  beforeEach(() => {
+    mockStoreState.nodes = [onlineNode];
+  });
+
+  it("mounts AgentCommsPanel on the Agent Comms tab (not the old placeholder)", async () => {
+    const getSpy = vi.spyOn(api, "get");
+    // 1st GET: useChatHistory (My Chat) on mount.
+    getSpy.mockResolvedValueOnce({ messages: [], reached_end: true });
+    // 2nd GET: AgentCommsPanel's activity load when the tab is shown.
+    // Empty list → panel renders its own empty state, which still
+    // proves AgentCommsPanel mounted (vs. the removed placeholder).
+    getSpy.mockResolvedValueOnce([]);
+
+    let rr: ReturnType<typeof renderChat>;
+    await act(async () => {
+      rr = renderChat(mockAgentId);
+    });
+    const { container } = rr!;
+
+    const commsTab = Array.from(container.querySelectorAll("button")).find(
+      (b) => b.textContent?.trim() === "Agent Comms",
+    );
+    expect(commsTab).toBeTruthy();
+    await act(async () => {
+      commsTab!.click();
+    });
+
+    await waitFor(() => {
+      const text = container.textContent ?? "";
+      // The panel's empty state — proves AgentCommsPanel mounted.
+      expect(text).toContain("No agent-to-agent communications yet.");
+    });
+    // The old hard-coded placeholder must be gone.
+    expect(container.textContent ?? "").not.toContain(
+      "peer-to-peer A2A traffic surfaces in the Comms tab",
+    );
+    // The panel hit its activity endpoint.
+    expect(getSpy).toHaveBeenCalledWith(
+      expect.stringContaining(`/workspaces/${mockAgentId}/activity`),
+    );
+  });
+
+  it("renders a peer message on the Agent Comms tab", async () => {
+    const getSpy = vi.spyOn(api, "get");
+    getSpy.mockResolvedValueOnce({ messages: [], reached_end: true });
+    // a2a_receive from a peer → AgentCommsPanel.toCommMessage maps it
+    // to an inbound bubble with the request text.
+    getSpy.mockResolvedValueOnce([
+      {
+        id: "act-1",
+        activity_type: "a2a_receive",
+        source_id: "peer-ws-uuid",
+        target_id: mockAgentId,
+        method: "message/send",
+        summary: "peer asked something",
+        request_body: { task: "Please review PR 42" },
+        response_body: null,
+        status: "ok",
+        created_at: new Date().toISOString(),
+      },
+    ]);
+
+    let rr: ReturnType<typeof renderChat>;
+    await act(async () => {
+      rr = renderChat(mockAgentId);
+    });
+    const { container } = rr!;
+
+    const commsTab = Array.from(container.querySelectorAll("button")).find(
+      (b) => b.textContent?.trim() === "Agent Comms",
+    );
+    await act(async () => {
+      commsTab!.click();
+    });
+
+    await waitFor(() => {
+      expect(container.textContent ?? "").toContain("Please review PR 42");
+    });
+  });
+});
@@ -93,6 +93,24 @@ describe("MobileSpawn — render", () => {
    expect(input).toBeTruthy();
  });

+  // Regression #224 / #225: the agent-name input must render with a
+  // font-size ≥ 16px. iOS Safari and PWAs auto-zoom the viewport when a
+  // focused input has a computed font-size below 16px — the layout
+  // jumps and the page looks broken until the user pinches back.
+  it("renders the name input at font-size 16px or greater (iOS focus-zoom regression)", () => {
+    apiGetSpy.mockResolvedValue(mockTemplates);
+    render(<MobileSpawn dark={true} onClose={vi.fn()} />);
+    const input = document.querySelector(
+      'input[aria-label="Agent name"]',
+    ) as HTMLInputElement | null;
+    expect(input).toBeTruthy();
+    // Parse the inline style font-size — jsdom doesn't run a layout
+    // engine, so getComputedStyle reports the inline value verbatim.
+    const fs = Number.parseFloat(input!.style.fontSize);
+    expect(Number.isFinite(fs)).toBe(true);
+    expect(fs).toBeGreaterThanOrEqual(16);
+  });
+
  it("renders all 4 tier buttons", () => {
    apiGetSpy.mockResolvedValue(mockTemplates);
    render(<MobileSpawn dark={true} onClose={vi.fn()} />);
@@ -133,6 +133,7 @@ export function TabBar({
            aria-label={t.label}
            onClick={() => onChange(t.id)}
            onKeyDown={(e) => handleKeyDown(e, idx)}
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            style={{
              background: "none",
              border: "none",
@@ -291,6 +292,7 @@ export function AgentCard({
      data-testid="workspace-card"
      aria-label={`${agent.name}, status: ${agent.status}, tier ${agent.tier}${agent.remote ? ", remote" : ""}`}
      onClick={onClick}
+      className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
      style={{
        display: "block",
        width: "100%",
@@ -444,6 +446,7 @@ export function FilterChips({
            type="button"
            aria-checked={on}
            onClick={() => onChange(o.id)}
+            className="focus:outline-none focus-visible:ring-2 focus-visible:ring-emerald-500 focus-visible:ring-offset-2 focus-visible:ring-offset-zinc-100 dark:focus-visible:ring-offset-zinc-900"
            style={{
              display: "inline-flex",
              alignItems: "center",
@@ -160,14 +160,14 @@ export function OrgTokensTab() {
            </code>
            <button
              onClick={handleCopy}
-              className="shrink-0 px-2 py-1.5 bg-emerald-800/40 hover:bg-emerald-700/50 border border-emerald-700/40 rounded text-[10px] text-good transition-colors"
+              className="shrink-0 px-2 py-1.5 bg-emerald-800/40 hover:bg-emerald-700/50 border border-emerald-700/40 rounded text-[10px] text-good transition-colors focus:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1"
            >
              {copied ? 'Copied' : 'Copy'}
            </button>
          </div>
          <button
            onClick={() => setNewToken(null)}
-            className="text-[9px] text-good/60 hover:text-good transition-colors"
+            className="text-[9px] text-good/60 hover:text-good transition-colors focus:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1"
          >
            Dismiss
          </button>
@@ -219,7 +219,7 @@ export function OrgTokensTab() {
              </div>
              <button
                onClick={() => setRevokeTarget(t)}
-                className="text-[10px] text-bad/70 hover:text-bad transition-colors px-2 py-1 shrink-0"
+                className="text-[10px] text-bad/70 hover:text-bad transition-colors px-2 py-1 shrink-0 focus:outline-none focus-visible:ring-2 focus-visible:ring-red-400 focus-visible:ring-offset-1"
              >
                Revoke
              </button>
@@ -3,16 +3,24 @@ import { useState, useCallback, useRef, useEffect } from 'react';
 import type { Secret, SecretGroup } from '@/types/secrets';
 import { useSecretsStore } from '@/stores/secrets-store';
 import { StatusBadge } from '@/components/ui/StatusBadge';
-import { RevealToggle } from '@/components/ui/RevealToggle';
 import { KeyValueField } from '@/components/ui/KeyValueField';
 import { ValidationHint } from '@/components/ui/ValidationHint';
 import { TestConnectionButton } from '@/components/ui/TestConnectionButton';
 import { validateSecretValue } from '@/lib/validation/secret-formats';
 import { SERVICES } from '@/lib/services';

-const AUTO_HIDE_MS = 30_000;
 const VALIDATION_DEBOUNCE_MS = 400;

+// Secret values are write-only from the browser: the server List endpoint
+// "Never exposes values", there is no per-secret decrypt route, and the
+// only decrypted path (GET /secrets/values) is bulk + token-gated for
+// remote agents. The old eye/RevealToggle was a dead affordance — it
+// flipped its own icon but could never reveal anything, which read as
+// "this doesn't work" (esp. once clicked → eye-with-slash). We show an
+// honest static indicator instead; rotation is via Edit.
+const WRITE_ONLY_TITLE =
+  'Value is write-only and cannot be revealed — use Edit to replace/rotate it';
+
 interface SecretRowProps {
  secret: Secret;
  workspaceId: string;
@@ -31,28 +39,12 @@ export function SecretRow({ secret, workspaceId }: SecretRowProps) {
  const setSecretStatus = useSecretsStore((s) => s.setSecretStatus);

  const isEditing = editingKey === secret.name;
-  const [revealed, setRevealed] = useState(false);
  const [editValue, setEditValue] = useState('');
  const [validationError, setValidationError] = useState<string | null>(null);
  const [isSaving, setIsSaving] = useState(false);
  const [saveError, setSaveError] = useState<string | null>(null);
  const debounceRef = useRef<ReturnType<typeof setTimeout>>(undefined);
  const editBtnRef = useRef<HTMLButtonElement>(null);
-  const revealTimerRef = useRef<ReturnType<typeof setTimeout>>(undefined);
-
-  // Auto-hide revealed value after 30s
-  useEffect(() => {
-    if (revealed) {
-      clearTimeout(revealTimerRef.current);
-      revealTimerRef.current = setTimeout(() => setRevealed(false), AUTO_HIDE_MS);
-      return () => clearTimeout(revealTimerRef.current);
-    }
-  }, [revealed]);
-
-  // Reset revealed state when panel closes (session-only)
-  useEffect(() => {
-    return () => setRevealed(false);
-  }, []);

  // Debounced validation
  useEffect(() => {
@@ -133,11 +125,15 @@ export function SecretRow({ secret, workspaceId }: SecretRowProps) {
          {secret.masked_value}
        </span>
        <div className="secret-row__actions">
-          <RevealToggle
-            revealed={revealed}
-            onToggle={() => setRevealed((r) => !r)}
-            label={`Toggle reveal ${secret.name}`}
-          />
+          <span
+            data-testid="write-only-indicator"
+            className="secret-row__write-only"
+            role="img"
+            aria-label={`${secret.name} value is write-only and cannot be revealed; use Edit to replace it`}
+            title={WRITE_ONLY_TITLE}
+          >
+            🔒
+          </span>
          <StatusBadge status={secret.status} />
          <button
            type="button"
@@ -16,7 +16,40 @@ interface TokensTabProps {
  workspaceId: string;
 }

+// The settings panel passes the literal sentinel "global" when no canvas
+// node is selected. Workspace tokens are inherently per-workspace — there
+// is no /workspaces/global/tokens endpoint (querying the uuid column with
+// "global" 500s on Postgres). The org-wide equivalent lives in the
+// separate "Org API Keys" tab. Mirrors the sentinel-awareness that
+// api/secrets.ts already has (workspaceId === 'global' → /settings/secrets).
+const GLOBAL_WORKSPACE_ID = 'global';
+
 export function TokensTab({ workspaceId }: TokensTabProps) {
+  if (workspaceId === GLOBAL_WORKSPACE_ID) {
+    return (
+      <div className="p-4 space-y-4">
+        <div>
+          <h3 className="text-sm font-semibold text-ink">API Tokens</h3>
+          <p className="text-[10px] text-ink-mid mt-0.5">
+            Bearer tokens for authenticating API calls to this workspace.
+          </p>
+        </div>
+        <div className="text-center py-6">
+          <p className="text-xs text-ink-mid">Select a workspace node first</p>
+          <p className="text-[10px] text-ink-mid mt-1">
+            Workspace tokens are scoped to a single workspace. Select a node
+            on the canvas to manage its tokens, or use the{' '}
+            <span className="text-accent font-medium">Org API Keys</span> tab
+            for org-wide API keys.
+          </p>
+        </div>
+      </div>
+    );
+  }
+  return <WorkspaceTokensTab workspaceId={workspaceId} />;
+}
+
+function WorkspaceTokensTab({ workspaceId }: TokensTabProps) {
  const [tokens, setTokens] = useState<Token[]>([]);
  const [loading, setLoading] = useState(true);
  const [creating, setCreating] = useState(false);
@@ -107,14 +140,14 @@ export function TokensTab({ workspaceId }: TokensTabProps) {
            </code>
            <button
              onClick={handleCopy}
-              className="shrink-0 px-2 py-1.5 bg-emerald-800/40 hover:bg-emerald-700/50 border border-emerald-700/40 rounded text-[10px] text-good transition-colors"
+              className="shrink-0 px-2 py-1.5 bg-emerald-800/40 hover:bg-emerald-700/50 border border-emerald-700/40 rounded text-[10px] text-good transition-colors focus:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1"
            >
              {copied ? 'Copied' : 'Copy'}
            </button>
          </div>
          <button
            onClick={() => setNewToken(null)}
-            className="text-[9px] text-good/60 hover:text-good transition-colors"
+            className="text-[9px] text-good/60 hover:text-good transition-colors focus:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1"
          >
            Dismiss
          </button>
@@ -159,7 +192,7 @@ export function TokensTab({ workspaceId }: TokensTabProps) {
              </div>
              <button
                onClick={() => setRevokeTarget(t)}
-                className="text-[10px] text-bad/70 hover:text-bad transition-colors px-2 py-1"
+                className="text-[10px] text-bad/70 hover:text-bad transition-colors px-2 py-1 focus:outline-none focus-visible:ring-2 focus-visible:ring-red-400 focus-visible:ring-offset-1"
              >
                Revoke
              </button>
@@ -138,14 +138,54 @@ describe("SecretRow — display mode", () => {
    expect(document.querySelector('[role="row"]')).toBeTruthy();
  });

-  it("has Reveal, Copy, Edit, Delete buttons", () => {
+  it("has Copy, Edit, Delete buttons", () => {
    render(<SecretRow secret={GITHUB_SECRET} workspaceId="ws-1" />);
-    expect(screen.getByTestId("reveal-toggle")).toBeTruthy();
    expect(screen.getByRole("button", { name: /copy/i })).toBeTruthy();
    expect(screen.getByRole("button", { name: /edit/i })).toBeTruthy();
    expect(screen.getByRole("button", { name: /delete/i })).toBeTruthy();
  });

+  // Regression: the reveal/eye control was a dead affordance. Clicking it
+  // flipped its own icon (eye → eye-with-slash) but never revealed the value,
+  // because secret values are write-only from the browser (server List
+  // "Never exposes values"; there is no per-secret decrypt endpoint and the
+  // client has no plaintext-fetch function). The honest fix removes the
+  // toggle and shows a static "write-only / cannot be revealed" indicator.
+  // See internal tracking issue + internal#210/#211.
+  it("does NOT render a reveal/eye toggle (values are write-only)", () => {
+    render(<SecretRow secret={GITHUB_SECRET} workspaceId="ws-1" />);
+    expect(screen.queryByTestId("reveal-toggle")).toBeNull();
+    expect(
+      screen.queryByRole("button", { name: /toggle reveal/i }),
+    ).toBeNull();
+  });
+
+  it("shows a write-only indicator explaining the value cannot be revealed", () => {
+    render(<SecretRow secret={ANTHROPIC_SECRET} workspaceId="ws-1" />);
+    const indicator = screen.getByTestId("write-only-indicator");
+    expect(indicator).toBeTruthy();
+    // Affordance must be honest: explain it cannot be revealed and that
+    // Edit is the rotate path. It must not be a clickable button.
+    const title = indicator.getAttribute("title") ?? "";
+    expect(title.toLowerCase()).toMatch(/write-only|cannot be revealed/);
+    expect(indicator.tagName).not.toBe("BUTTON");
+  });
+
+  it("write-only indicator is present for the Anthropic/OAuth-token row too", () => {
+    // The reported bug singled out CLAUDE_CODE_OAUTH_TOKEN (anthropic group);
+    // the fix is group-agnostic — every row gets the same honest affordance.
+    const OAUTH_SECRET = {
+      name: "CLAUDE_CODE_OAUTH_TOKEN",
+      masked_value: "••••••••••••••••9d2a",
+      group: "anthropic" as const,
+      status: "unverified" as const,
+      updated_at: "2024-01-04",
+    };
+    render(<SecretRow secret={OAUTH_SECRET} workspaceId="ws-1" />);
+    expect(screen.queryByTestId("reveal-toggle")).toBeNull();
+    expect(screen.getByTestId("write-only-indicator")).toBeTruthy();
+  });
+
  it("shows invalid status correctly", () => {
    render(<SecretRow secret={CUSTOM_SECRET} workspaceId="ws-1" />);
    expect(screen.getByTestId("status-badge").getAttribute("data-status")).toBe("invalid");
@@ -302,3 +302,35 @@ describe("TokensTab — error", () => {
    expect(document.querySelector('[role="status"]')).toBeNull();
  });
 });
+
+// ─── "global" sentinel (no node selected) ────────────────────────────────────
+//
+// Regression: SettingsPanel passes the literal "global" when no canvas
+// node is selected. workspace tokens are per-workspace and there is no
+// /workspaces/global/tokens endpoint — calling it 500'd
+// ("invalid input syntax for type uuid: global"). The tab must NOT call
+// the API in that state and must point the user at the Org API Keys tab.
+describe("TokensTab — global sentinel (no node selected)", () => {
+  beforeEach(() => {
+    mockApiGet.mockReset();
+    mockApiPost.mockReset();
+    mockApiGet.mockRejectedValue(new Error("should not be called"));
+  });
+
+  it("does not call the API and shows a pointer to Org API Keys", async () => {
+    render(<TokensTab workspaceId="global" />);
+    await flush();
+    expect(mockApiGet).not.toHaveBeenCalled();
+    expect(mockApiPost).not.toHaveBeenCalled();
+    expect(document.body.textContent).toContain("Select a workspace node");
+    expect(document.body.textContent).toContain("Org API Keys");
+    // No error banner, no scary 500 surfacing.
+    expect(document.querySelector(".text-bad")).toBeNull();
+  });
+
+  it("has no create button in the global state", async () => {
+    render(<TokensTab workspaceId="global" />);
+    await flush();
+    expect(document.body.textContent).not.toContain("New Token");
+  });
+});
@@ -185,7 +185,7 @@ export function ActivityTab({ workspaceId }: Props) {
      {/* Activity list */}
      <div className="flex-1 overflow-y-auto p-3 space-y-1.5">
        {loading && activities.length === 0 && (
-          <div className="text-xs text-ink-mid text-center py-8">Loading activity...</div>
+          <div role="status" aria-live="polite" className="text-xs text-ink-mid text-center py-8">Loading activity...</div>
        )}

        {error && (
@@ -262,7 +262,7 @@ export function ChannelsTab({ workspaceId }: Props) {
      </div>

      {error && (
-        <div className="px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
+        <div role="alert" aria-live="assertive" className="px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
          {error}
        </div>
      )}
@@ -10,6 +10,7 @@ import { downloadChatFile, isPlatformAttachment } from "./chat/uploads";
 import { PendingAttachmentPill } from "./chat/AttachmentViews";
 import { AttachmentPreview } from "./chat/AttachmentPreview";
 import { AgentCommsPanel } from "./chat/AgentCommsPanel";
+import { ChatErrorBanner } from "./chat/ChatErrorBanner";
 import { appendActivityLine } from "./chat/activityLog";
 import { runtimeDisplayName } from "@/lib/runtime-names";
 import { ConfirmDialog } from "@/components/ConfirmDialog";
@@ -592,22 +593,19 @@ function MyChatPanel({ workspaceId, data }: Props) {
        <div ref={bottomRef} />
      </div>

-      {/* Error banner */}
-      {displayError && (
-        <div className="px-3 py-2 bg-red-900/20 border-t border-red-800/30">
-          <div className="flex items-center justify-between">
-            <span className="text-[10px] text-red-300">{displayError}</span>
-            {!isOnline && (
-              <button
-                onClick={() => setConfirmRestart(true)}
-                className="text-[11px] px-2 py-0.5 bg-red-800 text-red-200 rounded hover:bg-red-700"
-              >
-                Restart
-              </button>
-            )}
-          </div>
-        </div>
-      )}
+      {/* Error banner — internal#212: surfaces the secret-safe
+          actionable failure reason that ws-server places on
+          ACTIVITY_LOGGED.error_detail (propagated via
+          useChatSocket → onSendError → setError) and offers a
+          "View activity log" affordance that navigates the user to
+          the Activity tab where the full row lives. The previous
+          inline JSX hardcoded "see workspace logs for details" with
+          no link — there is no separate Logs tab. */}
+      <ChatErrorBanner
+        message={displayError}
+        isOnline={isOnline}
+        onRestart={() => setConfirmRestart(true)}
+      />

      {/* Input */}
      <div className="p-3 border-t border-line">
@@ -81,7 +81,7 @@ function AgentCardSection({ workspaceId }: { workspaceId: string }) {
            spellCheck={false} rows={12}
            className="w-full bg-surface-card border border-line rounded p-2 text-[10px] font-mono text-ink focus:outline-none focus:border-accent resize-none"
          />
-          {error && <div className="px-2 py-1 bg-red-900/30 border border-red-800 rounded text-[10px] text-bad">{error}</div>}
+          {error && <div role="alert" aria-live="assertive" className="px-2 py-1 bg-red-900/30 border border-red-800 rounded text-[10px] text-bad">{error}</div>}
          <div className="flex gap-2">
            <button type="button" onClick={handleSave} disabled={saving}
              className="px-2 py-1 bg-accent hover:bg-accent-strong text-[10px] rounded text-white disabled:opacity-50 transition-colors focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1 focus-visible:ring-offset-surface">
@@ -109,6 +109,130 @@ function AgentCardSection({ workspaceId }: { workspaceId: string }) {
  );
 }

+// --- Agent Abilities Section ---
+//
+// Always-visible on/off controls for the two workspace-level ability flags
+// (broadcast_enabled, talk_to_user_enabled). Both are mutated through the
+// same admin endpoint the ChatTab recovery banner already uses
+// (PATCH /workspaces/:id/abilities) and reflected into the canvas store node
+// data (broadcastEnabled / talkToUserEnabled) so every surface that reads
+// useCanvasStore.nodes stays consistent without a full re-hydrate.
+//
+// Before this section there was NO canvas control for either flag: the
+// backend was fully wired (workspace_abilities.go / workspace_broadcast.go /
+// agent_message_writer.go, see commit 29b4bffb + internal#510/#511) but the
+// only frontend affordance was the ChatTab recovery banner, which renders
+// solely when talk_to_user_enabled===false and so is invisible under the
+// TRUE default and never existed at all for broadcast.
+function AgentAbilitiesSection({ workspaceId }: { workspaceId: string }) {
+  // Read the live ability flags off the canvas store node — the platform
+  // event stream hydrates these (canvas-topology.ts maps the workspace row's
+  // broadcast_enabled/talk_to_user_enabled onto node data), so this stays in
+  // sync with the recovery banner and avoids a duplicate GET. Mirrors the
+  // store-read pattern used by AgentCardSection above.
+  const node = useCanvasStore((s) =>
+    s.nodes?.find?.((n) => n.id === workspaceId),
+  );
+  // Defaults match the backend column defaults + canvas-topology mapping:
+  // broadcast_enabled defaults FALSE, talk_to_user_enabled defaults TRUE.
+  const broadcastEnabled = node?.data.broadcastEnabled ?? false;
+  const talkToUserEnabled = node?.data.talkToUserEnabled ?? true;
+
+  // Track an in-flight PATCH per field so a double-click can't fire two
+  // racing writes, and surface a one-line error if the server rejects.
+  const [pending, setPending] = useState<null | "broadcast" | "talk">(null);
+  const [error, setError] = useState<string | null>(null);
+
+  const patchAbility = async (
+    which: "broadcast" | "talk",
+    body: { broadcast_enabled: boolean } | { talk_to_user_enabled: boolean },
+    optimistic: Partial<{ broadcastEnabled: boolean; talkToUserEnabled: boolean }>,
+  ) => {
+    setError(null);
+    setPending(which);
+    // Optimistic store update — the toggle flips immediately; on failure we
+    // roll back to the server-truth value the store last held.
+    const prev = {
+      broadcastEnabled,
+      talkToUserEnabled,
+    };
+    useCanvasStore.getState().updateNodeData(workspaceId, optimistic);
+    try {
+      await api.patch(`/workspaces/${workspaceId}/abilities`, body);
+    } catch (e) {
+      // Roll back the optimistic change to last-known server truth.
+      useCanvasStore.getState().updateNodeData(workspaceId, {
+        broadcastEnabled: prev.broadcastEnabled,
+        talkToUserEnabled: prev.talkToUserEnabled,
+      });
+      setError(
+        e instanceof Error ? e.message : "Failed to update ability — try again",
+      );
+    } finally {
+      setPending(null);
+    }
+  };
+
+  return (
+    <Section title="Agent Abilities">
+      <p className="text-[10px] text-ink-mid px-1 pb-1">
+        Workspace-level permissions for this agent. Changes apply immediately
+        (no restart required).
+      </p>
+      <div className="space-y-2">
+        <div>
+          <Toggle
+            label="Talk to user"
+            checked={talkToUserEnabled}
+            onChange={(v) =>
+              pending
+                ? undefined
+                : patchAbility(
+                    "talk",
+                    { talk_to_user_enabled: v },
+                    { talkToUserEnabled: v },
+                  )
+            }
+          />
+          <p className="text-[10px] text-ink-mid mt-0.5 ml-6">
+            When off, the agent&apos;s <code className="font-mono">send_message_to_user</code>{" "}
+            and <code className="font-mono">POST /notify</code> calls are
+            rejected (403) — it must route updates through a parent workspace.
+          </p>
+        </div>
+        <div>
+          <Toggle
+            label="Broadcast to peers"
+            checked={broadcastEnabled}
+            onChange={(v) =>
+              pending
+                ? undefined
+                : patchAbility(
+                    "broadcast",
+                    { broadcast_enabled: v },
+                    { broadcastEnabled: v },
+                  )
+            }
+          />
+          <p className="text-[10px] text-ink-mid mt-0.5 ml-6">
+            When on, the agent may <code className="font-mono">POST /broadcast</code>{" "}
+            to message all non-removed agent workspaces in the org. Off by
+            default — only privileged orchestrators should hold this.
+          </p>
+        </div>
+      </div>
+      {pending && (
+        <div className="mt-2 text-[10px] text-ink-mid">Saving…</div>
+      )}
+      {error && (
+        <div role="alert" aria-live="assertive" className="mt-2 px-2 py-1 bg-red-900/30 border border-red-800 rounded text-[10px] text-bad">
+          {error}
+        </div>
+      )}
+    </Section>
+  );
+}
+
 // --- Main ConfigTab ---

 interface ModelSpec {
@@ -795,6 +919,7 @@ export function ConfigTab({ workspaceId }: Props) {
                  <label className="text-[10px] text-ink-mid block mb-1">Model</label>
                  <input
                    type="text"
+                    aria-label="Model"
                    value={currentModelId}
                    onChange={(e) => {
                      const v = e.target.value;
@@ -885,6 +1010,8 @@ export function ConfigTab({ workspaceId }: Props) {
            )}
          </Section>

+          <AgentAbilitiesSection workspaceId={workspaceId} />
+
          {/* Claude Settings — shown for claude-code runtime or claude/anthropic model names */}
          {(config.runtime === "claude-code" ||
            (config.runtime_config?.model || config.model || "").toLowerCase().includes("claude") ||
@@ -995,7 +1122,7 @@ export function ConfigTab({ workspaceId }: Props) {
      )}

      {error && (
-        <div className="mx-3 mb-2 px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">{error}</div>
+        <div role="alert" aria-live="assertive" className="mx-3 mb-2 px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">{error}</div>
      )}
      {!error && RUNTIMES_WITH_OWN_CONFIG.has(config.runtime || "") && (
        <div className="mx-3 mb-2 px-3 py-1.5 bg-surface-sunken/50 border border-line rounded text-xs text-ink-mid">
@@ -157,7 +157,7 @@ export function DetailsTab({ workspaceId, data }: Props) {
              </select>
            </Field>
            {saveError && (
-              <div className="px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
+              <div role="alert" aria-live="assertive" className="px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
                {saveError}
              </div>
            )}
@@ -203,7 +203,7 @@ export function DetailsTab({ workspaceId, data }: Props) {
            {isRestartable && (
              <div className="pt-2">
                {restartError && (
-                  <div className="mb-2 px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
+                  <div role="alert" aria-live="assertive" className="mb-2 px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
                    {restartError}
                  </div>
                )}
@@ -307,7 +307,7 @@ export function DetailsTab({ workspaceId, data }: Props) {
      {/* Delete */}
      <Section title="Danger Zone">
        {deleteError && (
-          <div className="mb-2 px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
+          <div role="alert" aria-live="assertive" className="mb-2 px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
            {deleteError}
          </div>
        )}
@@ -82,7 +82,7 @@ export function EventsTab({ workspaceId }: Props) {
      </div>

      {error && (
-        <div className="px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
+        <div role="alert" aria-live="assertive" className="px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
          {error}
        </div>
      )}
@@ -102,7 +102,7 @@ export function ExternalConnectionSection({ workspaceId }: Props) {
      </div>

      {error && (
-        <div className="mt-2 px-2 py-1 bg-red-900/30 border border-red-800 rounded text-[10px] text-bad">
+        <div role="alert" aria-live="assertive" className="mt-2 px-2 py-1 bg-red-900/30 border border-red-800 rounded text-[10px] text-bad">
          {error}
        </div>
      )}
@@ -45,11 +45,54 @@ export function FilesTab({ workspaceId, data }: Props) {
  if (data && isExternalLikeRuntime(data.runtime)) {
    return <NotAvailablePanel runtime={data.runtime} />;
  }
-  return <PlatformOwnedFilesTab workspaceId={workspaceId} />;
+  return <PlatformOwnedFilesTab workspaceId={workspaceId} runtime={data?.runtime} />;
 }

-function PlatformOwnedFilesTab({ workspaceId }: { workspaceId: string }) {
-  const [root, setRoot] = useState("/configs");
+/** Picks the initial root for the FilesTab dropdown based on the
+ *  workspace's runtime. Decision: per-runtime default (Hongming
+ *  2026-05-15, internal#425 Decisions §2).
+ *
+ *  - openclaw → `/agent-home` (the agent's identity/state — the
+ *    user-facing interesting files for that runtime live in
+ *    `~/.openclaw/` inside the container, which `/agent-home` maps to
+ *    via the Phase 2b docker-exec backend).
+ *  - everything else (claude-code, hermes, external-like, undefined)
+ *    → `/configs` (the legacy default — managed config that flows
+ *    through the per-runtime indirection in
+ *    workspace-server/internal/handlers/template_files_eic.go).
+ *
+ *  When the runtime is undefined (legacy callers that don't thread
+ *  `data` through, or a workspace whose runtime field hasn't loaded
+ *  yet) the default is `/configs` — matches today's behaviour, no
+ *  surprise.
+ *
+ *  Note on `/agent-home` pre-Phase-2b: the backend short-circuits
+ *  with HTTP 501 and the canonical "implementation pending" body.
+ *  The tab renders empty + the error banner explains. This is by
+ *  design — lets us land the canvas UX before the backend ships,
+ *  per the RFC's phased rollout. The 501 is graceful: it doesn't
+ *  poison error toasts or generate "workspace not found" noise.
+ *
+ *  Adding a new runtime that should default to `/agent-home`: add it
+ *  to the agentHomeDefaultRuntimes set below. Adding a runtime that
+ *  should default to a different root: extend this function. */
+const agentHomeDefaultRuntimes = new Set(["openclaw"]);
+
+function defaultRootForRuntime(runtime: string | undefined): string {
+  if (runtime && agentHomeDefaultRuntimes.has(runtime)) {
+    return "/agent-home";
+  }
+  return "/configs";
+}
+
+function PlatformOwnedFilesTab({
+  workspaceId,
+  runtime,
+}: {
+  workspaceId: string;
+  runtime?: string;
+}) {
+  const [root, setRoot] = useState(() => defaultRootForRuntime(runtime));
  const [selectedFile, setSelectedFile] = useState<string | null>(null);
  const [fileContent, setFileContent] = useState("");
  const [editContent, setEditContent] = useState("");
@@ -223,7 +266,7 @@ function PlatformOwnedFilesTab({ workspaceId }: { workspaceId: string }) {
        // immediately. Delete-All hovers DARKER (bg-red-700) — same AA
        // contrast trap that bit ConfirmDialog/ApprovalBanner. Cancel
        // lifts to surface-elevated instead of the prior no-op hover.
-        <div role="alertdialog" aria-labelledby="files-delete-all-msg" className="mx-3 mt-2 px-3 py-2 bg-red-950/30 border border-red-800/40 rounded space-y-1.5">
+        <div role="alertdialog" aria-modal="false" aria-labelledby="files-delete-all-msg" className="mx-3 mt-2 px-3 py-2 bg-red-950/30 border border-red-800/40 rounded space-y-1.5">
          <p id="files-delete-all-msg" className="text-xs text-bad">Delete all {files.filter((f) => !f.dir).length} files? This cannot be undone.</p>
          <div className="flex gap-2">
            <button type="button" onClick={() => { handleDeleteAll(); setShowDeleteAll(false); }} className="px-2 py-0.5 bg-red-700 hover:bg-red-600 text-[10px] rounded text-white transition-colors focus:outline-none focus-visible:ring-2 focus-visible:ring-red-500/60 focus-visible:ring-offset-1 focus-visible:ring-offset-surface">Delete All</button>
@@ -237,7 +280,7 @@ function PlatformOwnedFilesTab({ workspaceId }: { workspaceId: string }) {
      )}

      {confirmDelete && (
-        <div role="alertdialog" aria-labelledby="files-delete-one-msg" className="mx-3 mt-2 px-3 py-2 bg-amber-950/30 border border-amber-800/40 rounded space-y-1.5">
+        <div role="alertdialog" aria-modal="false" aria-labelledby="files-delete-one-msg" className="mx-3 mt-2 px-3 py-2 bg-amber-950/30 border border-amber-800/40 rounded space-y-1.5">
          <p id="files-delete-one-msg" className="text-xs text-warm">Delete <span className="font-mono">{confirmDelete}</span>{files.find((f) => f.path === confirmDelete && f.dir) ? " and all its contents" : ""}?</p>
          <div className="flex gap-2">
            <button type="button" onClick={confirmDeleteFile} className="px-2 py-0.5 bg-red-700 hover:bg-red-600 text-[10px] rounded text-white transition-colors focus:outline-none focus-visible:ring-2 focus-visible:ring-red-500/60 focus-visible:ring-offset-1 focus-visible:ring-offset-surface">Delete</button>
@@ -3,6 +3,22 @@
 import { useRef } from "react";
 import { getIcon } from "./tree";

+// secretShapeMarker is the canonical body the workspace-server Files
+// API returns when a file's path OR content matched a credential
+// regex (internal#425 RFC, Phase 2b — backed by
+// workspace-server/internal/secrets.ScanBytes). The marker is a
+// fixed prefix so the canvas can detect it without parsing JSON and
+// without round-tripping the matched bytes through the editor (which
+// would defeat the purpose — clipboard, browser history, log
+// surfaces would all see them).
+//
+// Today (Phase 1 / before 2b ships) the backend returns 501 for the
+// only root that uses this path, so the marker is dead code until
+// 2b lands. Wiring it in now keeps the canvas + backend contracts
+// aligned in one PR rather than a follow-up. The constant is
+// importable so a future test can pin the exact string.
+export const SECRET_SHAPE_DENIED_MARKER = "<denied: secret-shape>";
+
 interface Props {
  selectedFile: string | null;
  fileContent: string;
@@ -31,6 +47,22 @@ export function FileEditor({
  const editorRef = useRef<HTMLTextAreaElement>(null);
  const isDirty = editContent !== fileContent;

+  // internal#425 Phase 3: detect the secret-shape denial marker and
+  // render a placeholder instead of the editor. The marker comes
+  // from workspace-server Phase 2b (secrets.ScanBytes) which refuses
+  // to surface the file's bytes. We deliberately don't expose
+  // the matched pattern's Name here — the canvas just shows the
+  // generic denial. The Files API log surface has the Pattern.Name
+  // for operators who need to debug a false positive.
+  const isSecretShapeDenied = fileContent === SECRET_SHAPE_DENIED_MARKER;
+
+  // /agent-home is read-only from the canvas (Phase 2b ships read +
+  // delete; Phase-2b-followup may add write). Edits to /configs are
+  // unchanged. Until 2b ships, /agent-home returns 501 so this
+  // read-only gate is also dead code, but wiring it in now keeps
+  // the UI honest the moment 2b lands without a follow-up canvas PR.
+  const isReadOnlyRoot = root !== "/configs";
+
  if (!selectedFile) {
    return (
      <div className="flex-1 flex items-center justify-center">
@@ -75,11 +107,42 @@ export function FileEditor({
      {/* Editor area */}
      {loadingFile ? (
        <div className="p-4 text-xs text-ink-mid">Loading...</div>
+      ) : isSecretShapeDenied ? (
+        // Files API refused to surface this file's bytes because its
+        // path or content matched a credential regex
+        // (workspace-server/internal/secrets, internal#425 Phase 2b).
+        // We render a placeholder INSTEAD OF the textarea so the
+        // matched bytes never enter the DOM. Clipboard / view-source
+        // / element-inspector all see the placeholder, not the
+        // credential.
+        <div
+          role="region"
+          aria-label="File content denied"
+          className="flex-1 flex items-center justify-center p-6 bg-surface"
+        >
+          <div className="max-w-md text-center space-y-2">
+            <div className="text-2xl opacity-40">🛡️</div>
+            <p className="text-[11px] font-mono text-warm">
+              {SECRET_SHAPE_DENIED_MARKER}
+            </p>
+            <p className="text-[10px] text-ink-mid leading-relaxed">
+              The platform refused to surface this file because its
+              path or content matched a credential-shape pattern.
+              The bytes never left the workspace container.
+            </p>
+            <p className="text-[10px] text-ink-mid leading-relaxed">
+              If this is a false positive (test fixture, docs example,
+              or content that happens to share a credential's shape),
+              rename the file or adjust the content via the workspace
+              terminal so the regex no longer matches, then refresh.
+            </p>
+          </div>
+        </div>
      ) : (
        <textarea
          ref={editorRef}
          value={editContent}
-          readOnly={root !== "/configs"}
+          readOnly={isReadOnlyRoot}
          onChange={(e) => setEditContent(e.target.value)}
          onKeyDown={(e) => {
            if ((e.metaKey || e.ctrlKey) && e.key === "s") {
@@ -38,6 +38,15 @@ export function FilesToolbar({
          <option value="/home">/home</option>
          <option value="/workspace">/workspace</option>
          <option value="/plugins">/plugins</option>
+          {/* internal#425 Phase 1+3: container-internal $HOME root.
+              Backend lands the docker-exec dispatch in Phase 2b. Until
+              then the stub returns 501 with a canonical
+              "implementation pending" message — the dropdown renders
+              the option so the canvas affordance is design-frozen
+              even before the backend ships.
+              Runtime-default selection logic in FilesTab.tsx picks
+              this as the initial value for openclaw workspaces. */}
+          <option value="/agent-home">/agent-home</option>
        </select>
        <span className="text-[10px] text-ink-mid">{fileCount} files</span>
      </div>
@@ -0,0 +1,181 @@
+// @vitest-environment jsdom
+/**
+ * Tests for the /agent-home root selector + per-runtime default-root
+ * + secret-shape denial placeholder (internal#425 Phase 3).
+ *
+ * Separate file so the diff is reviewable as a unit and the existing
+ * FilesToolbar / FileEditor / FilesTab tests don't have to grow
+ * agent-home-specific cases. Once Phase 2b lands, the read-only +
+ * 501-stub assertions here can be tightened (or moved into the main
+ * test file as the agent-home root becomes a first-class affordance).
+ */
+import React from "react";
+import { render, screen, cleanup } from "@testing-library/react";
+import { afterEach, describe, expect, it, vi } from "vitest";
+
+import { FilesToolbar } from "../FilesToolbar";
+import {
+  FileEditor,
+  SECRET_SHAPE_DENIED_MARKER,
+} from "../FileEditor";
+
+afterEach(cleanup);
+
+describe("internal#425 Phase 3 — /agent-home root selector", () => {
+  it("dropdown includes /agent-home as an option", () => {
+    // Pins the affordance is in the DOM even pre-Phase-2b — the
+    // canvas design freezes today, the backend lands the dispatch
+    // later. Without this, a future refactor that drops the option
+    // would silently regress the RFC's Phase 1 contract (canvas
+    // visibility) without breaking any other test.
+    render(
+      <FilesToolbar
+        root="/configs"
+        setRoot={vi.fn()}
+        fileCount={0}
+        onNewFile={vi.fn()}
+        onUpload={vi.fn()}
+        onDownloadAll={vi.fn()}
+        onClearAll={vi.fn()}
+        onRefresh={vi.fn()}
+      />,
+    );
+    const select = screen.getByRole("combobox", {
+      name: /file root directory/i,
+    }) as HTMLSelectElement;
+    const values = Array.from(select.options).map((o) => o.value);
+    expect(values).toContain("/agent-home");
+  });
+
+  it("dropdown shows /agent-home as the SELECTED root when prop is /agent-home", () => {
+    render(
+      <FilesToolbar
+        root="/agent-home"
+        setRoot={vi.fn()}
+        fileCount={0}
+        onNewFile={vi.fn()}
+        onUpload={vi.fn()}
+        onDownloadAll={vi.fn()}
+        onClearAll={vi.fn()}
+        onRefresh={vi.fn()}
+      />,
+    );
+    const select = screen.getByRole("combobox", {
+      name: /file root directory/i,
+    }) as HTMLSelectElement;
+    expect(select.value).toBe("/agent-home");
+  });
+});
+
+describe("internal#425 Phase 3 — secret-shape denial placeholder", () => {
+  // Files API Phase 2b returns SECRET_SHAPE_DENIED_MARKER as the file
+  // body when the file's path or content matched a credential regex.
+  // The editor MUST render the marker as a placeholder, not pump it
+  // through the textarea — that would put the marker (and any future
+  // matched bytes if the backend contract changes) into the DOM
+  // value, clipboard, and inspector.
+
+  it("renders the denial placeholder INSTEAD of the textarea when fileContent is the marker", () => {
+    render(
+      <FileEditor
+        selectedFile="agent/.openclaw/secrets.env"
+        fileContent={SECRET_SHAPE_DENIED_MARKER}
+        editContent={SECRET_SHAPE_DENIED_MARKER}
+        setEditContent={vi.fn()}
+        loadingFile={false}
+        saving={false}
+        success={null}
+        root="/agent-home"
+        onSave={vi.fn()}
+        onDownload={vi.fn()}
+      />,
+    );
+    // Placeholder region present
+    expect(
+      screen.getByRole("region", { name: /file content denied/i }),
+    ).toBeTruthy();
+    // Marker text visible (so a debugging operator sees the canonical
+    // contract string without having to dig into the source).
+    expect(screen.getByText(SECRET_SHAPE_DENIED_MARKER)).toBeTruthy();
+    // Critically: NO textarea — the bytes never reach a controlled
+    // input. A regression that re-introduces the textarea path would
+    // make the matched marker (and any future content) selectable +
+    // copyable.
+    expect(screen.queryByRole("textbox")).toBeNull();
+  });
+
+  it("renders the textarea normally when fileContent is regular content", () => {
+    render(
+      <FileEditor
+        selectedFile="config.yaml"
+        fileContent="name: openclaw\n"
+        editContent="name: openclaw\n"
+        setEditContent={vi.fn()}
+        loadingFile={false}
+        saving={false}
+        success={null}
+        root="/configs"
+        onSave={vi.fn()}
+        onDownload={vi.fn()}
+      />,
+    );
+    expect(screen.getByRole("textbox")).toBeTruthy();
+    expect(screen.queryByRole("region", { name: /file content denied/i }))
+      .toBeNull();
+  });
+
+  it("/agent-home renders textarea READ-ONLY for non-denied content", () => {
+    // Phase 2b ships read + delete on /agent-home; write semantics
+    // are decided later. Until then, the canvas presents the editor
+    // as read-only so a user can't type into a buffer that the
+    // backend will refuse to PUT. Without this gate, the user would
+    // edit, hit Save, get a 501, and lose their context for why.
+    render(
+      <FileEditor
+        selectedFile=".openclaw/agent-card.json"
+        fileContent='{"name":"openclaw"}'
+        editContent='{"name":"openclaw"}'
+        setEditContent={vi.fn()}
+        loadingFile={false}
+        saving={false}
+        success={null}
+        root="/agent-home"
+        onSave={vi.fn()}
+        onDownload={vi.fn()}
+      />,
+    );
+    const textarea = screen.getByRole("textbox") as HTMLTextAreaElement;
+    expect(textarea.readOnly).toBe(true);
+  });
+
+  it("/configs renders textarea WRITABLE (regression guard for the read-only gate)", () => {
+    render(
+      <FileEditor
+        selectedFile="config.yaml"
+        fileContent="name: x\n"
+        editContent="name: x\n"
+        setEditContent={vi.fn()}
+        loadingFile={false}
+        saving={false}
+        success={null}
+        root="/configs"
+        onSave={vi.fn()}
+        onDownload={vi.fn()}
+      />,
+    );
+    const textarea = screen.getByRole("textbox") as HTMLTextAreaElement;
+    expect(textarea.readOnly).toBe(false);
+  });
+});
+
+describe("internal#425 Phase 3 — marker constant is the canonical string", () => {
+  // The marker string is part of the canvas <-> workspace-server
+  // contract. The workspace-server emits this exact body; the canvas
+  // detects it by exact-equality. A typo on either side would
+  // silently break detection — the canvas would render the literal
+  // string in the textarea instead of the placeholder. Pin the
+  // contract value here.
+  it("matches the contract value '<denied: secret-shape>'", () => {
+    expect(SECRET_SHAPE_DENIED_MARKER).toBe("<denied: secret-shape>");
+  });
+});
@@ -275,7 +275,7 @@ export function ScheduleTab({ workspaceId }: Props) {
              Enabled
            </label>
          </div>
-          {error && <div className="text-[10px] text-bad">{error}</div>}
+          {error && <div role="alert" aria-live="assertive" className="text-[10px] text-bad">{error}</div>}
          <div className="flex gap-2">
            <button
              type="button"
@@ -67,7 +67,7 @@ export function TracesTab({ workspaceId }: Props) {
      </div>

      {error && (
-        <div className="px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
+        <div role="alert" aria-live="assertive" className="px-3 py-1.5 bg-red-900/30 border border-red-800 rounded text-xs text-bad">
          {error}
        </div>
      )}
@@ -0,0 +1,99 @@
+// @vitest-environment jsdom
+//
+// Pins internal#212 — the chat error banner must:
+//
+//   1. Render the secret-safe failure reason (e.g. the provider's own
+//      "403 oauth_org_not_allowed: ..." string), NOT the opaque
+//      hardcoded "Agent error (Exception) — see workspace logs for
+//      details." that points at a workspace-logs tab that doesn't
+//      exist.
+//
+//   2. Offer a working "View activity log" affordance that navigates
+//      the user to the Activity tab where the full row lives.
+//
+// Tested at the banner-component seam (ChatErrorBanner). The
+// hook-level path is pinned separately by
+// chat/hooks/__tests__/useChatSocket.test.tsx — together they cover
+// wire-payload → callback → render without each test needing to drive
+// the full ChatTab send-state machinery.
+
+import { describe, it, expect, vi, afterEach, beforeEach } from "vitest";
+import { render, screen, cleanup, fireEvent } from "@testing-library/react";
+
+afterEach(cleanup);
+
+const mocks = vi.hoisted(() => ({
+  setPanelTabMock: vi.fn(),
+}));
+
+vi.mock("@/store/canvas", () => {
+  const state = {
+    setPanelTab: mocks.setPanelTabMock,
+    panelTab: "chat",
+  };
+  const hook = (selector?: (s: typeof state) => unknown) =>
+    selector ? selector(state) : state;
+  hook.getState = () => state;
+  return { useCanvasStore: hook };
+});
+
+beforeEach(() => {
+  mocks.setPanelTabMock.mockClear();
+});
+
+import { ChatErrorBanner } from "../chat/ChatErrorBanner";
+
+describe("ChatErrorBanner — surfaces actionable reason (internal#212)", () => {
+  it("renders the secret-safe failure reason verbatim, not a hardcoded opaque message", () => {
+    const reason =
+      "Anthropic 403 oauth_org_not_allowed: Your organization has disabled Claude subscription access for Claude Code — use an Anthropic API key or ask your admin to enable access.";
+    render(<ChatErrorBanner message={reason} isOnline={true} onRestart={() => {}} />);
+    expect(screen.getByText(/oauth_org_not_allowed/i)).toBeDefined();
+    expect(screen.getByText(/disabled Claude subscription access/i)).toBeDefined();
+    // The legacy boilerplate must NOT leak through when a real reason
+    // is provided.
+    expect(screen.queryByText(/see workspace logs for details/i)).toBeNull();
+  });
+
+  it("falls back to the message when it IS the legacy boilerplate (older ws-server)", () => {
+    // Graceful degradation: an older ws-server passes through the
+    // hardcoded text; the banner still renders SOMETHING — never
+    // silently swallow.
+    render(
+      <ChatErrorBanner
+        message="Agent error (Exception) — see workspace logs for details."
+        isOnline={true}
+        onRestart={() => {}}
+      />,
+    );
+    expect(
+      screen.getByText(/Agent error \(Exception\) — see workspace logs for details\./),
+    ).toBeDefined();
+  });
+
+  it("offers a 'View activity log' button that calls setPanelTab('activity')", () => {
+    render(
+      <ChatErrorBanner message="kimi 401 invalid_api_key" isOnline={true} onRestart={() => {}} />,
+    );
+    const btn = screen.getByRole("button", { name: /view activity log/i });
+    fireEvent.click(btn);
+    expect(mocks.setPanelTabMock).toHaveBeenCalledWith("activity");
+  });
+
+  it("still shows the Restart button when offline (existing behavior preserved)", () => {
+    const onRestart = vi.fn();
+    render(
+      <ChatErrorBanner message="Agent is offline" isOnline={false} onRestart={onRestart} />,
+    );
+    const btn = screen.getByRole("button", { name: /^restart$/i });
+    fireEvent.click(btn);
+    expect(onRestart).toHaveBeenCalledTimes(1);
+  });
+
+  it("renders nothing when message is null", () => {
+    const { container } = render(
+      <ChatErrorBanner message={null} isOnline={true} onRestart={() => {}} />,
+    );
+    expect(container.textContent).toBe("");
+  });
+});
@@ -0,0 +1,165 @@
+// @vitest-environment jsdom
+//
+// Tests for the always-visible "Agent Abilities" section added to ConfigTab
+// (internal#510 broadcast_enabled, internal#511 talk_to_user_enabled; backend
+// wired in commit 29b4bffb).
+//
+// Problem this pins: the two workspace ability flags had complete wired
+// backends but NO canvas control — broadcast had none at all, talk-to-user
+// only surfaced as a ChatTab recovery banner that is invisible under its
+// TRUE default. The CTO could not see or toggle either from canvas.
+//
+// What this suite pins:
+//   1. An "Agent Abilities" section renders (always visible, not gated).
+//   2. Both toggles render and reflect the store node's ability fields,
+//      including the asymmetric defaults (broadcast FALSE, talk TRUE).
+//   3. Toggling a switch calls PATCH /workspaces/:id/abilities with the
+//      correct snake_case body and optimistically updates the store.
+
+import { describe, it, expect, vi, afterEach, beforeEach } from "vitest";
+import { render, screen, cleanup, waitFor, fireEvent } from "@testing-library/react";
+import React from "react";
+
+afterEach(cleanup);
+
+const apiGet = vi.fn();
+const apiPatch = vi.fn();
+vi.mock("@/lib/api", () => ({
+  api: {
+    get: (path: string) => apiGet(path),
+    patch: (path: string, body?: unknown) => apiPatch(path, body),
+    put: vi.fn(),
+    post: vi.fn(),
+    del: vi.fn(),
+  },
+}));
+
+// Store node carries the ability flags hydrated by the platform stream
+// (canvas-topology.ts maps broadcast_enabled/talk_to_user_enabled onto
+// node.data). Mirror that shape so the section reads real values.
+const storeUpdateNodeData = vi.fn();
+const storeRestartWorkspace = vi.fn();
+let nodeData: { broadcastEnabled?: boolean; talkToUserEnabled?: boolean } = {};
+const makeState = () => ({
+  nodes: [{ id: "ws-test", data: nodeData }],
+  restartWorkspace: storeRestartWorkspace,
+  updateNodeData: storeUpdateNodeData,
+});
+vi.mock("@/store/canvas", () => ({
+  useCanvasStore: Object.assign(
+    (selector: (s: unknown) => unknown) => selector(makeState()),
+    { getState: () => makeState() },
+  ),
+}));
+
+vi.mock("../AgentCardSection", () => ({
+  AgentCardSection: () => <div data-testid="agent-card-stub" />,
+}));
+
+import { ConfigTab } from "../ConfigTab";
+
+beforeEach(() => {
+  apiGet.mockReset();
+  apiPatch.mockReset();
+  apiPatch.mockResolvedValue({ status: "updated" });
+  storeUpdateNodeData.mockReset();
+  apiGet.mockImplementation((path: string) => {
+    if (path === `/workspaces/ws-test`) {
+      return Promise.resolve({ runtime: "claude-code" });
+    }
+    if (path === `/workspaces/ws-test/model`) {
+      return Promise.resolve({ model: "claude-opus-4-7" });
+    }
+    if (path === `/workspaces/ws-test/provider`) {
+      return Promise.resolve({ provider: "anthropic-oauth", source: "default" });
+    }
+    if (path === `/workspaces/ws-test/files/config.yaml`) {
+      return Promise.resolve({ content: "name: test\nruntime: claude-code\n" });
+    }
+    if (path === "/templates") {
+      return Promise.resolve([
+        { id: "claude-code", name: "Claude Code", runtime: "claude-code", providers: [] },
+      ]);
+    }
+    return Promise.reject(new Error(`unmocked api.get: ${path}`));
+  });
+});
+
+describe("ConfigTab Agent Abilities section", () => {
+  it("renders an always-visible 'Agent Abilities' section with both toggles", async () => {
+    nodeData = {}; // unset → defaults
+    render(<ConfigTab workspaceId="ws-test" />);
+    await waitFor(() => expect(apiGet).toHaveBeenCalled());
+    expect(
+      await screen.findByRole("button", { name: /Agent Abilities/i }),
+    ).toBeTruthy();
+    expect(screen.getByText("Talk to user")).toBeTruthy();
+    expect(screen.getByText("Broadcast to peers")).toBeTruthy();
+  });
+
+  it("reflects the asymmetric defaults: talk-to-user ON, broadcast OFF", async () => {
+    nodeData = {}; // unset → backend defaults
+    render(<ConfigTab workspaceId="ws-test" />);
+    await waitFor(() => expect(apiGet).toHaveBeenCalled());
+    const talk = (await screen.findByText("Talk to user"))
+      .closest("label")!
+      .querySelector("input") as HTMLInputElement;
+    const broadcast = screen
+      .getByText("Broadcast to peers")
+      .closest("label")!
+      .querySelector("input") as HTMLInputElement;
+    expect(talk.checked).toBe(true);
+    expect(broadcast.checked).toBe(false);
+  });
+
+  it("reflects explicit store values", async () => {
+    nodeData = { broadcastEnabled: true, talkToUserEnabled: false };
+    render(<ConfigTab workspaceId="ws-test" />);
+    await waitFor(() => expect(apiGet).toHaveBeenCalled());
+    const talk = (await screen.findByText("Talk to user"))
+      .closest("label")!
+      .querySelector("input") as HTMLInputElement;
+    const broadcast = screen
+      .getByText("Broadcast to peers")
+      .closest("label")!
+      .querySelector("input") as HTMLInputElement;
+    expect(talk.checked).toBe(false);
+    expect(broadcast.checked).toBe(true);
+  });
+
+  it("PATCHes /abilities with talk_to_user_enabled and optimistically updates the store", async () => {
+    nodeData = {}; // talk defaults true
+    render(<ConfigTab workspaceId="ws-test" />);
+    await waitFor(() => expect(apiGet).toHaveBeenCalled());
+    const talk = (await screen.findByText("Talk to user"))
+      .closest("label")!
+      .querySelector("input") as HTMLInputElement;
+    fireEvent.click(talk); // true → false
+    await waitFor(() =>
+      expect(apiPatch).toHaveBeenCalledWith("/workspaces/ws-test/abilities", {
+        talk_to_user_enabled: false,
+      }),
+    );
+    expect(storeUpdateNodeData).toHaveBeenCalledWith("ws-test", {
+      talkToUserEnabled: false,
+    });
+  });
+
+  it("PATCHes /abilities with broadcast_enabled when the broadcast toggle is flipped", async () => {
+    nodeData = {}; // broadcast defaults false
+    render(<ConfigTab workspaceId="ws-test" />);
+    await waitFor(() => expect(apiGet).toHaveBeenCalled());
+    const broadcast = (await screen.findByText("Broadcast to peers"))
+      .closest("label")!
+      .querySelector("input") as HTMLInputElement;
+    fireEvent.click(broadcast); // false → true
+    await waitFor(() =>
+      expect(apiPatch).toHaveBeenCalledWith("/workspaces/ws-test/abilities", {
+        broadcast_enabled: true,
+      }),
+    );
+    expect(storeUpdateNodeData).toHaveBeenCalledWith("ws-test", {
+      broadcastEnabled: true,
+    });
+  });
+});
@@ -405,7 +405,7 @@ export function AgentCommsPanel({ workspaceId }: { workspaceId: string }) {
        </p>
        <button
          onClick={loadInitial}
-          className="text-[10px] px-2 py-0.5 rounded bg-red-800/40 text-bad hover:bg-red-700/50 transition-colors"
+          className="text-[10px] px-2 py-0.5 rounded bg-red-800/40 text-bad hover:bg-red-700/50 transition-colors focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-red-500/60 focus-visible:ring-offset-1"
        >
          Retry
        </button>
@@ -610,7 +610,7 @@ function PeerTabButton({
      aria-selected={active}
      tabIndex={active ? 0 : -1}
      onClick={onClick}
-      className={`shrink-0 px-3 py-1.5 text-[10px] font-medium transition-colors whitespace-nowrap ${
+      className={`shrink-0 px-3 py-1.5 text-[10px] font-medium transition-colors whitespace-nowrap focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-cyan-500/60 focus-visible:ring-offset-1 ${
        active
          ? "border-b-2 border-cyan-500 text-cyan-200"
          : "border-b-2 border-transparent text-ink-mid hover:text-ink-mid"
@@ -33,7 +33,7 @@ export function PendingAttachmentPill({
      <button
        onClick={onRemove}
        aria-label={`Remove ${file.name}`}
-        className="ml-0.5 text-ink-mid hover:text-ink transition-colors shrink-0"
+        className="ml-0.5 text-ink-mid hover:text-ink transition-colors shrink-0 focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent/60 focus-visible:ring-offset-1"
      >
        <svg width="10" height="10" viewBox="0 0 16 16" fill="none" aria-hidden="true">
          <path d="M4 4l8 8M12 4l-8 8" stroke="currentColor" strokeWidth="1.6" strokeLinecap="round" />
@@ -62,8 +62,9 @@ export function AttachmentChip({
  return (
    <button
      onClick={() => onDownload(attachment)}
+      aria-label={`Download ${attachment.name}`}
      title={`Download ${attachment.name}`}
-      className={`flex items-center gap-1.5 rounded-md border px-2 py-1 text-[10px] transition-colors max-w-full ${toneClasses}`}
+      className={`flex items-center gap-1.5 rounded-md border px-2 py-1 text-[10px] transition-colors max-w-full focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent/60 focus-visible:ring-offset-1 ${toneClasses}`}
    >
      <FileGlyph className="shrink-0 opacity-70" />
      <span className="truncate">{attachment.name}</span>
@@ -0,0 +1,85 @@
+"use client";
+
+/**
+ * ChatErrorBanner — error-state banner rendered under the chat
+ * message list when an agent turn fails or the workspace is offline.
+ *
+ * internal#212 closes the "see workspace logs for details" pointer-to-
+ * nowhere defect:
+ *
+ *   - The banner now renders the actionable, secret-safe failure
+ *     reason that ws-server places on `ACTIVITY_LOGGED.error_detail`
+ *     (provider HTTP status + error code + provider's own human
+ *     message). The hook (`useChatSocket`) forwards this through
+ *     `onSendError`, which the ChatTab routes into this banner's
+ *     `message` prop. No hardcoded opaque text in this component.
+ *
+ *   - A "View activity log" button navigates the user to the Activity
+ *     tab where the full row (request body, response body, timing,
+ *     full error_detail) lives. Until internal#212, the banner
+ *     mentioned "workspace logs" with no link — there is no separate
+ *     Logs tab in the side panel; the Activity tab IS the workspace-
+ *     logs surface. Routing through the existing tab makes the
+ *     reference real instead of dangling.
+ *
+ *   - The existing Restart button (shown only when the workspace is
+ *     offline) is preserved unchanged so the recovery affordance the
+ *     old banner offered does not regress.
+ *
+ * Pure presentational — no socket subscription, no state machine. Easy
+ * to unit-test in isolation and easy to compose into the ChatTab.
+ */
+
+import { useCanvasStore } from "@/store/canvas";
+
+export interface ChatErrorBannerProps {
+  /** The user-visible reason. Pass `null` to render nothing. */
+  message: string | null;
+  /** Workspace reachable state — gates the Restart affordance. */
+  isOnline: boolean;
+  /** Fires when the user clicks Restart (offline-only). */
+  onRestart: () => void;
+}
+
+export function ChatErrorBanner({ message, isOnline, onRestart }: ChatErrorBannerProps) {
+  // Pulled from the global store rather than threaded through props so
+  // the chat tab does not need to know about the side-panel tab state.
+  // Matches how Toolbar.tsx triggers the audit tab (the existing
+  // precedent for cross-tab navigation).
+  const setPanelTab = useCanvasStore((s) => s.setPanelTab);
+
+  if (!message) return null;
+
+  return (
+    <div
+      // role="alert" + aria-live mirrors the project's existing WCAG
+      // 4.1.3 banner pattern (see fix/canvas-errors-aria-alert) — a
+      // screen reader announces the failure as soon as it lands.
+      role="alert"
+      aria-live="assertive"
+      className="px-3 py-2 bg-red-900/20 border-t border-red-800/30"
+    >
+      <div className="flex items-center justify-between gap-2">
+        <span className="text-[10px] text-red-300 break-words flex-1">{message}</span>
+        <div className="flex items-center gap-1.5 shrink-0">
+          <button
+            type="button"
+            onClick={() => setPanelTab("activity")}
+            className="text-[10px] px-2 py-0.5 bg-red-900/40 hover:bg-red-800/60 border border-red-700/40 text-red-200 rounded transition-colors focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-accent focus-visible:ring-offset-1"
+          >
+            View activity log
+          </button>
+          {!isOnline && (
+            <button
+              type="button"
+              onClick={onRestart}
+              className="text-[11px] px-2 py-0.5 bg-red-800 text-red-200 rounded hover:bg-red-700"
+            >
+              Restart
+            </button>
+          )}
+        </div>
+      </div>
+    </div>
+  );
+}
@@ -0,0 +1,140 @@
+// @vitest-environment jsdom
+import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
+import { renderHook, act } from "@testing-library/react";
+
+// Capture the handler so we can drive WS events from tests. useSocketEvent
+// stores the latest handler in a ref under the hood, but since we mock
+// the hook entirely, just remember the last passed-in handler.
+let capturedHandler: ((msg: unknown) => void) | null = null;
+vi.mock("@/hooks/useSocketEvent", () => ({
+  useSocketEvent: (h: (msg: unknown) => void) => {
+    capturedHandler = h;
+  },
+}));
+
+// Canvas store mock — useChatSocket calls
+// useCanvasStore.getState().nodes for peer name resolution and reads
+// agentMessages via the selector form. Support both.
+vi.mock("@/store/canvas", () => {
+  const state = {
+    nodes: [
+      { id: "ws-self", data: { name: "Self" } },
+      { id: "ws-peer", data: { name: "Peer Agent" } },
+    ],
+    agentMessages: {} as Record<string, unknown[]>,
+    consumeAgentMessages: () => [],
+  };
+  const hook = (selector?: (s: typeof state) => unknown) =>
+    selector ? selector(state) : state;
+  hook.getState = () => state;
+  return { useCanvasStore: hook };
+});
+
+import { useChatSocket } from "../useChatSocket";
+
+beforeEach(() => {
+  capturedHandler = null;
+});
+
+afterEach(() => {
+  vi.clearAllMocks();
+});
+
+// Helper: assemble an ACTIVITY_LOGGED a2a_receive error event the way
+// the ws-server emits one when a peer call errors out. Fields mirror
+// workspace-server/internal/handlers/activity.go::logActivityExec
+// broadcast payload shape.
+function makeActivityErrorEvent(opts: { workspaceId: string; targetId?: string; errorDetail?: string | undefined }) {
+  return {
+    event: "ACTIVITY_LOGGED",
+    workspace_id: opts.workspaceId,
+    payload: {
+      activity_type: "a2a_receive",
+      method: "message/send",
+      status: "error",
+      target_id: opts.targetId ?? opts.workspaceId,
+      duration_ms: 1500,
+      ...(opts.errorDetail !== undefined ? { error_detail: opts.errorDetail } : {}),
+    },
+    timestamp: "2026-05-18T00:00:00Z",
+  };
+}
+
+describe("useChatSocket — surface error_detail to onSendError (internal#212)", () => {
+  it("forwards the secret-safe error_detail from the broadcast as the onSendError reason", () => {
+    const onSendError = vi.fn();
+    const onSendComplete = vi.fn();
+    renderHook(() =>
+      useChatSocket("ws-self", {
+        onSendError,
+        onSendComplete,
+      }),
+    );
+
+    expect(capturedHandler).not.toBeNull();
+    act(() => {
+      capturedHandler!(
+        makeActivityErrorEvent({
+          workspaceId: "ws-self",
+          errorDetail:
+            "Anthropic 403 oauth_org_not_allowed: Your organization has disabled Claude subscription access for Claude Code",
+        }),
+      );
+    });
+
+    // The hook must NOT fall back to the opaque hardcoded
+    // "Agent error (Exception) — see workspace logs for details." —
+    // that was internal#212. When the broadcast carries an
+    // error_detail, that string is the user-facing reason.
+    expect(onSendError).toHaveBeenCalledTimes(1);
+    const reason = onSendError.mock.calls[0][0] as string;
+    expect(reason).toContain("403");
+    expect(reason).toContain("oauth_org_not_allowed");
+    expect(reason).toContain("disabled Claude subscription");
+    expect(reason).not.toMatch(/see workspace logs for details/i);
+  });
+
+  it("gracefully degrades to the legacy opaque message when error_detail is absent (older ws-server)", () => {
+    // An older ws-server doesn't include error_detail in the payload.
+    // The hook must still fire onSendError with the legacy hardcoded
+    // text so the chat banner has SOMETHING to show. The fix is
+    // additive — never depend on the new field's presence.
+    const onSendError = vi.fn();
+    renderHook(() =>
+      useChatSocket("ws-self", {
+        onSendError,
+      }),
+    );
+
+    act(() => {
+      capturedHandler!(makeActivityErrorEvent({ workspaceId: "ws-self" }));
+    });
+
+    expect(onSendError).toHaveBeenCalledTimes(1);
+    const reason = onSendError.mock.calls[0][0] as string;
+    // Legacy boilerplate is the floor — never silently swallow.
+    expect(reason.length).toBeGreaterThan(0);
+  });
+
+  it("ignores errors targeted at a different workspace's peer", () => {
+    // Defense against a race where the WS hub fans out to all clients —
+    // each chat panel must only react when target_id matches its own
+    // workspace.
+    const onSendError = vi.fn();
+    renderHook(() =>
+      useChatSocket("ws-self", {
+        onSendError,
+      }),
+    );
+    act(() => {
+      capturedHandler!(
+        makeActivityErrorEvent({
+          workspaceId: "ws-self",
+          targetId: "ws-someone-else",
+          errorDetail: "irrelevant",
+        }),
+      );
+    });
+    expect(onSendError).not.toHaveBeenCalled();
+  });
+});
@@ -67,9 +67,23 @@ export function useChatSocket(
            const own = (targetId || msg.workspace_id) === workspaceId;
            if (own) {
              callbacksRef.current.onSendComplete?.();
-              callbacksRef.current.onSendError?.(
-                "Agent error (Exception) — see workspace logs for details.",
-              );
+              // internal#212 — surface the actionable, secret-safe
+              // failure reason (provider HTTP status + error code +
+              // human-readable message) the ws-server now puts on
+              // ACTIVITY_LOGGED.error_detail. The old hardcoded
+              // "Agent error (Exception) — see workspace logs for
+              // details." is the fallback only — it pointed at a
+              // workspace-logs tab that doesn't exist, telling the
+              // user nothing they could act on.
+              //
+              // Graceful degradation: older ws-server builds don't
+              // include error_detail, so the legacy boilerplate is
+              // still the floor (never silently swallow).
+              const detail = (p.error_detail as string) || "";
+              const reason = detail
+                ? detail
+                : "Agent error (Exception) — see workspace logs for details.";
+              callbacksRef.current.onSendError?.(reason);
            }
          }
        } else if (type === "a2a_send") {
@@ -351,8 +351,10 @@ export function SecretsSection({ workspaceId, requiredEnv }: { workspaceId: stri
          {showAdd ? (
            <div className="bg-surface-card/50 rounded p-2 space-y-1.5 border border-line/50">
              <input value={newKey} onChange={(e) => setNewKey(e.target.value.toUpperCase())} placeholder="KEY_NAME"
+                aria-label="Secret key name"
                className="w-full bg-surface-sunken border border-line rounded px-2 py-1 text-[10px] font-mono text-ink focus:outline-none focus:border-accent" />
              <input value={newValue} onChange={(e) => setNewValue(e.target.value)} placeholder="Value" type="password"
+                aria-label="Secret value"
                className="w-full bg-surface-sunken border border-line rounded px-2 py-1 text-[10px] text-ink focus:outline-none focus:border-accent" />
              <div className="flex gap-2">
                <button type="button" onClick={() => { if (newKey && newValue) handleSave(newKey, newValue); }} disabled={!newKey || !newValue}
@@ -2,7 +2,7 @@

 import { useState, useCallback, useRef, useEffect } from 'react';
 import type { TestConnectionState, SecretGroup } from '@/types/secrets';
-import { validateSecret } from '@/lib/api/secrets';
+import { validateSecret, ApiError } from '@/lib/api/secrets';

 interface TestConnectionButtonProps {
  provider: SecretGroup;
@@ -55,9 +55,23 @@ export function TestConnectionButton({
      }
      onResult?.(result.valid);
      resetTimerRef.current = setTimeout(() => setState('idle'), RESET_DELAYS[nextState]!);
-    } catch {
+    } catch (err) {
+      // Distinguish a real failure shape rather than always claiming a
+      // timeout. A reachable server that answered with an HTTP status
+      // (ApiError) did NOT time out — most commonly the validation route
+      // is not available (404/501), which must not masquerade as
+      // "service down". Only an actual thrown network/abort error is a
+      // connectivity failure.
      setState('failure');
-      setErrorDetail('Connection timed out. Service may be down.');
+      if (err instanceof ApiError) {
+        setErrorDetail(
+          err.status === 404 || err.status === 501
+            ? 'Key validation is not available for this service yet. The key was not tested.'
+            : `Could not verify key (server returned ${err.status}). Saving is unaffected.`,
+        );
+      } else {
+        setErrorDetail('Could not reach the validation service. Check your connection and try again.');
+      }
      onResult?.(false);
      resetTimerRef.current = setTimeout(() => setState('idle'), RESET_DELAYS.failure);
    }
@@ -85,7 +99,7 @@ export function TestConnectionButton({

 function Spinner() {
  return (
-    <svg className="spinner" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" strokeWidth="2">
+    <svg aria-hidden="true" className="spinner" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" strokeWidth="2">
      <path d="M12 2v4M12 18v4M4.93 4.93l2.83 2.83M16.24 16.24l2.83 2.83M2 12h4M18 12h4M4.93 19.07l2.83-2.83M16.24 7.76l2.83-2.83" />
    </svg>
  );
@@ -28,8 +28,20 @@ const mockValidateSecret = vi.fn();

 vi.mock("@/lib/api/secrets", () => ({
  validateSecret: (...args: unknown[]) => mockValidateSecret(...args),
+  ApiError: class ApiError extends Error {
+    status: number;
+    constructor(status: number, message: string) {
+      super(message);
+      this.name = "ApiError";
+      this.status = status;
+    }
+  },
 }));

+// Re-import the mocked ApiError so test cases construct the same class the
+// component's `instanceof` check sees.
+import { ApiError } from "@/lib/api/secrets";
+
 beforeEach(() => {
  vi.useFakeTimers();
  vi.clearAllMocks();
@@ -201,8 +213,27 @@ describe("TestConnectionButton — failure path", () => {
 });

 describe("TestConnectionButton — catch path", () => {
-  it("shows 'Connection timed out' on network error", async () => {
-    mockValidateSecret.mockRejectedValue(new Error("timeout"));
+  it("does NOT claim a timeout when the validate endpoint 404s (regression: internal#492)", async () => {
+    // The validate route is unimplemented on the server and returns a fast
+    // 404. Before the fix this rendered the misleading hardcoded string
+    // "Connection timed out. Service may be down." It must instead state
+    // honestly that validation isn't available and the key was not tested.
+    mockValidateSecret.mockRejectedValue(new ApiError(404, "Not Found"));
+    render(
+      <TestConnectionButton provider="anthropic" secretValue="sk-ant-xxx" />,
+    );
+    fireEvent.click(document.querySelector('button[type="button"]')!);
+    await act(async () => {
+      await vi.advanceTimersByTimeAsync(0);
+    });
+    expect(document.body.textContent).not.toContain("Connection timed out");
+    expect(document.body.textContent).not.toContain("Service may be down");
+    expect(document.body.textContent).toContain("not available");
+    expect(document.body.textContent).toContain("not tested");
+  });
+
+  it("reports a non-404 server error with its status, not a timeout", async () => {
+    mockValidateSecret.mockRejectedValue(new ApiError(500, "Internal Server Error"));
    render(
      <TestConnectionButton provider="github" secretValue="ghp_xxx" />,
    );
@@ -210,7 +241,20 @@ describe("TestConnectionButton — catch path", () => {
    await act(async () => {
      await vi.advanceTimersByTimeAsync(0);
    });
-    expect(document.body.textContent).toContain("Connection timed out");
+    expect(document.body.textContent).toContain("500");
+    expect(document.body.textContent).not.toContain("Connection timed out");
+  });
+
+  it("shows a connectivity message on a genuine network error", async () => {
+    mockValidateSecret.mockRejectedValue(new Error("network down"));
+    render(
+      <TestConnectionButton provider="github" secretValue="ghp_xxx" />,
+    );
+    fireEvent.click(document.querySelector('button[type="button"]')!);
+    await act(async () => {
+      await vi.advanceTimersByTimeAsync(0);
+    });
+    expect(document.body.textContent).toContain("Could not reach the validation service");
  });

  it("calls onResult(false) on network error", async () => {
@@ -63,7 +63,7 @@ class MockWebSocket {
 (globalThis as unknown as Record<string, unknown>).WebSocket = MockWebSocket;

 // Now import the socket module (uses globalThis.WebSocket at call time)
-import { connectSocket, disconnectSocket } from "../socket";
+import { connectSocket, disconnectSocket, wakeSocket } from "../socket";
 import { useCanvasStore } from "../canvas";

 // ---------------------------------------------------------------------------
@@ -416,3 +416,84 @@ describe("RehydrateDedup", () => {
    expect(d.shouldSkip(2_700)).toBe(true);
  });
 });
+
+// ---------------------------------------------------------------------------
+// wakeSocket() — visibility-wake reconnect (regression #223 / #228)
+// ---------------------------------------------------------------------------
+//
+// Mobile browsers (iOS Safari, Chrome on Android in deep-sleep) silently
+// drop the WebSocket when the tab is backgrounded; the in-page onclose
+// fires very late or never. Without a visibility wake, the canvas stays
+// frozen until the user manually refreshes.
+//
+// The real wiring lives at module level: connectSocket installs a
+// visibilitychange/pageshow listener that calls wake() on foreground.
+// We can't dispatch DOM events here because the suite runs under the
+// `node` test environment (no `document`/`window` — see canvas/vitest
+// .config.ts). Instead we test wake() directly through the wakeSocket
+// public export, which is the same code path the listener invokes.
+
+describe("wakeSocket → reconnect (#223 / #228 — mobile visibility wake)", () => {
+  it("wake on a healthy OPEN socket does not create a new WebSocket", () => {
+    connectSocket();
+    const ws = getLastWS();
+    ws.triggerOpen();
+    // OPEN === 1. wake() should take the healthy-no-op branch.
+    (ws as unknown as { readyState: number }).readyState = 1;
+    const before = MockWebSocket.instances.length;
+    wakeSocket();
+    expect(MockWebSocket.instances.length).toBe(before);
+  });
+
+  it("wake on a CLOSED socket creates a new WebSocket (the actual #223 fix)", () => {
+    connectSocket();
+    const ws = getLastWS();
+    ws.triggerOpen();
+    // CLOSED === 3. Simulates the OS killing the socket while the tab
+    // was backgrounded. We deliberately don't fire triggerClose() —
+    // the whole point of #223 is that mobile browsers don't fire
+    // onclose when they kill the WS, so reconnect never schedules.
+    (ws as unknown as { readyState: number }).readyState = 3;
+    const before = MockWebSocket.instances.length;
+    wakeSocket();
+    expect(MockWebSocket.instances.length).toBe(before + 1);
+  });
+
+  it("wake while CONNECTING (readyState=0) does not pile another handshake", () => {
+    connectSocket();
+    const ws = getLastWS();
+    // CONNECTING === 0 — a handshake is already in flight.
+    (ws as unknown as { readyState: number }).readyState = 0;
+    const before = MockWebSocket.instances.length;
+    wakeSocket();
+    expect(MockWebSocket.instances.length).toBe(before);
+  });
+
+  it("wake cancels any pending backoff reconnect", () => {
+    const clearTimeoutSpy = vi.spyOn(globalThis, "clearTimeout");
+    connectSocket();
+    const ws = getLastWS();
+    ws.triggerOpen();
+    // Drop the socket — onclose schedules a backoff reconnect.
+    ws.triggerClose();
+    // Now wake the page. wake() should pre-empt the backoff so the
+    // user sees the canvas come back immediately, not after the
+    // exponential delay window.
+    (ws as unknown as { readyState: number }).readyState = 3;
+    clearTimeoutSpy.mockClear();
+    wakeSocket();
+    expect(clearTimeoutSpy).toHaveBeenCalled();
+    clearTimeoutSpy.mockRestore();
+  });
+
+  it("wake after disconnectSocket is a no-op (no zombie reconnect)", () => {
+    connectSocket();
+    const ws = getLastWS();
+    ws.triggerOpen();
+    disconnectSocket();
+    const before = MockWebSocket.instances.length;
+    // Singleton is null now — wake() should silently do nothing.
+    expect(() => wakeSocket()).not.toThrow();
+    expect(MockWebSocket.instances.length).toBe(before);
+  });
+});
@@ -268,6 +268,46 @@ class ReconnectingSocket {
    }
    useCanvasStore.getState().setWsStatus("disconnected");
  }
+
+  /** Force a reconnect attempt now, skipping the backoff window.
+   *  Used by the visibilitychange / pageshow handler: when a mobile
+   *  browser backgrounds the tab, the OS silently kills the WebSocket
+   *  but the in-page onclose either fires very late or never fires at
+   *  all (iOS Safari, Chrome on Android in deep-sleep). Once the user
+   *  brings the tab back, the canvas needs to reconnect within human
+   *  perception — not on whatever backoff delay was last scheduled,
+   *  which can be up to 30s. (#223 / #228)
+   *
+   *  Idempotent: if the socket is already OPEN we leave it alone; the
+   *  WebSocket is still healthy and a reconnect would just churn. */
+  wake() {
+    if (this.disposed) return;
+    // OPEN === 1. Use the numeric literal so we don't have to import
+    // WebSocket type values; the runtime constant is well-defined.
+    if (this.ws && this.ws.readyState === 1) {
+      // Healthy. Run a rehydrate to catch any events we may have missed
+      // while the tab was backgrounded — the OS does deliver some
+      // packets late, but it can also drop them, and the dedup gate
+      // collapses this with any subsequent health-check rehydrate.
+      void this.rehydrate();
+      return;
+    }
+    // CONNECTING === 0 means a handshake is already in flight. Don't
+    // pile another one on; the existing attempt or its onclose-driven
+    // reconnect will resolve.
+    if (this.ws && this.ws.readyState === 0) return;
+    // Otherwise (CLOSING, CLOSED, or null) we're in limbo. Cancel any
+    // pending backoff and reconnect now.
+    if (this.reconnectTimer) {
+      clearTimeout(this.reconnectTimer);
+      this.reconnectTimer = null;
+    }
+    // Reset attempt counter so the *next* failure (if any) starts from
+    // a short delay again — we just had a real user interaction, not
+    // an unattended-tab failure cascade.
+    this.attempt = 0;
+    this.connect();
+  }
 }

 export interface WorkspaceData {
@@ -306,11 +346,49 @@ export interface WorkspaceData {

 let socket: ReconnectingSocket | null = null;

+/** visibilitychange / pageshow handler. Mobile browsers (iOS Safari,
+ *  Chrome on Android in deep-sleep) silently drop the WebSocket when
+ *  the tab is backgrounded — the in-page `onclose` fires very late or
+ *  never. Without this listener, the canvas appears frozen after the
+ *  user backgrounds the PWA and returns to it: status events, agent
+ *  messages, and cross-device chat broadcast don't arrive until a
+ *  manual refresh (#223 / #228).
+ *
+ *  Both events are wired: `visibilitychange` covers tab-switch on a
+ *  live page; `pageshow` covers Safari's bfcache restore, where the
+ *  page comes back from cache without firing visibilitychange. */
+function onPageWake() {
+  // document is undefined in SSR; the listener never installs there,
+  // but defensively guard anyway in case this code is run via a test
+  // harness that doesn't shim it.
+  if (typeof document !== "undefined" && document.hidden) return;
+  socket?.wake();
+}
+let visibilityHandlerInstalled = false;
+function installVisibilityHandler() {
+  if (visibilityHandlerInstalled) return;
+  if (typeof document === "undefined" || typeof window === "undefined") return;
+  document.addEventListener("visibilitychange", onPageWake);
+  // `pageshow` with `event.persisted === true` is the bfcache restore
+  // signal — relevant on iOS Safari. We don't need to inspect
+  // `persisted` because waking an OPEN socket is a no-op.
+  window.addEventListener("pageshow", onPageWake);
+  visibilityHandlerInstalled = true;
+}
+function uninstallVisibilityHandler() {
+  if (!visibilityHandlerInstalled) return;
+  if (typeof document === "undefined" || typeof window === "undefined") return;
+  document.removeEventListener("visibilitychange", onPageWake);
+  window.removeEventListener("pageshow", onPageWake);
+  visibilityHandlerInstalled = false;
+}
+
 export function connectSocket() {
  if (!socket) {
    socket = new ReconnectingSocket(WS_URL);
  }
  socket.connect();
+  installVisibilityHandler();
 }

 export function disconnectSocket() {
@@ -318,4 +396,14 @@ export function disconnectSocket() {
    socket.disconnect();
    socket = null;
  }
+  uninstallVisibilityHandler();
+}
+
+/** Manually trigger the visibility-wake path. Exported so the test suite
+ *  can exercise `ReconnectingSocket.wake()` without depending on a
+ *  jsdom DOM (the rest of this file's tests run under the node env).
+ *  Real-world callers don't need this — the visibility/pageshow listener
+ *  drives it. */
+export function wakeSocket() {
+  socket?.wake();
 }
@@ -584,6 +584,10 @@
 .secrets-tab__refresh-btn:hover {
  background: #1e40af;
 }
+.secrets-tab__refresh-btn:focus-visible {
+  outline: 2px solid #1d4ed8;
+  outline-offset: 2px;
+}

 .secrets-tab__no-results {
  text-align: center;
@@ -649,6 +653,10 @@
  border-radius: 6px;
  cursor: pointer;
 }
+.delete-dialog__cancel-btn:focus-visible {
+  outline: var(--focus-ring);
+  outline-offset: var(--focus-ring-offset);
+}

 .delete-dialog__confirm-btn {
  background: var(--status-invalid);
@@ -658,6 +666,10 @@
  border-radius: 6px;
  cursor: pointer;
 }
+.delete-dialog__confirm-btn:focus-visible {
+  outline: var(--focus-ring);
+  outline-offset: var(--focus-ring-offset);
+}

 .delete-dialog__confirm-btn:disabled { opacity: 0.4; cursor: not-allowed; }

@@ -58,6 +58,7 @@ TOP_LEVEL_MODULES = {
    "a2a_response",
    "a2a_tools",
    "a2a_tools_delegation",
+    "a2a_tools_identity",
    "a2a_tools_inbox",
    "a2a_tools_memory",
    "a2a_tools_messaging",
@@ -310,8 +311,17 @@ locally.
  deps from your system Python. Plain `pip install --user` works
  but the binary lands in `~/.local/bin` (Linux) or
  `~/Library/Python/3.X/bin` (macOS) which is often not on PATH on
-  a fresh shell — `claude mcp add molecule -- molecule-mcp` then
-  fails with "command not found" at first use.
+  a fresh shell — `claude mcp add molecule-<workspace-slug> -- molecule-mcp`
+  then fails with "command not found" at first use.
+
+* **Server name in `claude mcp add` is workspace-specific.** The
+  Canvas "Add to Claude Code" snippet stamps a unique slug
+  (`molecule-<workspace-name>`) so a single Claude Code session can
+  talk to N molecule workspaces concurrently — `claude mcp add` keys
+  entries by name in `~/.claude.json`, so re-running with a bare
+  `molecule` name silently overwrites the prior workspace's entry.
+  See [molecule-core#1535](https://git.moleculesai.app/molecule-ai/molecule-core/pulls/1535)
+  for the canonical generator.

 ### Install

@@ -335,8 +345,10 @@ WORKSPACE_ID=<uuid> \\
 That exposes the same 8 platform tools (`delegate_task`, `list_peers`,
 `send_message_to_user`, `commit_memory`, etc.) that container-bound
 runtimes already get via the workspace's auto-spawned MCP. Register
-the binary in your agent's MCP config (e.g. Claude Code's
-`claude mcp add molecule -- molecule-mcp` with the env above).
+the binary in your agent's MCP config — use a workspace-specific
+server name so multi-workspace setups don't collide (e.g. Claude Code:
+`claude mcp add molecule-<workspace-slug> -- molecule-mcp` with the env
+above; the Canvas modal stamps the right slug for you).

 ### Keeping the token out of shell history

@@ -374,8 +386,8 @@ hold:
   wheel does (see `_build_initialize_result`). Nothing for you to
   do.
 2. **Claude Code installs the server as a marketplace plugin** — a
-   plain `claude mcp add molecule -- molecule-mcp` produces a
-   non-plugin-sourced server, which Claude Code rejects with
+   plain `claude mcp add molecule-<workspace-slug> -- molecule-mcp`
+   produces a non-plugin-sourced server, which Claude Code rejects with
   `channel_enable requires a marketplace plugin`. Until the
   official `moleculesai/claude-code-plugin` marketplace lands
   (tracking [#2936](https://git.moleculesai.app/molecule-ai/molecule-core/issues/2936)),
@@ -0,0 +1,165 @@
+# shellcheck shell=bash
+# Shared peer-visibility assertion core — runtime/backend-AGNOSTIC.
+#
+# WHY THIS FILE EXISTS
+# --------------------
+# The peer-visibility gate (PR #1298) was staging-only. Per the standing
+# rule that the local prod-mimic stack must run a MANDATORY local-Postgres
+# E2E BEFORE staging E2E (memory: feedback_local_must_mimic_production,
+# feedback_mandatory_local_e2e_before_ship, feedback_local_test_before_
+# staging_e2e), peer-visibility must also run against the local stack.
+#
+# The ASSERTION must be byte-identical between local and staging — only
+# provisioning differs. So the literal MCP `list_peers` call + every
+# anti-proxy / anti-native-fallback guarantee lives HERE, sourced by both
+# tests/e2e/test_peer_visibility_mcp_staging.sh (staging/CP backend) and
+# tests/e2e/test_peer_visibility_mcp_local.sh (local docker-compose
+# backend). If this assertion ever diverges between the two, that is the
+# bug — keep it in one place.
+#
+# THIS IS NOT A PROXY. pv_assert_runtime issues the byte-for-byte
+# JSON-RPC `tools/call name=list_peers` envelope to `POST
+# /workspaces/:id/mcp` using the workspace's OWN bearer token, through
+# the real WorkspaceAuth + MCPRateLimiter middleware chain — the exact
+# call mcp_molecule_list_peers makes from a canvas agent. It does NOT
+# read a registry row, /health, the heartbeat table, or
+# GET /registry/:id/peers.
+#
+# Contract:
+#   pv_assert_runtime <runtime> <ws_id> <ws_bearer> <base_url> \
+#                     <org_id_or_empty> <all_ws_ids_space_separated>
+#
+#     <org_id_or_empty>  staging: the X-Molecule-Org-Id header value.
+#                        local:   "" (the local single-tenant stack does
+#                                 not gate on the org header; the header
+#                                 is simply omitted when empty).
+#     <all_ws_ids>       every provisioned workspace id (parent + every
+#                        runtime sibling). The expected peer set for this
+#                        runtime is every id in here EXCEPT <ws_id>.
+#
+#   Sets the global PV_VERDICT to one of:
+#     OK
+#     FAIL(http=<code>)
+#     FAIL(native-fallback)
+#     FAIL(rpc=<detail>)
+#     FAIL(peers=<detail>)
+#     FAIL(unknown)
+#   Returns 0 when PV_VERDICT=OK, 1 otherwise. Never exits — the caller
+#   owns aggregation + the gate exit code (10 = regression reproduced).
+#
+# The literal JSON-RPC envelope. Identical to what
+# workspace/platform_tools/registry.py's mcp_molecule_list_peers emits.
+PV_RPC_BODY='{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"list_peers","arguments":{}}}'
+
+pv_assert_runtime() {
+  local rt="$1" wid="$2" wtok="$3" base_url="$4" org_id="$5" all_ws_ids="$6"
+
+  # Expected peer set = every OTHER provisioned workspace, excluding the
+  # caller itself. Byte-identical selection to the original staging script.
+  local expect_ids
+  expect_ids=$(echo "$all_ws_ids" | tr ' ' '\n' | grep -v "^${wid}$" | grep -v '^$')
+
+  # X-Molecule-Org-Id only when the backend supplies one (staging multi-
+  # tenant). Local single-tenant omits it — the same WorkspaceAuth +
+  # MCPRateLimiter chain still runs; only the tenant-routing header differs.
+  local org_header=()
+  if [ -n "$org_id" ]; then
+    org_header=(-H "X-Molecule-Org-Id: $org_id")
+  fi
+
+  local resp http_code body
+  set +e
+  resp=$(curl -sS -X POST "$base_url/workspaces/$wid/mcp" \
+    -H "Authorization: Bearer $wtok" \
+    "${org_header[@]}" \
+    -H "Content-Type: application/json" \
+    -d "$PV_RPC_BODY" \
+    -o /tmp/pv_mcp_body.json -w "%{http_code}" 2>/dev/null)
+  set -e
+  http_code="$resp"
+  body=$(cat /tmp/pv_mcp_body.json 2>/dev/null || echo '')
+
+  echo "--- $rt (ws=$wid) ---"
+  echo "    HTTP $http_code"
+  echo "    body: $(echo "$body" | head -c 600)"
+
+  # (1) HTTP 200 — a 401 (WorkspaceAuth reject, the Hermes symptom) fails here.
+  if [ "$http_code" != "200" ]; then
+    echo "  ✗ $rt: list_peers MCP call returned HTTP $http_code (expected 200)"
+    PV_VERDICT="FAIL(http=$http_code)"
+    return 1
+  fi
+
+  # (2) JSON-RPC result present, not an error object; expected sibling IDs
+  #     present; not a native-sessions fallback. Byte-identical to the
+  #     original staging script's inline python.
+  local parse
+  parse=$(echo "$body" | python3 -c "
+import sys, json
+expect = set(filter(None, '''$expect_ids'''.split()))
+try:
+    d = json.load(sys.stdin)
+except Exception as e:
+    print('PARSE_ERROR:' + str(e)); sys.exit(0)
+if isinstance(d, dict) and d.get('error') is not None:
+    print('RPC_ERROR:' + json.dumps(d['error'])[:200]); sys.exit(0)
+res = d.get('result') if isinstance(d, dict) else None
+if res is None:
+    print('NO_RESULT'); sys.exit(0)
+# MCP tools/call result shape: {content:[{type:text,text:'<json or prose>'}]}
+text = ''
+if isinstance(res, dict):
+    for c in res.get('content', []):
+        if c.get('type') == 'text':
+            text += c.get('text', '')
+text_l = text.lower()
+# Native-sessions fallback signature (the OpenClaw symptom): the agent
+# answered from its own runtime session list, not the platform peer set.
+if 'sessions_list' in text_l or 'no platform peers' in text_l or 'native session' in text_l:
+    print('NATIVE_FALLBACK:' + text[:200]); sys.exit(0)
+# The expected sibling IDs must literally appear in the returned peer text.
+found = sorted(i for i in expect if i in text)
+missing = sorted(expect - set(found))
+if not expect:
+    print('NO_EXPECTED_PEERS_CONFIGURED'); sys.exit(0)
+if missing:
+    print('MISSING_PEERS:found=%d/%d missing=%s' % (len(found), len(expect), ','.join(m[:8] for m in missing)))
+    sys.exit(0)
+print('OK:found=%d/%d' % (len(found), len(expect)))
+" 2>/dev/null)
+
+  case "$parse" in
+    OK:*)
+      echo "  ✓ $rt: list_peers returned 200 and contains all expected peers ($parse)"
+      PV_VERDICT="OK"
+      return 0
+      ;;
+    NATIVE_FALLBACK:*)
+      echo "  ✗ $rt: list_peers fell back to NATIVE sessions — sees no platform peers ($parse)"
+      PV_VERDICT="FAIL(native-fallback)"
+      return 1
+      ;;
+    RPC_ERROR:*|NO_RESULT|PARSE_ERROR:*)
+      echo "  ✗ $rt: list_peers MCP call did not return a usable result ($parse)"
+      PV_VERDICT="FAIL(rpc=$parse)"
+      return 1
+      ;;
+    MISSING_PEERS:*)
+      echo "  ✗ $rt: list_peers returned 200 but peer set is wrong/empty ($parse)"
+      PV_VERDICT="FAIL(peers=$parse)"
+      return 1
+      ;;
+    NO_EXPECTED_PEERS_CONFIGURED)
+      # Caller bug, not a runtime regression — surface loudly so a
+      # mis-wired backend can't mint a false green.
+      echo "  ✗ $rt: no expected peers were configured for this caller"
+      PV_VERDICT="FAIL(rpc=NO_EXPECTED_PEERS_CONFIGURED)"
+      return 1
+      ;;
+    *)
+      echo "  ✗ $rt: unexpected verdict '$parse'"
+      PV_VERDICT="FAIL(unknown)"
+      return 1
+      ;;
+  esac
+}
@@ -0,0 +1,328 @@
+#!/usr/bin/env bash
+# LOCAL E2E — fresh-provision peer-visibility gate via the LITERAL MCP path.
+#
+# WHY THIS EXISTS
+# ---------------
+# tests/e2e/test_peer_visibility_mcp_staging.sh (PR #1298) codified the
+# literal user-facing peer-visibility path — but staging-only. The
+# standing rule is that the local prod-mimic stack runs a MANDATORY
+# local-Postgres E2E BEFORE staging E2E (memory:
+# feedback_local_must_mimic_production, feedback_mandatory_local_e2e_
+# before_ship, feedback_local_test_before_staging_e2e,
+# feedback_real_subprocess_test_for_boot_path). A staging-only gate means
+# regressions are caught late and expensively on EC2. This is the LOCAL
+# backend: same byte-identical assertion, local docker-compose stack.
+#
+# THE ASSERTION IS NOT A PROXY and is BYTE-IDENTICAL to staging — it is
+# the SAME tests/e2e/lib/peer_visibility_assert.sh::pv_assert_runtime that
+# the staging script calls. It issues the byte-for-byte JSON-RPC
+# `tools/call name=list_peers` envelope to `POST /workspaces/:id/mcp`
+# using each workspace's OWN bearer token, through the real WorkspaceAuth
+# + MCPRateLimiter middleware chain — the exact call
+# mcp_molecule_list_peers makes from a canvas agent. It does NOT read a
+# registry row, /health, the heartbeat table, or GET /registry/:id/peers.
+#
+# Only PROVISIONING differs from staging:
+#   - staging: POST /cp/admin/orgs (cold EC2 tenant) + per-tenant admin
+#     token + each workspace's auth_token from the POST /workspaces resp.
+#   - local:   POST /workspaces directly against the local stack
+#     (BASE, default http://localhost:8080), MCP bearer minted via
+#     GET /admin/workspaces/:id/test-token (e2e_mint_test_token —
+#     deterministic, gated by MOLECULE_ENV != production). Same model
+#     every other local E2E (test_priority_runtimes_e2e.sh,
+#     test_api.sh) already uses; no new credential/provision flow.
+#
+# It is written to FAIL on today's broken Hermes/OpenClaw behavior and go
+# green only when the in-flight root-cause fixes (Hermes-401 #162,
+# OpenClaw-never-online/MCP-wiring #165) actually land — same gate
+# semantics + exit codes as the staging script. NON-required by design
+# until then (flip-to-required tracked at molecule-core#1296), and NOT
+# masked with continue-on-error (feedback_fix_root_not_symptom).
+#
+# Required env: none (local stack only).
+# Optional env:
+#   BASE                    default http://localhost:8080
+#   PV_RUNTIMES             space list; default "hermes openclaw claude-code"
+#   E2E_PROVISION_TIMEOUT_SECS  per-workspace online budget; default 900
+#                            (hermes cold apt+uv is the slow path locally)
+#   E2E_KEEP_WS             1 → skip teardown (local debugging only)
+#   LLM provider keys (a workspace boots only if its provider key is set;
+#   a runtime whose key is absent is SKIPPED, not failed — a partially
+#   keyed local env must not false-fail the gate):
+#     CLAUDE_CODE_OAUTH_TOKEN  claude-code
+#     E2E_MINIMAX_API_KEY      hermes/openclaw (MiniMax, preferred)
+#     E2E_ANTHROPIC_API_KEY    hermes/openclaw (direct Anthropic)
+#     E2E_OPENAI_API_KEY       hermes/openclaw (OpenAI)
+#
+# Exit codes (match the staging script):
+#   0  every runtime under test saw its peers via the literal MCP call
+#   1  generic failure
+#   3  a workspace never reached online within the budget
+#   10 peer-visibility regression reproduced (the gate firing as designed)
+
+set -uo pipefail
+
+source "$(dirname "$0")/_lib.sh"
+# Byte-identical assertion shared with the staging backend.
+# shellcheck source=tests/e2e/lib/peer_visibility_assert.sh
+source "$(dirname "$0")/lib/peer_visibility_assert.sh"
+
+PV_RUNTIMES="${PV_RUNTIMES:-hermes openclaw claude-code}"
+PROVISION_TIMEOUT_SECS="${E2E_PROVISION_TIMEOUT_SECS:-900}"
+NAME_PREFIX="PV-Local-$$-$(date +%H%M%S)"
+
+log()  { echo "[$(date +%H:%M:%S)] $*"; }
+ok()   { echo "[$(date +%H:%M:%S)] ✅ $*"; }
+
+CREATED_WSIDS=()
+
+# ─── Scoped teardown ───────────────────────────────────────────────────
+# Deletes ONLY the workspaces THIS run created (tracked in CREATED_WSIDS),
+# one DELETE /workspaces/:id?confirm=true each. NEVER e2e_cleanup_all_
+# workspaces / any blanket sweep — honors feedback_cleanup_after_each_test
+# and feedback_never_run_cluster_cleanup_tests_on_live_platform (a local
+# stack can still be shared with other concurrent local E2E).
+teardown() {
+  local rc=$?
+  set +e
+  if [ "${E2E_KEEP_WS:-0}" = "1" ]; then
+    echo ""
+    log "[teardown] E2E_KEEP_WS=1 — leaving ${#CREATED_WSIDS[@]} ws for debugging (REMEMBER TO DELETE)"
+    exit $rc
+  fi
+  echo ""
+  log "[teardown] deleting ${#CREATED_WSIDS[@]} workspace(s) this run created (scoped)"
+  for wid in ${CREATED_WSIDS[@]+"${CREATED_WSIDS[@]}"}; do
+    [ -n "$wid" ] || continue
+    curl -s -X DELETE "$BASE/workspaces/$wid?confirm=true" >/dev/null 2>&1 || true
+  done
+  exit $rc
+}
+trap teardown EXIT INT TERM
+
+# Pre-sweep workspaces a prior crashed run of THIS script left behind
+# (name prefix match only — never a blanket delete). The trap fires on
+# normal exit, but a kill -9 / SIGPIPE can bypass it.
+PRIOR=$(curl -s "$BASE/workspaces" | python3 -c '
+import json, sys
+try:
+    print(" ".join(w["id"] for w in json.load(sys.stdin) if w.get("name","").startswith("PV-Local-")))
+except Exception:
+    pass
+' 2>/dev/null)
+for _wid in $PRIOR; do
+  log "Pre-sweeping prior PV-Local workspace: $_wid"
+  curl -s -X DELETE "$BASE/workspaces/$_wid?confirm=true" >/dev/null 2>&1 || true
+done
+
+# ─── Local-stack preflight ─────────────────────────────────────────────
+log "0/5 local stack preflight: $BASE/health"
+if ! curl -fsS "$BASE/health" -m 5 >/dev/null 2>&1; then
+  echo "::error::Local stack not healthy at $BASE/health — bring it up (make up) before this gate. Infra, not a workspace bug (feedback_fix_root_not_symptom)." >&2
+  exit 1
+fi
+# admin/test-token is the local MCP-bearer mint path; it 404s in
+# production. If it is off, this gate cannot drive the literal call.
+if ! curl -fsS "$BASE/admin/workspaces/preflight-probe/test-token" -m 5 >/dev/null 2>&1; then
+  # A 404 here is EITHER "no such ws" (fine — endpoint is enabled) OR the
+  # endpoint is disabled (MOLECULE_ENV=production). Distinguish by body.
+  PROBE=$(curl -s "$BASE/admin/workspaces/preflight-probe/test-token" -m 5 2>/dev/null)
+  if echo "$PROBE" | grep -qi 'production\|disabled\|not found.*endpoint'; then
+    echo "::error::GET /admin/workspaces/:id/test-token disabled (MOLECULE_ENV=production?). Cannot mint a local MCP bearer." >&2
+    exit 1
+  fi
+fi
+ok "    local stack healthy"
+
+# ─── Resolve per-runtime provisioning secrets ──────────────────────────
+# Mirrors test_priority_runtimes_e2e.sh / test_staging_full_saas.sh's
+# provider-key chain. A runtime whose key is absent is SKIPPED (not
+# failed) so a partially keyed local env doesn't false-fail the gate.
+runtime_secrets() {
+  local rt="$1"
+  case "$rt" in
+    claude-code)
+      [ -n "${CLAUDE_CODE_OAUTH_TOKEN:-}" ] || { echo ""; return 1; }
+      python3 -c "import json,os;print(json.dumps({'CLAUDE_CODE_OAUTH_TOKEN':os.environ['CLAUDE_CODE_OAUTH_TOKEN']}))"
+      ;;
+    hermes|openclaw)
+      if [ -n "${E2E_MINIMAX_API_KEY:-}" ]; then
+        python3 -c "import json,os;k=os.environ['E2E_MINIMAX_API_KEY'];print(json.dumps({'ANTHROPIC_BASE_URL':'https://api.minimax.io/anthropic','ANTHROPIC_AUTH_TOKEN':k,'MINIMAX_API_KEY':k}))"
+      elif [ -n "${E2E_ANTHROPIC_API_KEY:-}" ]; then
+        python3 -c "import json,os;k=os.environ['E2E_ANTHROPIC_API_KEY'];print(json.dumps({'ANTHROPIC_API_KEY':k}))"
+      elif [ -n "${E2E_OPENAI_API_KEY:-}" ]; then
+        python3 -c "import json,os;k=os.environ['E2E_OPENAI_API_KEY'];print(json.dumps({'OPENAI_API_KEY':k,'OPENAI_BASE_URL':'https://api.openai.com/v1','MODEL_PROVIDER':'openai:gpt-4o','HERMES_INFERENCE_PROVIDER':'custom','HERMES_CUSTOM_BASE_URL':'https://api.openai.com/v1','HERMES_CUSTOM_API_KEY':k,'HERMES_CUSTOM_API_MODE':'chat_completions'}))"
+      else
+        echo ""; return 1
+      fi
+      ;;
+    *)
+      # Unknown runtime: provision with empty secrets and let the stack
+      # decide (kept permissive so PV_RUNTIMES can be widened later).
+      echo "{}"
+      ;;
+  esac
+}
+
+# Block until $1 reaches one of $2 (space-separated), or $3 sec elapse.
+wait_for_status() {
+  local wsid="$1" want="$2" budget="$3" start=$SECONDS last=""
+  while [ $((SECONDS - start)) -lt "$budget" ]; do
+    local s
+    s=$(curl -s "$BASE/workspaces/$wsid" | python3 -c 'import json,sys
+try:
+  d=json.load(sys.stdin); w=d.get("workspace") if isinstance(d.get("workspace"),dict) else d; print(w.get("status",""))
+except Exception:
+  print("")' 2>/dev/null || echo "")
+    [ "$s" != "$last" ] && { log "    $wsid → ${s:-<none>}"; last="$s"; }
+    for w in $want; do [ "$s" = "$w" ] && { echo "$s"; return 0; }; done
+    sleep 5
+  done
+  echo "$last"
+  return 1
+}
+
+# ─── 1. Provision parent (claude-code) + one sibling per runtime ───────
+# Same topology as the staging script: a claude-code parent plus one
+# sibling per runtime under test, so each runtime should see all others.
+log "1/5 provisioning parent (claude-code) + one sibling per runtime under test..."
+
+PARENT_SECRETS=$(runtime_secrets claude-code) || PARENT_SECRETS=""
+if [ -z "$PARENT_SECRETS" ]; then
+  # Parent still needs to exist as a peer target even without an LLM key;
+  # it never has to answer list_peers itself (it is excluded from the
+  # caller set), so an empty-secrets claude-code shell is sufficient.
+  PARENT_SECRETS="{}"
+fi
+P_RESP=$(curl -s -X POST "$BASE/workspaces" -H "Content-Type: application/json" \
+  -d "{\"name\":\"${NAME_PREFIX}-parent\",\"runtime\":\"claude-code\",\"tier\":3,\"secrets\":$PARENT_SECRETS}")
+PARENT_ID=$(echo "$P_RESP" | python3 -c 'import json,sys;print(json.load(sys.stdin).get("id",""))' 2>/dev/null)
+if [ -z "$PARENT_ID" ]; then
+  echo "::error::parent create failed: $(echo "$P_RESP" | head -c 300)" >&2
+  exit 1
+fi
+CREATED_WSIDS+=("$PARENT_ID")
+log "    PARENT_ID=$PARENT_ID"
+
+# NOTE: no `declare -A` — this script must also run on a local macOS dev
+# box (bash 3.2, no associative arrays) per feedback_local_must_mimic_
+# production. WS_IDS / VERDICT are kept as newline-delimited "rt<TAB>val"
+# maps with tiny get/set helpers (portable to bash 3.2+ AND ubuntu CI).
+WS_IDS_MAP=""
+VERDICT_MAP=""
+_map_set() { # _map_set <mapvarname> <key> <value>
+  local __m="$1" __k="$2" __v="$3" __cur
+  eval "__cur=\$$__m"
+  __cur=$(printf '%s' "$__cur" | grep -v "^${__k}	" || true)
+  if [ -n "$__cur" ]; then
+    eval "$__m=\$(printf '%s\n%s\t%s' \"\$__cur\" \"\$__k\" \"\$__v\")"
+  else
+    eval "$__m=\$(printf '%s\t%s' \"\$__k\" \"\$__v\")"
+  fi
+}
+_map_get() { # _map_get <mapvarname> <key>  -> stdout value (empty if absent)
+  local __m="$1" __k="$2" __cur
+  eval "__cur=\$$__m"
+  printf '%s\n' "$__cur" | awk -F'\t' -v k="$__k" '$1==k {print $2; exit}'
+}
+
+ALL_WS_IDS="$PARENT_ID"
+ACTIVE_RUNTIMES=""
+for rt in $PV_RUNTIMES; do
+  SEC=$(runtime_secrets "$rt") || SEC=""
+  if [ -z "$SEC" ]; then
+    log "    SKIP $rt — no provider key in env (partially-keyed local env; not a failure)"
+    continue
+  fi
+  R=$(curl -s -X POST "$BASE/workspaces" -H "Content-Type: application/json" \
+    -d "{\"name\":\"${NAME_PREFIX}-$rt\",\"runtime\":\"$rt\",\"tier\":2,\"parent_id\":\"$PARENT_ID\",\"secrets\":$SEC}")
+  WID=$(echo "$R" | python3 -c 'import json,sys;print(json.load(sys.stdin).get("id",""))' 2>/dev/null)
+  if [ -z "$WID" ]; then
+    echo "::error::$rt workspace create failed: $(echo "$R" | head -c 300)" >&2
+    exit 1
+  fi
+  _map_set WS_IDS_MAP "$rt" "$WID"
+  CREATED_WSIDS+=("$WID")
+  ALL_WS_IDS="$ALL_WS_IDS $WID"
+  ACTIVE_RUNTIMES="$ACTIVE_RUNTIMES $rt"
+  log "    $rt → $WID"
+done
+ACTIVE_RUNTIMES="$(echo "$ACTIVE_RUNTIMES" | xargs)"
+
+if [ -z "$ACTIVE_RUNTIMES" ]; then
+  echo "::error::No runtime had a provider key set — cannot run the local peer-visibility gate. Set CLAUDE_CODE_OAUTH_TOKEN and/or E2E_MINIMAX_API_KEY (or ANTHROPIC/OPENAI)." >&2
+  exit 1
+fi
+
+# ─── 2. Wait for the parent online (it is a peer target) ───────────────
+log "2/5 waiting for parent online (peer target)..."
+PF=$(wait_for_status "$PARENT_ID" "online" "$PROVISION_TIMEOUT_SECS") || true
+if [ "$PF" != "online" ]; then
+  echo "::error::parent ($PARENT_ID) never reached online (last=$PF) within ${PROVISION_TIMEOUT_SECS}s" >&2
+  exit 3
+fi
+ok "    parent online"
+
+# ─── 3. Wait for every sibling online ──────────────────────────────────
+# A runtime that never comes online locally is itself a finding: it
+# reproduces the openclaw-never-online class (#165) on the local stack.
+log "3/5 waiting for all siblings online (up to ${PROVISION_TIMEOUT_SECS}s each — cold boot)..."
+REGRESSED=0
+ONLINE_RUNTIMES=""
+for rt in $ACTIVE_RUNTIMES; do
+  wid="$(_map_get WS_IDS_MAP "$rt")"
+  S=$(wait_for_status "$wid" "online" "$PROVISION_TIMEOUT_SECS") || true
+  if [ "$S" != "online" ]; then
+    echo "  ✗ $rt ($wid): never reached online (last=$S) — reproduces the never-online class locally"
+    _map_set VERDICT_MAP "$rt" "FAIL(never-online:last=$S)"
+    REGRESSED=1
+    continue
+  fi
+  ok "    $rt online"
+  ONLINE_RUNTIMES="$ONLINE_RUNTIMES $rt"
+done
+
+# ─── 4. THE GATE — literal mcp_molecule_list_peers via POST /:id/mcp ────
+# Shared, byte-identical assertion. Local passes "" for the org id (the
+# single-tenant local stack does not gate on X-Molecule-Org-Id); the
+# literal MCP call + every anti-proxy / anti-native-fallback guarantee is
+# the SAME code the staging backend runs.
+log "4/5 driving the LITERAL list_peers MCP call per online runtime..."
+echo ""
+for rt in $ONLINE_RUNTIMES; do
+  wid="$(_map_get WS_IDS_MAP "$rt")"
+  WTOK=$(e2e_mint_test_token "$wid" 2>/dev/null || true)
+  if [ -z "$WTOK" ]; then
+    echo "--- $rt (ws=$wid) ---"
+    echo "  ✗ $rt: could not mint a local MCP bearer (admin/test-token) — cannot drive the literal call"
+    _map_set VERDICT_MAP "$rt" "FAIL(no-bearer)"
+    REGRESSED=1
+    echo ""
+    continue
+  fi
+  PV_VERDICT=""
+  pv_assert_runtime "$rt" "$wid" "$WTOK" "$BASE" "" "$ALL_WS_IDS" || REGRESSED=1
+  _map_set VERDICT_MAP "$rt" "$PV_VERDICT"
+  echo ""
+done
+
+# ─── 5. Summary + honest gate exit ─────────────────────────────────────
+echo "=== SUMMARY — LOCAL fresh-provision peer-visibility (literal MCP list_peers) ==="
+for rt in $ACTIVE_RUNTIMES; do
+  _v="$(_map_get VERDICT_MAP "$rt")"
+  printf '  %-14s %s\n' "$rt" "${_v:-NO_RUN}"
+done
+echo ""
+
+if [ "$REGRESSED" -ne 0 ]; then
+  echo "✗ GATE FAILED (LOCAL) — at least one runtime cannot see its peers via"
+  echo "  the literal mcp_molecule_list_peers call on the local prod-mimic"
+  echo "  stack. This is the SAME user-facing failure the proxy signals were"
+  echo "  hiding, reproduced locally (far faster than EC2). Expected RED until"
+  echo "  the Hermes-401 (#162) + OpenClaw-never-online/MCP-wiring (#165)"
+  echo "  root-cause fixes land; goes green only when they actually do."
+  exit 10
+fi
+
+ok "GATE PASSED (LOCAL) — every runtime under test sees its platform peers via the literal MCP call."
+exit 0
@@ -64,6 +64,13 @@

 set -uo pipefail

+# The literal MCP list_peers assertion lives in the shared, backend-
+# agnostic lib so it is BYTE-IDENTICAL between this staging backend and
+# the local docker-compose backend (tests/e2e/test_peer_visibility_mcp_
+# local.sh). Only provisioning/teardown differs per backend.
+# shellcheck source=tests/e2e/lib/peer_visibility_assert.sh
+source "$(dirname "${BASH_SOURCE[0]}")/lib/peer_visibility_assert.sh"
+
 CP_URL="${MOLECULE_CP_URL:-https://staging-api.moleculesai.app}"
 ADMIN_TOKEN="${MOLECULE_ADMIN_TOKEN:?MOLECULE_ADMIN_TOKEN required — Railway staging CP_ADMIN_API_TOKEN}"
 RUN_ID_SUFFIX="${E2E_RUN_ID:-$(date +%H%M%S)-$$}"
@@ -259,101 +266,19 @@ done
 # through WorkspaceAuth + MCPRateLimiter.
 log "6/6 driving the LITERAL list_peers MCP call per runtime..."
 echo ""
-RPC_BODY='{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"list_peers","arguments":{}}}'
 REGRESSED=0
 declare -A VERDICT

 for rt in $PV_RUNTIMES; do
  wid="${WS_IDS[$rt]}"
  wtok="${WS_TOKENS[$rt]}"
-  # The expected peer set = every OTHER provisioned workspace (parent +
-  # the sibling runtimes), excluding the caller itself.
-  EXPECT_IDS=$(echo "$ALL_WS_IDS" | tr ' ' '\n' | grep -v "^${wid}$" | grep -v '^$')
-
-  set +e
-  RESP=$(curl -sS -X POST "$TENANT_URL/workspaces/$wid/mcp" \
-    -H "Authorization: Bearer $wtok" \
-    -H "X-Molecule-Org-Id: $ORG_ID" \
-    -H "Content-Type: application/json" \
-    -d "$RPC_BODY" \
-    -o /tmp/pv_mcp_body.json -w "%{http_code}" 2>/dev/null)
-  set -e
-  HTTP_CODE="$RESP"
-  BODY=$(cat /tmp/pv_mcp_body.json 2>/dev/null || echo '')
-
-  echo "--- $rt (ws=$wid) ---"
-  echo "    HTTP $HTTP_CODE"
-  echo "    body: $(echo "$BODY" | head -c 600)"
-
-  # (1) HTTP 200 — a 401 (WorkspaceAuth reject, the Hermes symptom) fails here.
-  if [ "$HTTP_CODE" != "200" ]; then
-    echo "  ✗ $rt: list_peers MCP call returned HTTP $HTTP_CODE (expected 200)"
-    VERDICT[$rt]="FAIL(http=$HTTP_CODE)"
-    REGRESSED=1
-    continue
-  fi
-
-  # (2) JSON-RPC result present, not an error object.
-  PARSE=$(echo "$BODY" | python3 -c "
-import sys, json
-expect = set(filter(None, '''$EXPECT_IDS'''.split()))
-try:
-    d = json.load(sys.stdin)
-except Exception as e:
-    print('PARSE_ERROR:' + str(e)); sys.exit(0)
-if isinstance(d, dict) and d.get('error') is not None:
-    print('RPC_ERROR:' + json.dumps(d['error'])[:200]); sys.exit(0)
-res = d.get('result') if isinstance(d, dict) else None
-if res is None:
-    print('NO_RESULT'); sys.exit(0)
-# MCP tools/call result shape: {content:[{type:text,text:'<json or prose>'}]}
-text = ''
-if isinstance(res, dict):
-    for c in res.get('content', []):
-        if c.get('type') == 'text':
-            text += c.get('text', '')
-text_l = text.lower()
-# Native-sessions fallback signature (the OpenClaw symptom): the agent
-# answered from its own runtime session list, not the platform peer set.
-if 'sessions_list' in text_l or 'no platform peers' in text_l or 'native session' in text_l:
-    print('NATIVE_FALLBACK:' + text[:200]); sys.exit(0)
-# The expected sibling IDs must literally appear in the returned peer text.
-found = sorted(i for i in expect if i in text)
-missing = sorted(expect - set(found))
-if not expect:
-    print('NO_EXPECTED_PEERS_CONFIGURED'); sys.exit(0)
-if missing:
-    print('MISSING_PEERS:found=%d/%d missing=%s' % (len(found), len(expect), ','.join(m[:8] for m in missing)))
-    sys.exit(0)
-print('OK:found=%d/%d' % (len(found), len(expect)))
-" 2>/dev/null)
-
-  case "$PARSE" in
-    OK:*)
-      echo "  ✓ $rt: list_peers returned 200 and contains all expected peers ($PARSE)"
-      VERDICT[$rt]="OK"
-      ;;
-    NATIVE_FALLBACK:*)
-      echo "  ✗ $rt: list_peers fell back to NATIVE sessions — sees no platform peers ($PARSE)"
-      VERDICT[$rt]="FAIL(native-fallback)"
-      REGRESSED=1
-      ;;
-    RPC_ERROR:*|NO_RESULT|PARSE_ERROR:*)
-      echo "  ✗ $rt: list_peers MCP call did not return a usable result ($PARSE)"
-      VERDICT[$rt]="FAIL(rpc=$PARSE)"
-      REGRESSED=1
-      ;;
-    MISSING_PEERS:*)
-      echo "  ✗ $rt: list_peers returned 200 but peer set is wrong/empty ($PARSE)"
-      VERDICT[$rt]="FAIL(peers=$PARSE)"
-      REGRESSED=1
-      ;;
-    *)
-      echo "  ✗ $rt: unexpected verdict '$PARSE'"
-      VERDICT[$rt]="FAIL(unknown)"
-      REGRESSED=1
-      ;;
-  esac
+  # Byte-identical assertion via the shared lib. Staging passes ORG_ID as
+  # the X-Molecule-Org-Id header value; the literal MCP call + every
+  # anti-proxy / anti-native-fallback guarantee is the SAME code the
+  # local backend runs.
+  PV_VERDICT=""
+  pv_assert_runtime "$rt" "$wid" "$wtok" "$TENANT_URL" "$ORG_ID" "$ALL_WS_IDS" || REGRESSED=1
+  VERDICT[$rt]="$PV_VERDICT"
  echo ""
 done

@@ -0,0 +1,35 @@
+// Command t4-contract-dump prints the T4 privilege contract as YAML.
+//
+// Usage:
+//
+//	go run ./workspace-server/cmd/t4-contract-dump > t4_capabilities.yaml
+//
+// This is the seam that template-repo CI workflows consume:
+//
+//	- Template CI fetches molecule-core at pinned ref
+//	- Runs `go run ./workspace-server/cmd/t4-contract-dump` to produce
+//	  t4_capabilities.yaml
+//	- Iterates capabilities and runs each Probe inside a freshly-built
+//	  privileged container
+//	- Aggregates structured pass/fail; fails the gate on any hard miss.
+//
+// Keeping this trivial and pure-stdlib means a fork user does not need
+// a Molecule-AI Gitea token or any internal infrastructure to consume
+// the contract — `go run` against molecule-core's public source is
+// enough.
+package main
+
+import (
+	"fmt"
+	"os"
+
+	"github.com/Molecule-AI/molecule-monorepo/platform/internal/provisioner"
+)
+
+func main() {
+	caps := provisioner.T4PrivilegeContract()
+	if _, err := os.Stdout.WriteString(provisioner.AsYAML(caps)); err != nil {
+		fmt.Fprintln(os.Stderr, "t4-contract-dump: write failed:", err)
+		os.Exit(1)
+	}
+}
@@ -399,7 +399,21 @@ func (h *WorkspaceHandler) proxyA2ARequest(ctx context.Context, workspaceID stri
 	// (no Do(), no maybeMarkContainerDead). The response is a synthetic
 	// {status:"queued"} envelope so the caller (canvas, another workspace)
 	// knows delivery is acknowledged but pending consumption.
-	if lookupDeliveryMode(ctx, workspaceID) == models.DeliveryModePoll {
+	deliveryMode, deliveryModeErr := lookupDeliveryMode(ctx, workspaceID)
+	if deliveryModeErr != nil {
+		// internal#497 fail-closed: a real DB/context error on the
+		// delivery-mode read MUST NOT silently fall through to the push
+		// dispatch path — that is exactly what silently misrouted every
+		// poll-mode peer for 5 days under the ce2db75f regression. Surface
+		// a structured error so the delegation is marked failed (loud +
+		// retryable) instead of dispatched to the wrong path.
+		log.Printf("ProxyA2A: delivery-mode lookup failed for %s: %v — failing closed", workspaceID, deliveryModeErr)
+		return 0, nil, &proxyA2AError{
+			Status:   http.StatusServiceUnavailable,
+			Response: gin.H{"error": "delivery-mode lookup failed; refusing to dispatch to avoid silent misrouting"},
+		}
+	}
+	if deliveryMode == models.DeliveryModePoll {
 		if logActivity {
 			h.logA2AReceiveQueued(ctx, workspaceID, callerID, body, a2aMethod)
 		}
@@ -168,6 +168,21 @@ func (h *WorkspaceHandler) maybeMarkContainerDead(ctx context.Context, workspace
 	if !h.HasProvisioner() {
 		return false
 	}
+	// Restart-aware short-circuit: during the 20-30s EC2-pending window of
+	// an in-flight restart, the workspace's url='' and IsRunning() returns
+	// false → looks indistinguishable from a dead container. Pre-fix this
+	// fired a fresh RestartByID for the just-launched instance, which
+	// coalesceRestart's pending-flag drained by running ANOTHER full
+	// stop+provision cycle (= ec2_stopped of the still-pending instance
+	// → re-provision). That's the 4x reprov thrash class. Skip the
+	// container-dead path while a restart is in flight; the in-flight
+	// restart's own provisionWorkspaceAutoSync will surface a real failure
+	// (markProvisionFailed) if the new container never comes up. Issue
+	// internal#544.
+	if isRestarting(workspaceID) {
+		log.Printf("ProxyA2A: maybeMarkContainerDead skipped for %s — restart already in flight (self-fire guard)", workspaceID)
+		return false
+	}

 	var running bool
 	var inspectErr error
@@ -194,6 +209,11 @@ func (h *WorkspaceHandler) maybeMarkContainerDead(ctx context.Context, workspace
 	}
 	db.ClearWorkspaceKeys(ctx, workspaceID)
 	h.broadcaster.RecordAndBroadcast(ctx, string(events.EventWorkspaceOffline), workspaceID, map[string]interface{}{})
+	// Tracked via goAsync (not bare `go`) so the asyncWG can be drained
+	// before a test swaps the global db.DB. runRestartCycle reads db.DB
+	// before its provisioner gate, so an untracked detached goroutine
+	// races setupTestDB's t.Cleanup db.DB restore. Matches the already-
+	// correct site at a2a_proxy.go:648.
 	h.goAsync(func() { h.RestartByID(workspaceID) })
 	return true
 }
@@ -218,6 +238,18 @@ func (h *WorkspaceHandler) maybeMarkContainerDead(ctx context.Context, workspace
 // shape post-EC2-replace (see molecule-controlplane#20 incident
 // 2026-05-07) where the reconciler hasn't respawned the agent yet.
 func (h *WorkspaceHandler) preflightContainerHealth(ctx context.Context, workspaceID string) *proxyA2AError {
+	// Restart-aware short-circuit (mirror of maybeMarkContainerDead): if a
+	// restart cycle is in flight for this workspace, do not run the
+	// IsRunning probe — it would observe the EC2-pending state as "not
+	// running" and trigger RestartByID for an already-restarting workspace,
+	// closing the self-fire loop. Returning nil lets the optimistic
+	// forward proceed; the upstream Do() call will fail with a connection
+	// error or 502, and the *post-restart* reactive path can decide what
+	// to do once the cycle has actually completed. Issue internal#544.
+	if isRestarting(workspaceID) {
+		log.Printf("ProxyA2A preflight: %s — skipped, restart already in flight (self-fire guard)", workspaceID)
+		return nil
+	}
 	running, err := h.provisioner.IsRunning(ctx, workspaceID)
 	if err != nil {
 		// Transient daemon error. Provisioner.IsRunning returns (true, err)
@@ -241,6 +273,9 @@ func (h *WorkspaceHandler) preflightContainerHealth(ctx context.Context, workspa
 	}
 	db.ClearWorkspaceKeys(ctx, workspaceID)
 	h.broadcaster.RecordAndBroadcast(ctx, string(events.EventWorkspaceOffline), workspaceID, map[string]interface{}{})
+	// Tracked via goAsync (see maybeMarkContainerDead): preflight's
+	// detached restart must be drainable so it doesn't race the global
+	// db.DB swap in test cleanup.
 	h.goAsync(func() { h.RestartByID(workspaceID) })
 	return &proxyA2AError{
 		Status: http.StatusServiceUnavailable,
@@ -262,8 +297,9 @@ func (h *WorkspaceHandler) logA2AFailure(ctx context.Context, workspaceID, calle
 		errWsName = workspaceID
 	}
 	summary := "A2A request to " + errWsName + " failed: " + errMsg
+	parent := ctx
 	h.goAsync(func() {
-		logCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx), 30*time.Second)
+		logCtx, cancel := context.WithTimeout(context.WithoutCancel(parent), 30*time.Second)
 		defer cancel()
 		LogActivity(logCtx, h.broadcaster, ActivityParams{
 			WorkspaceID:  workspaceID,
@@ -309,8 +345,9 @@ func (h *WorkspaceHandler) logA2ASuccess(ctx context.Context, workspaceID, calle
 	}
 	summary := a2aMethod + " → " + wsNameForLog
 	toolTrace := extractToolTrace(respBody)
+	parent := ctx
 	h.goAsync(func() {
-		logCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx), 30*time.Second)
+		logCtx, cancel := context.WithTimeout(context.WithoutCancel(parent), 30*time.Second)
 		defer cancel()
 		LogActivity(logCtx, h.broadcaster, ActivityParams{
 			WorkspaceID:  workspaceID,
@@ -458,40 +495,64 @@ func parseUsageFromA2AResponse(body []byte) (inputTokens, outputTokens int64) {
 	return 0, 0
 }

-// lookupDeliveryMode returns the workspace's delivery_mode. On any DB
-// error or missing row it returns DeliveryModePush — the fail-closed
-// default. "Closed" here means "fall back to today's behavior (synchronous
-// dispatch)" rather than "fall back to drop the request silently into
-// activity_logs where the agent might never see it." A poll-mode workspace
-// that briefly reads as push will get its A2A request dispatched to the
-// stored URL (or a 502 if no URL); a push-mode workspace that briefly
-// reads as poll would get its request silently queued with no dispatch.
-// The first failure is loud + recoverable; the second is silent.
+// lookupDeliveryMode returns the workspace's delivery_mode.
+//
+// internal#497 / RFC#497 fail-closed (SURGICAL scope): the *specific*
+// failure mode that hid the ce2db75f regression for 5 days is now
+// propagated instead of silently swallowed — a CONTEXT error
+// (context.Canceled / context.DeadlineExceeded). Under ce2db75f the
+// detached delegation goroutine ran on a cancelled request context, every
+// `SELECT delivery_mode` failed `context canceled`, this function returned
+// push, the poll-mode short-circuit in proxyA2ARequest was skipped, and
+// poll-mode peers (e.g. an operator laptop on molecule-mcp-claude-channel)
+// silently never got their a2a_receive inbox row. A transient,
+// systematic-once-triggered context cancellation became permanent
+// invisible misrouting. Returning that error lets the caller fail loud
+// (mark the delegation failed) instead of mis-dispatching.
+//
+// Scope is deliberately narrow: only ctx errors propagate. Other DB
+// errors retain the long-standing documented "fall back to push (today's
+// synchronous behavior)" contract — that path is loud + recoverable
+// (502 / SSRF reject / restart), unlike the silent poll-mode drop, and
+// the surrounding proxy (incl. the sibling checkWorkspaceBudget) is
+// intentionally built around that fail-open-to-push behavior. Widening
+// further is an RFC#497 follow-up, not part of this P0 fix.
+//
+// A genuinely *absent* configuration is NOT an error and still resolves to
+// push (the safe synchronous default): sql.ErrNoRows, a NULL/empty column,
+// or an unrecognised value all return (push, nil).
 //
 // The function is intentionally lookup-only — it never mutates the row.
 // The register handler (registry.go) is the only writer for delivery_mode.
 //
 // See #2339 PR 1 for the column + register-flow side; this is the
 // proxy-side read used for the short-circuit in proxyA2ARequest.
-func lookupDeliveryMode(ctx context.Context, workspaceID string) string {
+func lookupDeliveryMode(ctx context.Context, workspaceID string) (string, error) {
 	var mode sql.NullString
 	err := db.DB.QueryRowContext(ctx,
 		`SELECT delivery_mode FROM workspaces WHERE id = $1`, workspaceID,
 	).Scan(&mode)
 	if err != nil {
-		if !errors.Is(err, sql.ErrNoRows) {
-			log.Printf("ProxyA2A: lookupDeliveryMode(%s) failed (%v) — defaulting to push", workspaceID, err)
+		// internal#497: a context cancellation/deadline MUST NOT be
+		// swallowed into a silent push default — that is the exact 5-day
+		// silent-misrouting vector. Propagate so the caller fails closed.
+		if errors.Is(err, context.Canceled) || errors.Is(err, context.DeadlineExceeded) {
+			log.Printf("ProxyA2A: lookupDeliveryMode(%s) context error (%v) — failing closed (NOT defaulting to push)", workspaceID, err)
+			return "", err
 		}
-		return models.DeliveryModePush
+		if !errors.Is(err, sql.ErrNoRows) {
+			log.Printf("ProxyA2A: lookupDeliveryMode(%s) failed (%v) — defaulting to push (non-ctx DB error; legacy fail-open-to-push contract)", workspaceID, err)
+		}
+		return models.DeliveryModePush, nil
 	}
 	if !mode.Valid || mode.String == "" {
-		return models.DeliveryModePush
+		return models.DeliveryModePush, nil
 	}
 	if !models.IsValidDeliveryMode(mode.String) {
 		log.Printf("ProxyA2A: workspace %s has invalid delivery_mode=%q — defaulting to push", workspaceID, mode.String)
-		return models.DeliveryModePush
+		return models.DeliveryModePush, nil
 	}
-	return mode.String
+	return mode.String, nil
 }

 // logA2AReceiveQueued records a poll-mode "queued" A2A receive into
@@ -2235,12 +2235,18 @@ func TestProxyA2A_PushMode_NoShortCircuit(t *testing.T) {
 	}
 }

-// TestProxyA2A_PollMode_FailsClosedToPush verifies the safety contract:
-// a DB error reading delivery_mode must default to push (the existing
-// behavior), NOT poll. Failing to push means a poll-mode workspace
-// briefly attempts a real dispatch — visible failure (502 / SSRF
-// rejection / restart cascade), not a silent drop into activity_logs
-// where the agent might never look. Loud > silent, recoverable > lost.
+// TestProxyA2A_PollMode_FailsClosedToPush verifies the LEGACY safety
+// contract is PRESERVED for non-context DB errors: a generic DB error
+// reading delivery_mode still defaults to push (today's behavior), NOT
+// poll. Failing to push means a poll-mode workspace briefly attempts a
+// real dispatch — visible failure (502 / SSRF rejection / restart
+// cascade), not a silent drop into activity_logs where the agent might
+// never look. Loud > silent, recoverable > lost.
+//
+// internal#497 narrows the fail-closed change to *context* errors only
+// (the actual ce2db75f regression vector); generic DB errors keep this
+// long-standing fail-open-to-push contract. The ctx-error fail-closed is
+// covered by TestLookupDeliveryMode_ContextCanceled_FailsClosed.
 func TestProxyA2A_PollMode_FailsClosedToPush(t *testing.T) {
 	mock := setupTestDB(t)
 	setupTestRedis(t) // empty Redis — forces resolveAgentURL DB lookup
@@ -2251,7 +2257,8 @@ func TestProxyA2A_PollMode_FailsClosedToPush(t *testing.T) {

 	expectBudgetCheck(mock, wsID)

-	// lookupDeliveryMode hits a transient DB error → must default push.
+	// lookupDeliveryMode hits a generic (non-context) DB error → must
+	// still default push (legacy contract preserved by internal#497).
 	mock.ExpectQuery("SELECT delivery_mode FROM workspaces WHERE id").
 		WithArgs(wsID).
 		WillReturnError(sql.ErrConnDone)
@@ -2275,7 +2282,7 @@ func TestProxyA2A_PollMode_FailsClosedToPush(t *testing.T) {
 		var resp map[string]interface{}
 		_ = json.Unmarshal(w.Body.Bytes(), &resp)
 		if resp["status"] == "queued" {
-			t.Errorf("DB error on delivery_mode lookup silently queued the request — must fail-closed-to-push, got body: %s", w.Body.String())
+			t.Errorf("generic DB error on delivery_mode lookup silently queued the request — must fail-open-to-push, got body: %s", w.Body.String())
 		}
 	}

@@ -2284,6 +2291,37 @@ func TestProxyA2A_PollMode_FailsClosedToPush(t *testing.T) {
 	}
 }

+// TestLookupDeliveryMode_ContextCanceled_FailsClosed is the internal#497
+// regression test for the SECONDARY defect. It pins the exact invariant
+// that hid the ce2db75f regression for 5 days: when the delivery_mode read
+// fails because the context was cancelled (precisely what happened in the
+// detached delegation goroutine running on a returned request context),
+// lookupDeliveryMode MUST return an error and MUST NOT silently return
+// "push". Returning push there is what skipped the poll-mode short-circuit
+// and silently dropped 100% of poll-mode peer deliveries.
+//
+// A pre-cancelled context makes QueryRowContext fail with
+// context.Canceled deterministically — no DB rows are mocked because the
+// query never reaches a result.
+func TestLookupDeliveryMode_ContextCanceled_FailsClosed(t *testing.T) {
+	mock := setupTestDB(t)
+	// The query fails on the cancelled ctx before matching; provide a
+	// permissive expectation so sqlmock doesn't complain about the attempt.
+	mock.ExpectQuery("SELECT delivery_mode FROM workspaces WHERE id").
+		WillReturnError(context.Canceled)
+
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel() // simulate the HTTP handler having returned (request ctx dead)
+
+	mode, err := lookupDeliveryMode(ctx, "ws-poll-peer")
+	if err == nil {
+		t.Fatalf("internal#497 regression: lookupDeliveryMode swallowed a context error and returned mode=%q with nil err — this is the exact 5-day silent-misrouting vector", mode)
+	}
+	if mode == models.DeliveryModePush {
+		t.Errorf("internal#497 regression: context error must NOT default to push (got mode=%q)", mode)
+	}
+}
+
 // ==================== a2aClient ResponseHeaderTimeout config ====================

 func TestA2AClientResponseHeaderTimeout(t *testing.T) {
@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"log"
 	"net/http"
+	"regexp"
 	"strconv"
 	"strings"
 	"time"
@@ -18,6 +19,46 @@ import (
 	"github.com/google/uuid"
 )

+// internal#212 — secret-safe scrubber applied to error_detail strings
+// before they cross the canvas WebSocket. Defense in depth: the
+// workspace runtime already runs `_sanitize_for_external` on its side
+// (workspace/executor_helpers.py), but the broadcast layer is the last
+// stop before the string reaches the user's browser, so we re-scrub
+// here in case any caller path forgot.
+//
+// The scrubber is intentionally surgical — it MUST preserve the
+// actionable parts (HTTP status codes, error codes like
+// `oauth_org_not_allowed`, human-readable provider messages) and
+// remove only what looks credential-ish. Over-redacting defeats the
+// whole point of internal#212 (giving the user a reason they can act on).
+
+// Capture (auth-key prefix) (value) so the prefix can be preserved in
+// the output. The keyword anchor prevents false positives on regular
+// text that happens to contain a long alphanumeric run.
+var errorDetailSecretRE = regexp.MustCompile(`(?i)((?:bearer|token|api[_-]?key|sk-proj-|sk-)[ :=]*)[A-Za-z0-9_/.-]{20,}`)
+
+// Stringly-typed JWT-shape: 3 dot-separated base64url segments, second
+// and third at least 16 chars. Matches eyJ-prefixed tokens that the
+// keyword-anchored rule above would miss when they appear bare.
+var errorDetailJWTRE = regexp.MustCompile(`eyJ[A-Za-z0-9_-]{8,}\.[A-Za-z0-9_-]{16,}\.[A-Za-z0-9_-]{16,}`)
+
+const errorDetailBroadcastCap = 4096
+
+func sanitizeErrorDetailForBroadcast(s string) string {
+	if s == "" {
+		return s
+	}
+	// Cap first — a huge error body shouldn't tax every websocket
+	// client's buffer. 4096 matches the workspace-side _MAX_STDERR
+	// budget (it's actually larger here so the runtime's cap dominates).
+	if len(s) > errorDetailBroadcastCap {
+		s = s[:errorDetailBroadcastCap] + "…[truncated]"
+	}
+	s = errorDetailSecretRE.ReplaceAllString(s, "${1}[REDACTED]")
+	s = errorDetailJWTRE.ReplaceAllString(s, "[REDACTED]")
+	return s
+}
+
 type ActivityHandler struct {
 	broadcaster *events.Broadcaster
 }
@@ -691,6 +732,16 @@ func logActivityExec(ctx context.Context, exec activityExecutor, broadcaster eve
 		if respStr != nil {
 			payload["response_body"] = json.RawMessage(respJSON)
 		}
+		// internal#212 — surface the secret-safe failure reason on the
+		// live broadcast so the canvas chat-tab error banner can show
+		// the user *why* (provider HTTP status, error code, the
+		// provider's own human message) instead of the opaque
+		// "Agent error (Exception) — see workspace logs for details."
+		// hardcoded fallback. Omitted when nil so the canvas's "has
+		// actionable reason" guard doesn't trip on empty-string keys.
+		if params.ErrorDetail != nil && *params.ErrorDetail != "" {
+			payload["error_detail"] = sanitizeErrorDetailForBroadcast(*params.ErrorDetail)
+		}
 	}

 	return func() {
@@ -934,6 +934,184 @@ func TestLogActivity_Broadcast_IncludesRequestAndResponseBodies(t *testing.T) {
 	}
 }

+// TestLogActivity_Broadcast_IncludesErrorDetail pins the internal#212
+// UX fix: when an a2a_receive row is logged with status="error" and a
+// non-empty error_detail, the live broadcast MUST carry that detail so
+// the canvas chat-tab error bubble can render the actionable reason
+// (e.g. the provider's own 403 message) instead of the opaque
+// "Agent error (Exception) — see workspace logs for details." string.
+// Without this, the canvas falls back to the hardcoded boilerplate;
+// the row's error_detail is in the DB but never reaches the user
+// without a manual refresh of the Activity tab.
+func TestLogActivity_Broadcast_IncludesErrorDetail(t *testing.T) {
+	mock := setupTestDB(t)
+	defer mock.ExpectationsWereMet()
+
+	mock.ExpectExec("INSERT INTO activity_logs").
+		WillReturnResult(sqlmock.NewResult(1, 1))
+
+	cb := &recordingBroadcaster{}
+	srcID := "ws-source"
+	tgtID := "ws-target"
+	method := "message/send"
+	// Realistic actionable reason: provider HTTP status + provider's
+	// own message. Secret-safe (no token, no api key, just the cause).
+	detail := "Anthropic 403 oauth_org_not_allowed: Your organization has disabled Claude subscription access for Claude Code — use an Anthropic API key or ask your admin to enable access."
+
+	LogActivity(context.Background(), cb, ActivityParams{
+		WorkspaceID:  "ws-source",
+		ActivityType: "a2a_receive",
+		SourceID:     &srcID,
+		TargetID:     &tgtID,
+		Method:       &method,
+		Status:       "error",
+		ErrorDetail:  &detail,
+	})
+
+	if len(cb.calls) != 1 {
+		t.Fatalf("expected 1 broadcast, got %d", len(cb.calls))
+	}
+	payload := cb.calls[0].payload
+	got, ok := payload["error_detail"].(string)
+	if !ok {
+		t.Fatalf("error_detail missing from broadcast payload: got %#v", payload["error_detail"])
+	}
+	if got != detail {
+		t.Errorf("error_detail = %q, want %q", got, detail)
+	}
+}
+
+// TestLogActivity_Broadcast_OmitsErrorDetailWhenNil pins the inverse:
+// rows logged without an error_detail (the common ok-path) must not
+// have an empty "error_detail":"" key in the broadcast, which would
+// false-positive the canvas's "has actionable reason" guard and render
+// an empty Underlying-Error block. The omission rule matches how
+// request_body/response_body are handled.
+func TestLogActivity_Broadcast_OmitsErrorDetailWhenNil(t *testing.T) {
+	mock := setupTestDB(t)
+	defer mock.ExpectationsWereMet()
+
+	mock.ExpectExec("INSERT INTO activity_logs").
+		WillReturnResult(sqlmock.NewResult(1, 1))
+
+	cb := &recordingBroadcaster{}
+	srcID := "ws-source"
+
+	LogActivity(context.Background(), cb, ActivityParams{
+		WorkspaceID:  "ws-source",
+		ActivityType: "a2a_send",
+		SourceID:     &srcID,
+		Status:       "ok",
+		ErrorDetail:  nil,
+	})
+
+	if len(cb.calls) != 1 {
+		t.Fatalf("expected 1 broadcast, got %d", len(cb.calls))
+	}
+	if _, present := cb.calls[0].payload["error_detail"]; present {
+		t.Errorf("error_detail should be omitted when nil, got %v", cb.calls[0].payload["error_detail"])
+	}
+}
+
+// TestSanitizeErrorDetail_StripsSecretShapes pins the secret-safe
+// scrubber's contract: the broadcast layer is the last defense before
+// a string crosses the canvas WebSocket and lands in the user's
+// browser, so anything that *looks* like an API key / bearer token /
+// JWT must be replaced with [REDACTED] even if upstream (the runtime,
+// the provider) failed to scrub it. The non-secret parts of the
+// message — provider status, error code, human-readable cause — MUST
+// survive intact, otherwise the whole point of internal#212 (giving
+// the user an actionable reason) is defeated.
+func TestSanitizeErrorDetail_StripsSecretShapes(t *testing.T) {
+	cases := []struct {
+		name     string
+		in       string
+		mustHave []string // substrings that must survive — the actionable parts
+		mustMiss []string // substrings that must NOT survive — the secret shapes
+	}{
+		{
+			name:     "preserves actionable provider reason",
+			in:       "Anthropic 403 oauth_org_not_allowed: Your organization has disabled Claude subscription access for Claude Code",
+			mustHave: []string{"403", "oauth_org_not_allowed", "disabled Claude subscription"},
+			mustMiss: []string{"[REDACTED]"},
+		},
+		{
+			name:     "redacts sk- API key embedded in error",
+			in:       "openai 401 invalid_api_key: Incorrect API key provided: sk-proj-abcdefghijklmnop1234567890abcdef. Check your key.",
+			mustHave: []string{"401", "invalid_api_key", "Incorrect API key provided"},
+			mustMiss: []string{"sk-proj-abcdefghijklmnop1234567890abcdef"},
+		},
+		{
+			name:     "redacts Bearer token in stringified header dump",
+			in:       "auth failed; headers: Authorization: Bearer eyJhbGciOiJIUzI1NiJ9.aaaaaaaaaaaaaaaaaaaa.bbbbbbbbbbbbbbbbbbbb",
+			mustHave: []string{"auth failed"},
+			mustMiss: []string{"eyJhbGciOiJIUzI1NiJ9.aaaaaaaaaaaaaaaaaaaa.bbbbbbbbbbbbbbbbbbbb"},
+		},
+		{
+			name:     "truncates absurdly long detail to bound payload size",
+			in:       "kimi 500 internal_error: " + strings.Repeat("x", 8000),
+			mustHave: []string{"kimi 500 internal_error"},
+			mustMiss: []string{strings.Repeat("x", 5000)}, // 5000 in a row must NOT survive — cap is 4096
+		},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			got := sanitizeErrorDetailForBroadcast(tc.in)
+			for _, s := range tc.mustHave {
+				if !strings.Contains(got, s) {
+					t.Errorf("expected %q to survive scrub, got: %q", s, got)
+				}
+			}
+			for _, s := range tc.mustMiss {
+				if strings.Contains(got, s) {
+					t.Errorf("expected %q to be scrubbed, got: %q", s, got)
+				}
+			}
+		})
+	}
+}
+
+// TestLogActivity_Broadcast_ErrorDetailIsSanitized pins the integration
+// of the scrubber into the broadcast path: if an upstream caller
+// somehow passes through an error_detail with a secret-shaped token,
+// the wire payload (what reaches the canvas WebSocket) must already
+// be scrubbed. Defense in depth — the runtime should never let this
+// happen, but the canvas is the trust boundary, not the runtime.
+func TestLogActivity_Broadcast_ErrorDetailIsSanitized(t *testing.T) {
+	mock := setupTestDB(t)
+	defer mock.ExpectationsWereMet()
+
+	mock.ExpectExec("INSERT INTO activity_logs").
+		WillReturnResult(sqlmock.NewResult(1, 1))
+
+	cb := &recordingBroadcaster{}
+	srcID := "ws-source"
+	// Upstream leaked a token into the detail string. The DB still
+	// stores the unscrubbed copy (workspace logs are an internal
+	// audit surface), but the broadcast that reaches the canvas
+	// must already be sanitized.
+	detail := "anthropic 401 invalid_api_key: provided key sk-proj-leakedsecretvalueabcdefghij is wrong"
+
+	LogActivity(context.Background(), cb, ActivityParams{
+		WorkspaceID:  "ws-source",
+		ActivityType: "a2a_receive",
+		SourceID:     &srcID,
+		Status:       "error",
+		ErrorDetail:  &detail,
+	})
+
+	if len(cb.calls) != 1 {
+		t.Fatalf("expected 1 broadcast, got %d", len(cb.calls))
+	}
+	got, _ := cb.calls[0].payload["error_detail"].(string)
+	if strings.Contains(got, "sk-proj-leakedsecretvalueabcdefghij") {
+		t.Errorf("broadcast leaked secret-shaped token: %q", got)
+	}
+	if !strings.Contains(got, "invalid_api_key") {
+		t.Errorf("scrubber over-redacted: lost the actionable code from %q", got)
+	}
+}
+
 // TestLogActivityTx_DefersBroadcastUntilCommitHook pins the #149
 // contract: LogActivityTx returns a commitHook that the caller MUST
 // invoke after tx.Commit(); the broadcast MUST NOT fire from inside
@@ -0,0 +1,113 @@
+package handlers
+
+import "encoding/json"
+
+// agent_card_reconcile.go — server-side repair for the fleet-wide
+// agent-card identity gap.
+//
+// Root cause: the runtime builds its AgentCard from config.name
+// (workspace/main.py:198), and config.name is read from the
+// CP-regenerated /configs/config.yaml whose `name:` field is the raw
+// workspace UUID — NOT the friendly name the operator sees. The friendly
+// name IS captured: POST /workspaces and PATCH /workspaces/:id (the
+// canvas Details tab) write it to the trusted workspaces.name DB column.
+// But /registry/register stores the runtime-supplied card verbatim
+// (registry.go: `agent_card = EXCLUDED.agent_card`), so the stored card
+// served at /.well-known/agent-card.json and returned to peers via
+// agent_card_url ends up with name = UUID, description = "", role = null.
+//
+// Fix shape (deliberately minimal, no contract weakening): when the
+// runtime-supplied card's `name` is empty or equals the workspace UUID
+// (the placeholder the runtime had no better value for), the PLATFORM —
+// not the agent — substitutes the friendly value from the trusted
+// workspaces row. Identity stays platform-controlled: the agent never
+// gains the ability to self-set its own name/role; the platform sources
+// it from the operator-controlled DB column. We only ever FILL gaps
+// (empty / UUID-placeholder); a card that already carries a real
+// friendly name is never downgraded.
+//
+// list_peers / the /registry/:id/peers endpoint already resolve display
+// names from workspaces.name directly (discovery.go / mcp_tools.go
+// `SELECT w.id, w.name, ...`), so peer_name in delivered message tags
+// was already correct — this fix closes the remaining surface: the
+// agent_card blob itself (canvas Agent Card / Skills view, peer
+// agent_card_url fetches, the well-known card).
+//
+// description / role degrade discovery the same way: an empty
+// description and null role give peers nothing to reason about. We
+// default description from the (now reconciled) name when blank and
+// role from workspaces.role when the operator set one.
+
+// reconcileAgentCardIdentity patches identity gaps in a runtime-supplied
+// agent card from the trusted workspace DB row. It returns the
+// (possibly rewritten) card bytes and whether anything changed. On any
+// failure (malformed JSON, nothing to fill) it returns the input bytes
+// unchanged with changed=false so the caller can store them verbatim —
+// this is strictly no-worse-than-before, never a regression.
+//
+// Pure function: no DB / HTTP / globals, so it is exhaustively
+// unit-testable (agent_card_reconcile_test.go) without booting the
+// handler or a sqlmock.
+func reconcileAgentCardIdentity(card json.RawMessage, workspaceID, dbName, dbRole string) (json.RawMessage, bool) {
+	var m map[string]any
+	if err := json.Unmarshal(card, &m); err != nil || m == nil {
+		// Malformed card — not this function's job to reject it (the
+		// upsert stores it as-is and downstream readers handle bad
+		// JSON). Return verbatim so byte-for-byte behaviour is
+		// preserved on the failure path.
+		return card, false
+	}
+
+	changed := false
+
+	// name: fill only when empty or the UUID placeholder. A dbName that
+	// is itself the UUID is a placeholder row (registry.go INSERT seeds
+	// name = id before the canvas sets a friendly one) — not a friendly
+	// name, so it is not an eligible source.
+	cardName, _ := m["name"].(string)
+	if (cardName == "" || cardName == workspaceID) &&
+		dbName != "" && dbName != workspaceID {
+		m["name"] = dbName
+		changed = true
+	}
+
+	// description: when blank, default to the (reconciled) name so peers
+	// and the canvas Agent Card view have a non-empty human label
+	// instead of "". Mirrors the runtime's own
+	// `config.description or config.name` fallback (main.py:199) but
+	// applied to the registry copy where the runtime's fallback was the
+	// UUID.
+	if desc, _ := m["description"].(string); desc == "" {
+		if n, _ := m["name"].(string); n != "" && n != workspaceID {
+			m["description"] = n
+			changed = true
+		}
+	}
+
+	// role: surface the operator-set workspaces.role when the card
+	// carries none. Discovery (peer_role) and the canvas Role row read
+	// workspaces.role directly; this just makes the standalone card
+	// self-describing too. Never overwrite a role the card already has.
+	if dbRole != "" {
+		if r, ok := m["role"].(string); !ok || r == "" {
+			m["role"] = dbRole
+			changed = true
+		}
+	}
+
+	if !changed {
+		// No-op: return the original bytes untouched so callers that
+		// compare/store get byte-identical input (re-marshalling would
+		// reorder keys for no reason).
+		return card, false
+	}
+
+	out, err := json.Marshal(m)
+	if err != nil {
+		// Re-marshal of a map we just unmarshalled should never fail;
+		// if it somehow does, fall back to the verbatim input rather
+		// than storing nothing.
+		return card, false
+	}
+	return out, true
+}
@@ -0,0 +1,166 @@
+package handlers
+
+import (
+	"encoding/json"
+	"testing"
+)
+
+// TestReconcileAgentCardIdentity covers the server-side backfill that
+// repairs the fleet-wide agent-card identity gap (internal#XXX): the
+// runtime POSTs /registry/register with agent_card.name = the workspace
+// UUID (because the CP-regenerated /configs/config.yaml sets name: <uuid>)
+// while the trusted workspaces.name DB column — the value the canvas
+// Details tab shows and lets the operator edit — holds the friendly
+// name ("Claude Code Agent"). The platform reconciles them from the DB
+// row (NOT from the agent — identity stays platform-controlled, not
+// self-mutable).
+func TestReconcileAgentCardIdentity(t *testing.T) {
+	const wsID = "3b81321b-1ec7-488c-96f7-72c42a968da6"
+
+	tests := []struct {
+		name        string
+		card        string
+		dbName      string
+		dbRole      string
+		wantName    string
+		wantDesc    string
+		wantRole    string
+		wantChanged bool
+	}{
+		{
+			name:        "name is the workspace UUID — backfill from DB",
+			card:        `{"name":"3b81321b-1ec7-488c-96f7-72c42a968da6","description":"","capabilities":{"streaming":true}}`,
+			dbName:      "Claude Code Agent",
+			dbRole:      "",
+			wantName:    "Claude Code Agent",
+			wantDesc:    "Claude Code Agent",
+			wantRole:    "",
+			wantChanged: true,
+		},
+		{
+			name:        "empty name — backfill from DB",
+			card:        `{"name":"","description":"x"}`,
+			dbName:      "ops-agent",
+			dbRole:      "sre",
+			wantName:    "ops-agent",
+			wantDesc:    "x",
+			wantRole:    "sre",
+			wantChanged: true,
+		},
+		{
+			name:        "role null in card, DB has role — backfill role only",
+			card:        `{"name":"Reviewer","description":"Senior reviewer"}`,
+			dbName:      "Reviewer",
+			dbRole:      "code-reviewer",
+			wantName:    "Reviewer",
+			wantDesc:    "Senior reviewer",
+			wantRole:    "code-reviewer",
+			wantChanged: true,
+		},
+		{
+			name: "card already has a real friendly name — do NOT clobber it",
+			// A richer card (e.g. an external channel agent) must win;
+			// the platform only fills gaps, never downgrades.
+			card:        `{"name":"Claude Code (channel)","description":"Local Claude Code session bridged","role":"assistant"}`,
+			dbName:      "hongming-pc",
+			dbRole:      "operator",
+			wantName:    "Claude Code (channel)",
+			wantDesc:    "Local Claude Code session bridged",
+			wantRole:    "assistant",
+			wantChanged: false,
+		},
+		{
+			name:        "no DB name available — leave UUID name untouched (no worse than before)",
+			card:        `{"name":"3b81321b-1ec7-488c-96f7-72c42a968da6","description":""}`,
+			dbName:      "",
+			dbRole:      "",
+			wantName:    "3b81321b-1ec7-488c-96f7-72c42a968da6",
+			wantDesc:    "",
+			wantRole:    "",
+			wantChanged: false,
+		},
+		{
+			name:        "dbName equals UUID (placeholder row) — not a friendly name, leave untouched",
+			card:        `{"name":"3b81321b-1ec7-488c-96f7-72c42a968da6"}`,
+			dbName:      "3b81321b-1ec7-488c-96f7-72c42a968da6",
+			dbRole:      "",
+			wantName:    "3b81321b-1ec7-488c-96f7-72c42a968da6",
+			wantDesc:    "",
+			wantRole:    "",
+			wantChanged: false,
+		},
+		{
+			name:        "malformed card JSON — return unchanged, no panic",
+			card:        `{not json`,
+			dbName:      "Claude Code Agent",
+			dbRole:      "",
+			wantChanged: false,
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			out, changed := reconcileAgentCardIdentity(
+				json.RawMessage(tc.card), wsID, tc.dbName, tc.dbRole,
+			)
+			if changed != tc.wantChanged {
+				t.Fatalf("changed = %v, want %v", changed, tc.wantChanged)
+			}
+			if !tc.wantChanged {
+				// Unchanged path must return the input bytes verbatim.
+				if string(out) != tc.card {
+					t.Fatalf("unchanged path mutated bytes:\n got  %s\n want %s", out, tc.card)
+				}
+				return
+			}
+			var got map[string]any
+			if err := json.Unmarshal(out, &got); err != nil {
+				t.Fatalf("output not valid JSON: %v (%s)", err, out)
+			}
+			if g, _ := got["name"].(string); g != tc.wantName {
+				t.Errorf("name = %q, want %q", g, tc.wantName)
+			}
+			if g, _ := got["description"].(string); g != tc.wantDesc {
+				t.Errorf("description = %q, want %q", g, tc.wantDesc)
+			}
+			if tc.wantRole != "" {
+				if g, _ := got["role"].(string); g != tc.wantRole {
+					t.Errorf("role = %q, want %q", g, tc.wantRole)
+				}
+			}
+		})
+	}
+}
+
+// TestReconcileAgentCardIdentity_PreservesOtherFields ensures the
+// reconcile is a minimal in-place patch — capabilities, version,
+// skills and any unknown future fields survive untouched.
+func TestReconcileAgentCardIdentity_PreservesOtherFields(t *testing.T) {
+	card := `{"name":"ws-uuid","description":"","version":"1.0.0",` +
+		`"capabilities":{"streaming":true,"pushNotifications":true},` +
+		`"skills":[{"id":"a","name":"a"}],"configuration_status":"ready"}`
+	out, changed := reconcileAgentCardIdentity(
+		json.RawMessage(card), "ws-uuid", "Friendly Name", "",
+	)
+	if !changed {
+		t.Fatal("expected changed = true")
+	}
+	var got map[string]any
+	if err := json.Unmarshal(out, &got); err != nil {
+		t.Fatalf("invalid JSON: %v", err)
+	}
+	if got["version"] != "1.0.0" {
+		t.Errorf("version not preserved: %v", got["version"])
+	}
+	if got["configuration_status"] != "ready" {
+		t.Errorf("configuration_status not preserved: %v", got["configuration_status"])
+	}
+	caps, ok := got["capabilities"].(map[string]any)
+	if !ok || caps["streaming"] != true {
+		t.Errorf("capabilities not preserved: %v", got["capabilities"])
+	}
+	skills, ok := got["skills"].([]any)
+	if !ok || len(skills) != 1 {
+		t.Errorf("skills not preserved: %v", got["skills"])
+	}
+}
@@ -1,6 +1,9 @@
 package handlers

 import (
+	"log"
+	"os"
+	"path/filepath"
 	"regexp"
 	"strings"
 )
@@ -17,6 +20,17 @@ var gitIdentitySlugPattern = regexp.MustCompile(`[^a-z0-9]+`)
 // docs/authorship.md (when it exists).
 const gitIdentityEmailDomain = "agents.moleculesai.app"

+// gitAskpassHelperPath is the in-container path of the askpass helper
+// installed by every workspace runtime image (workspace/Dockerfile in
+// molecule-core; scripts/git-askpass.sh → /usr/local/bin/molecule-askpass
+// in each external template-* repo). The helper reads GIT_HTTP_USERNAME
+// / GIT_HTTP_PASSWORD (falling back to GITEA_USER / GITEA_TOKEN) from
+// env and emits them on the git credential-prompt protocol. Setting
+// GIT_ASKPASS to this path is what wires container-side HTTPS git auth
+// to the persona credentials already arriving via workspace_secrets,
+// with no on-disk .gitconfig / .git-credentials mutation required.
+const gitAskpassHelperPath = "/usr/local/bin/molecule-askpass"
+
 // applyAgentGitIdentity sets GIT_AUTHOR_* / GIT_COMMITTER_* env vars so
 // every commit from this workspace container carries a distinct author
 // in `git log` and `git blame`. Git reads these env vars before falling
@@ -50,6 +64,125 @@ func applyAgentGitIdentity(envVars map[string]string, workspaceName string) {
 	setIfEmpty(envVars, "GIT_AUTHOR_EMAIL", authorEmail)
 	setIfEmpty(envVars, "GIT_COMMITTER_NAME", authorName)
 	setIfEmpty(envVars, "GIT_COMMITTER_EMAIL", authorEmail)
+
+	applyGitAskpass(envVars)
+}
+
+// applyGitAskpass points git at the in-image askpass helper so that any
+// HTTPS git operation against a remote without a pre-configured
+// credential.helper picks up the persona credentials already present in
+// the container env (GIT_HTTP_USERNAME / GIT_HTTP_PASSWORD, or
+// GITEA_USER / GITEA_TOKEN as fallback — the latter pair is what
+// loadPersonaEnvFile delivers from the operator-host bootstrap kit).
+//
+// Idempotent: if GIT_ASKPASS is already set (e.g. by an operator-
+// supplied workspace_secret or an env-mutator plugin), the existing
+// value wins. This lets a workspace opt out by setting GIT_ASKPASS=""
+// or pointing at a different helper.
+//
+// No vendor-specific behaviour lives in this function — the host the
+// credentials apply to is determined entirely by the deployer choosing
+// when to populate GIT_HTTP_USERNAME / GIT_HTTP_PASSWORD (or
+// GITEA_USER / GITEA_TOKEN). The helper script itself is generic and
+// has no hardcoded hostnames, so it's safe to ship inside the
+// open-source workspace template images alongside the platform-managed
+// claude-code image.
+func applyGitAskpass(envVars map[string]string) {
+	if envVars == nil {
+		return
+	}
+	setIfEmpty(envVars, "GIT_ASKPASS", gitAskpassHelperPath)
+}
+
+// applyAgentGitHTTPCreds reads the persona's HTTPS git credential from
+// the operator-host bootstrap dir and injects it as GIT_HTTP_USERNAME /
+// GIT_HTTP_PASSWORD so the in-container askpass helper can emit it on
+// git's auth challenge.
+//
+// Why a dedicated env-var pair instead of reusing GITEA_USER / GITEA_TOKEN:
+// the provisioner's forensic #145 denylist (provisioner.scmWriteTokenKeys)
+// strips any env var named GITEA_TOKEN / GITHUB_TOKEN / GH_TOKEN /
+// GITLAB_TOKEN / GL_TOKEN / BITBUCKET_TOKEN from tenant container env
+// before docker run. That denylist is by exact key name, so the same
+// token survives transport when shipped under the generic
+// GIT_HTTP_USERNAME / GIT_HTTP_PASSWORD names that the askpass helper
+// reads first (scripts/git-askpass.sh in each template-*). The username
+// half stays an identifier (the persona's Gitea login), the password
+// half carries the bytes from the persona token file.
+//
+// The fallback pair GITEA_USER / GITEA_TOKEN is ALSO set — GITEA_USER
+// survives the denylist (it's an identity, not a credential) and
+// GITEA_TOKEN is the no-op write that buildContainerEnv will drop.
+// Both pairs in lockstep means the askpass helper's GIT_HTTP_*-first /
+// GITEA_*-fallback chain works regardless of which lane lands first in
+// the container env on any future provisioner refactor.
+//
+// Idempotent: existing GIT_HTTP_USERNAME / GIT_HTTP_PASSWORD keys are
+// preserved. Operator-supplied workspace_secrets win over the persona
+// token file by virtue of running BEFORE this helper in
+// prepareProvisionContext.
+//
+// Silent no-op when:
+//   - personaKey is empty (no role → no persona dir to consult)
+//   - personaKey fails the safe-segment check (defense-in-depth against
+//     a crafted role escaping the persona dir)
+//   - the persona token file does not exist or is empty (legitimate
+//     case for personas that don't ship a git-write credential — e.g.
+//     read-only PM/Reviewer/Researcher identities or a partially-
+//     provisioned bootstrap)
+//
+// No vendor-specific behaviour: this function reads bytes from a path
+// and emits them as the standard askpass env-var pair. The host the
+// credential applies to is determined by the deployer choosing which
+// remote to push to — the askpass helper has no hardcoded hostnames.
+func applyAgentGitHTTPCreds(envVars map[string]string, personaKey string) {
+	if envVars == nil {
+		return
+	}
+	personaKey = strings.TrimSpace(personaKey)
+	if !isSafeRoleName(personaKey) {
+		// Silent no-op for empty / unsafe keys — same shape as
+		// loadPersonaTokenFile. Descriptive-role payloads (multi-word
+		// "Frontend Engineer" etc.) take this branch and pick up
+		// creds via workspace_secrets / org-import persona-env merge,
+		// not the direct persona-token file path.
+		return
+	}
+	root := os.Getenv("MOLECULE_PERSONA_ROOT")
+	if root == "" {
+		root = "/etc/molecule-bootstrap/personas"
+	}
+	tokenPath := filepath.Join(root, personaKey, "token")
+	data, err := os.ReadFile(tokenPath)
+	if err != nil {
+		// Persona dir / file absent: legitimate for the host shapes
+		// that don't ship the bootstrap kit (dev laptops, CI nodes)
+		// or for personas that intentionally carry no git-write
+		// credential. Caller decides whether the resulting
+		// "Authentication failed" at first push is a configuration
+		// error or expected behaviour.
+		return
+	}
+	token := strings.TrimSpace(string(data))
+	if token == "" {
+		return
+	}
+
+	// Primary lane — survives forensic #145 by virtue of the generic
+	// GIT_HTTP_* names not being on the SCM-write denylist.
+	setIfEmpty(envVars, "GIT_HTTP_USERNAME", personaKey)
+	setIfEmpty(envVars, "GIT_HTTP_PASSWORD", token)
+
+	// Fallback lane — askpass reads GITEA_USER / GITEA_TOKEN second.
+	// GITEA_USER survives the denylist; GITEA_TOKEN will be stripped
+	// by buildContainerEnv but is set here for completeness so the
+	// (envVars map[string]string) contract is consistent for callers
+	// inspecting it before the provisioner-level filter runs (e.g.
+	// the env-mutator plugin chain).
+	setIfEmpty(envVars, "GITEA_USER", personaKey)
+	setIfEmpty(envVars, "GITEA_TOKEN", token)
+
+	log.Printf("applyAgentGitHTTPCreds: injected GIT_HTTP_USERNAME/PASSWORD for persona %q (token %d bytes)", personaKey, len(token))
 }

 // slugifyForEmail collapses a workspace name to a safe email localpart:
@@ -1,6 +1,8 @@
 package handlers

 import (
+	"os"
+	"path/filepath"
 	"testing"
 )

@@ -75,6 +77,261 @@ func TestApplyAgentGitIdentity_NilMapIsSafe(t *testing.T) {
 	applyAgentGitIdentity(nil, "PM")
 }

+func TestApplyAgentGitIdentity_SetsGitAskpass(t *testing.T) {
+	// GIT_ASKPASS is what wires container-side HTTPS git auth to the
+	// persona credentials (GITEA_USER/GITEA_TOKEN, etc.) that
+	// loadPersonaEnvFile delivers via workspace_secrets. Without this,
+	// `git push` inside the container would fall through to interactive
+	// prompts (impossible) or a missing credential.helper (401).
+	env := map[string]string{}
+	applyAgentGitIdentity(env, "Frontend Engineer")
+	if env["GIT_ASKPASS"] != "/usr/local/bin/molecule-askpass" {
+		t.Errorf("GIT_ASKPASS: got %q, want %q",
+			env["GIT_ASKPASS"], "/usr/local/bin/molecule-askpass")
+	}
+}
+
+func TestApplyAgentGitIdentity_RespectsAskpassOverride(t *testing.T) {
+	// A workspace_secret or env-mutator plugin must be able to point at
+	// a custom askpass helper without us clobbering it. Symmetric with
+	// the GIT_AUTHOR_NAME override test above.
+	env := map[string]string{
+		"GIT_ASKPASS": "/opt/custom/askpass",
+	}
+	applyAgentGitIdentity(env, "Backend Engineer")
+	if env["GIT_ASKPASS"] != "/opt/custom/askpass" {
+		t.Errorf("GIT_ASKPASS should not be overwritten, got %q", env["GIT_ASKPASS"])
+	}
+}
+
+func TestApplyAgentGitIdentity_AskpassSkippedOnEmptyName(t *testing.T) {
+	// The empty-name early-return covers GIT_ASKPASS too — a provisioning
+	// glitch that dropped the workspace name shouldn't half-configure the
+	// container (identity vars empty but askpass wired). All-or-nothing.
+	env := map[string]string{}
+	applyAgentGitIdentity(env, "")
+	if _, ok := env["GIT_ASKPASS"]; ok {
+		t.Errorf("empty name should not set GIT_ASKPASS, got %q", env["GIT_ASKPASS"])
+	}
+}
+
+func TestApplyGitAskpass_NilMapIsSafe(t *testing.T) {
+	defer func() {
+		if r := recover(); r != nil {
+			t.Errorf("applyGitAskpass panicked on nil map: %v", r)
+		}
+	}()
+	applyGitAskpass(nil)
+}
+
+// TestApplyAgentGitHTTPCreds_HappyPath: the prod-team shape — a persona
+// dir at /etc/molecule-bootstrap/personas/<role>/token ships a write
+// token. applyAgentGitHTTPCreds reads it and emits both the
+// askpass-preferred GIT_HTTP_* pair and the GITEA_* fallback.
+func TestApplyAgentGitHTTPCreds_HappyPath(t *testing.T) {
+	root := t.TempDir()
+	roleDir := filepath.Join(root, "agent-dev-a")
+	if err := os.MkdirAll(roleDir, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(filepath.Join(roleDir, "token"),
+		[]byte("token-bytes-redacted\n"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("MOLECULE_PERSONA_ROOT", root)
+
+	env := map[string]string{}
+	applyAgentGitHTTPCreds(env, "agent-dev-a")
+
+	cases := map[string]string{
+		"GIT_HTTP_USERNAME": "agent-dev-a",
+		"GIT_HTTP_PASSWORD": "token-bytes-redacted",
+		"GITEA_USER":        "agent-dev-a",
+		"GITEA_TOKEN":       "token-bytes-redacted",
+	}
+	for k, want := range cases {
+		if got := env[k]; got != want {
+			t.Errorf("%s: got %q, want %q", k, got, want)
+		}
+	}
+}
+
+// TestApplyAgentGitHTTPCreds_TrimsWhitespace: bootstrap-kit-written
+// token files canonically end in \n. Must trim like loadPersonaTokenFile
+// does — Gitea PAT validator rejects embedded whitespace.
+func TestApplyAgentGitHTTPCreds_TrimsWhitespace(t *testing.T) {
+	root := t.TempDir()
+	roleDir := filepath.Join(root, "agent-dev-b")
+	if err := os.MkdirAll(roleDir, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(filepath.Join(roleDir, "token"),
+		[]byte("\n  raw-token-bytes  \n\n"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("MOLECULE_PERSONA_ROOT", root)
+
+	env := map[string]string{}
+	applyAgentGitHTTPCreds(env, "agent-dev-b")
+	if env["GIT_HTTP_PASSWORD"] != "raw-token-bytes" {
+		t.Errorf("GIT_HTTP_PASSWORD: token whitespace not trimmed; got %q", env["GIT_HTTP_PASSWORD"])
+	}
+}
+
+// TestApplyAgentGitHTTPCreds_RespectsOperatorOverride: if a workspace
+// secret (loaded earlier by loadWorkspaceSecrets) already set the
+// askpass pair, those values must win — operator intent ranks above
+// persona-file defaults. Symmetric with applyAgentGitIdentity's
+// GIT_AUTHOR_* override semantics.
+func TestApplyAgentGitHTTPCreds_RespectsOperatorOverride(t *testing.T) {
+	root := t.TempDir()
+	roleDir := filepath.Join(root, "agent-dev-a")
+	if err := os.MkdirAll(roleDir, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(filepath.Join(roleDir, "token"),
+		[]byte("file-token\n"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("MOLECULE_PERSONA_ROOT", root)
+
+	env := map[string]string{
+		"GIT_HTTP_USERNAME": "operator-user",
+		"GIT_HTTP_PASSWORD": "operator-secret",
+	}
+	applyAgentGitHTTPCreds(env, "agent-dev-a")
+
+	if env["GIT_HTTP_USERNAME"] != "operator-user" {
+		t.Errorf("GIT_HTTP_USERNAME should not be overwritten, got %q", env["GIT_HTTP_USERNAME"])
+	}
+	if env["GIT_HTTP_PASSWORD"] != "operator-secret" {
+		t.Errorf("GIT_HTTP_PASSWORD should not be overwritten, got %q", env["GIT_HTTP_PASSWORD"])
+	}
+	// Fallback pair was not pre-set, so persona-file fills it in.
+	if env["GITEA_TOKEN"] != "file-token" {
+		t.Errorf("GITEA_TOKEN fallback should be filled, got %q", env["GITEA_TOKEN"])
+	}
+}
+
+// TestApplyAgentGitHTTPCreds_EmptyKeyIsNoop: a workspace with an empty
+// payload.Role (descriptive multi-word role, or no role) must take the
+// silent-no-op branch — no FS read, no env keys touched.
+func TestApplyAgentGitHTTPCreds_EmptyKeyIsNoop(t *testing.T) {
+	root := t.TempDir()
+	t.Setenv("MOLECULE_PERSONA_ROOT", root)
+
+	env := map[string]string{}
+	applyAgentGitHTTPCreds(env, "")
+	if len(env) != 0 {
+		t.Errorf("empty persona key should leave env untouched, got %v", env)
+	}
+	applyAgentGitHTTPCreds(env, "   ")
+	if len(env) != 0 {
+		t.Errorf("whitespace persona key should leave env untouched, got %v", env)
+	}
+	applyAgentGitHTTPCreds(env, "Frontend Engineer")
+	if len(env) != 0 {
+		t.Errorf("multi-word descriptive role should leave env untouched (silent no-op via isSafeRoleName), got %v", env)
+	}
+}
+
+// TestApplyAgentGitHTTPCreds_MissingTokenFile: persona dir exists but
+// ships no token (legitimate for read-only personas like agent-pm pre-
+// CTO-cred or partially-provisioned bootstrap). Silent no-op — no env
+// keys set so first push surfaces "Authentication failed" cleanly
+// instead of half-configured creds.
+func TestApplyAgentGitHTTPCreds_MissingTokenFile(t *testing.T) {
+	root := t.TempDir()
+	if err := os.MkdirAll(filepath.Join(root, "agent-pm"), 0o755); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("MOLECULE_PERSONA_ROOT", root)
+
+	env := map[string]string{}
+	applyAgentGitHTTPCreds(env, "agent-pm")
+	if len(env) != 0 {
+		t.Errorf("missing token file should leave env untouched, got %v", env)
+	}
+}
+
+// TestApplyAgentGitHTTPCreds_EmptyTokenIsNoop: a whitespace-only token
+// file (botched bootstrap) must be treated as absent — never emit
+// GIT_HTTP_PASSWORD="" because the askpass helper would then return
+// empty on the password prompt and git would surface a confusing 401
+// rather than a clean "no credentials" state.
+func TestApplyAgentGitHTTPCreds_EmptyTokenIsNoop(t *testing.T) {
+	root := t.TempDir()
+	roleDir := filepath.Join(root, "agent-dev-a")
+	if err := os.MkdirAll(roleDir, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(filepath.Join(roleDir, "token"),
+		[]byte("   \t\n  \n"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("MOLECULE_PERSONA_ROOT", root)
+
+	env := map[string]string{}
+	applyAgentGitHTTPCreds(env, "agent-dev-a")
+	if len(env) != 0 {
+		t.Errorf("whitespace-only token should leave env untouched, got %v", env)
+	}
+}
+
+// TestApplyAgentGitHTTPCreds_RejectsUnsafeRole: defense-in-depth — a
+// crafted role with path separators / "../" must NOT touch the FS,
+// even if a token file exists at the traversed location.
+func TestApplyAgentGitHTTPCreds_RejectsUnsafeRole(t *testing.T) {
+	root := t.TempDir()
+	// Plant a token at <root>/token so a successful traversal would land here.
+	if err := os.WriteFile(filepath.Join(root, "token"),
+		[]byte("stolen-token\n"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("MOLECULE_PERSONA_ROOT", filepath.Join(root, "personas"))
+
+	for _, bad := range []string{"..", "../personas", "/abs", "with/slash", "."} {
+		env := map[string]string{}
+		applyAgentGitHTTPCreds(env, bad)
+		if len(env) != 0 {
+			t.Errorf("unsafe role %q must leave env untouched, got %v", bad, env)
+		}
+	}
+}
+
+// TestApplyAgentGitHTTPCreds_NilMapIsSafe: defensive — never panic
+// on a nil map. Symmetric with applyAgentGitIdentity's nil-map test.
+func TestApplyAgentGitHTTPCreds_NilMapIsSafe(t *testing.T) {
+	defer func() {
+		if r := recover(); r != nil {
+			t.Errorf("applyAgentGitHTTPCreds panicked on nil map: %v", r)
+		}
+	}()
+	applyAgentGitHTTPCreds(nil, "agent-dev-a")
+}
+
+// TestApplyAgentGitHTTPCreds_DefaultPersonaRoot: when
+// MOLECULE_PERSONA_ROOT is unset, the helper falls back to
+// /etc/molecule-bootstrap/personas — the canonical operator-host path
+// per the bootstrap kit shape. We can't write into /etc in a test,
+// but we CAN assert the helper takes the silent-no-op branch when
+// that real path is absent (the prod-default case on a dev laptop).
+func TestApplyAgentGitHTTPCreds_DefaultPersonaRoot(t *testing.T) {
+	t.Setenv("MOLECULE_PERSONA_ROOT", "")
+
+	env := map[string]string{}
+	applyAgentGitHTTPCreds(env, "agent-dev-a")
+	// The /etc/molecule-bootstrap/personas/agent-dev-a/token path
+	// almost certainly does not exist on a dev/CI host. The contract
+	// here is "silent no-op when token unreadable", not "exact env
+	// state" — so we only assert no panic + no half-state pair.
+	if _, ok := env["GIT_HTTP_USERNAME"]; ok {
+		if _, ok2 := env["GIT_HTTP_PASSWORD"]; !ok2 {
+			t.Errorf("USERNAME set without PASSWORD — half-state; got %v", env)
+		}
+	}
+}
+
 func TestSlugifyForEmail(t *testing.T) {
 	cases := []struct {
 		in, want string
@@ -163,8 +163,32 @@ func (h *DelegationHandler) Delegate(c *gin.Context) {
 		},
 	})

-	// Fire-and-forget: send A2A in background goroutine
-	go h.executeDelegation(ctx, sourceID, body.TargetID, delegationID, a2aBody)
+	// Fire-and-forget: send A2A in a background goroutine.
+	//
+	// internal#497 — the goroutine MUST NOT inherit the HTTP request's
+	// cancellation. `ctx` here is c.Request.Context(); the handler returns
+	// 202 a few lines below, which cancels that context immediately. Before
+	// this fix (regression ce2db75f) executeDelegation ran on the
+	// request-scoped ctx, so every DB op + proxy call in the detached
+	// goroutine failed `context canceled` the instant the 202 was written.
+	// That silently broke 100% of A2A peer delegations fleet-wide since
+	// 2026-05-12 (poll-mode peers never got their a2a_receive inbox row;
+	// lookupDeliveryMode swallowed the ctx error and defaulted to push).
+	//
+	// context.WithoutCancel detaches cancellation/deadline while PRESERVING
+	// all context values (trace/correlation/tenant ids that proxyA2ARequest
+	// and the broadcaster read off ctx) — this is the established pattern in
+	// this package (a2a_proxy.go:850, a2a_proxy_helpers.go:525,
+	// registry.go:822). The 30-minute ceiling matches the prior internal
+	// budget executeDelegation used before ce2db75f and the proxy's own
+	// absolute agent-dispatch ceiling (a2a_proxy.go forwardCtx).
+	delegationCtx, cancelDelegation := context.WithTimeout(
+		context.WithoutCancel(ctx), 30*time.Minute,
+	)
+	go func() {
+		defer cancelDelegation()
+		h.executeDelegation(delegationCtx, sourceID, body.TargetID, delegationID, a2aBody)
+	}()

 	// Broadcast event so canvas shows delegation in real-time
 	h.broadcaster.RecordAndBroadcast(ctx, string(events.EventDelegationSent), sourceID, map[string]interface{}{
@@ -16,6 +16,65 @@ import (
 	"github.com/gin-gonic/gin"
 )

+// ---------- internal#497 regression: detached goroutine ctx must outlive the handler ----------
+
+// TestDelegate_DetachedContext_SurvivesRequestCancellation pins the
+// load-bearing invariant that regression ce2db75f violated: the context
+// handed to executeDelegation in the fire-and-forget goroutine must NOT be
+// cancelled when the HTTP handler returns 202 (which cancels
+// c.Request.Context()). Before the fix, executeDelegation ran on the
+// request-scoped ctx, so every DB op + proxy call failed `context
+// canceled` the instant the 202 was written — silently breaking 100% of
+// A2A peer delegations fleet-wide since 2026-05-12.
+//
+// This test asserts the exact ctx-derivation contract used by Delegate
+// (context.WithoutCancel(parent) + a timeout budget): the derived context
+// (a) stays alive after the parent is cancelled, and (b) still carries
+// parent values (trace/correlation/tenant ids the downstream proxy +
+// broadcaster read off ctx). It is intentionally DB-free and fast.
+func TestDelegate_DetachedContext_SurvivesRequestCancellation(t *testing.T) {
+	type ctxKey string
+	const traceKey ctxKey = "trace-id"
+
+	// Simulate c.Request.Context() carrying a correlation value.
+	parent, cancelParent := context.WithCancel(
+		context.WithValue(context.Background(), traceKey, "trace-abc-123"),
+	)
+
+	// Exact derivation Delegate uses for the detached goroutine.
+	delegationCtx, cancelDelegation := context.WithTimeout(
+		context.WithoutCancel(parent), 30*time.Minute,
+	)
+	defer cancelDelegation()
+
+	// The HTTP handler "returns 202" → request context is cancelled.
+	cancelParent()
+
+	if err := parent.Err(); err == nil {
+		t.Fatal("precondition: parent context should be cancelled after the handler returns")
+	}
+
+	// (a) Cancellation MUST NOT propagate to the detached context.
+	select {
+	case <-delegationCtx.Done():
+		t.Fatalf("regression: detached delegation ctx was cancelled by the handler returning (err=%v) — executeDelegation would fail every DB op with `context canceled`", delegationCtx.Err())
+	default:
+		// alive — correct
+	}
+
+	// (b) Parent values MUST still be readable (WithoutCancel preserves
+	// values; trace/correlation/tenant ids the proxy + broadcaster use).
+	if got, _ := delegationCtx.Value(traceKey).(string); got != "trace-abc-123" {
+		t.Errorf("detached ctx lost the parent trace value: got %q, want %q", got, "trace-abc-123")
+	}
+
+	// And it still has a real deadline (the 30m budget), so it is not an
+	// unbounded background context.
+	if _, hasDeadline := delegationCtx.Deadline(); !hasDeadline {
+		t.Error("detached ctx must carry the 30-minute timeout budget, but has no deadline")
+	}
+}
+
 // ---------- Delegate: missing target_id → 400 ----------

 func TestDelegate_MissingTargetID(t *testing.T) {
@@ -24,17 +24,30 @@ import (

 // BuildExternalConnectionPayload assembles the gin.H payload that the
 // canvas's ExternalConnectModal consumes. Pure data — caller owns DB
-// reads (workspace_id) and token minting (auth_token).
+// reads (workspace_id, workspace_name) and token minting (auth_token).
 //
 // authToken may be empty for the read-only "show instructions again"
 // path; the modal masks the field in that case rather than displaying
 // an empty string.
-func BuildExternalConnectionPayload(platformURL, workspaceID, authToken string) gin.H {
+//
+// workspaceName feeds the per-workspace MCP server-name in the snippets
+// that wire molecule-mcp into an external Claude Code (or other
+// MCP-stdio) client. Without a unique server name a second
+// `claude mcp add molecule` call REPLACES the first entry, collapsing
+// multi-workspace use into a single per-session slot — see
+// mcpServerNameForWorkspace below. May be empty (re-show / rotate paths
+// that don't plumb the name); the helper falls back to the workspace
+// ID's short prefix so the snippet is always unique.
+func BuildExternalConnectionPayload(platformURL, workspaceID, workspaceName, authToken string) gin.H {
 	pURL := strings.TrimSuffix(platformURL, "/")
+	mcpName := mcpServerNameForWorkspace(workspaceID, workspaceName)
 	stamp := func(tmpl string) string {
 		return strings.ReplaceAll(
-			strings.ReplaceAll(tmpl, "{{PLATFORM_URL}}", pURL),
-			"{{WORKSPACE_ID}}", workspaceID,
+			strings.ReplaceAll(
+				strings.ReplaceAll(tmpl, "{{PLATFORM_URL}}", pURL),
+				"{{WORKSPACE_ID}}", workspaceID,
+			),
+			"{{MCP_SERVER_NAME}}", mcpName,
 		)
 	}
 	return gin.H{
@@ -77,6 +90,81 @@ func externalPlatformURL(c *gin.Context) string {
 	return scheme + "://" + host
 }

+// mcpServerNameForWorkspace derives the unique MCP server name used in
+// the Universal MCP snippet's `claude mcp add <name> -- ...` line.
+//
+// Why per-workspace, not a fixed "molecule": `claude mcp add` keys
+// entries by name in ~/.claude.json, so re-running with the same name
+// silently REPLACES the previous entry. A single external Claude Code
+// session that connects to N molecule workspaces must therefore use N
+// distinct server names — otherwise the second install collapses the
+// first, and the user experiences "MCP is per-session". MCP itself
+// supports many servers per session; the install-snippet name was the
+// only thing standing in the way.
+//
+// Pattern: "molecule-<slug>" where slug comes from the workspace name
+// (lowercased, non-alphanumeric → hyphen, collapsed, trimmed, <=24
+// chars). Falls back to the workspace ID's first 8 chars when the name
+// is empty or slugifies to nothing — both produce a deterministic,
+// Claude-Code-name-safe (alphanumeric + hyphens, no spaces / dots /
+// slashes) identifier that disambiguates per-workspace.
+//
+// Two workspaces with identical names still produce identical slugs by
+// design — the user picked them to look the same. The
+// `claude mcp add` step will overwrite the older one in that case;
+// the workaround is to rename one, then re-run. Documented in the
+// snippet header so users aren't surprised.
+func mcpServerNameForWorkspace(workspaceID, workspaceName string) string {
+	const fallbackIDPrefixLen = 8
+	const maxSlugLen = 24
+	slug := slugifyForMcpName(workspaceName, maxSlugLen)
+	if slug == "" {
+		id := strings.ReplaceAll(workspaceID, "-", "")
+		if len(id) > fallbackIDPrefixLen {
+			id = id[:fallbackIDPrefixLen]
+		}
+		slug = id
+	}
+	if slug == "" {
+		// Defensive: empty workspaceID at this layer means the caller
+		// is misusing the API; we still return a usable (non-colliding
+		// in the common case) constant rather than producing "molecule-"
+		// which Claude Code would reject.
+		return "molecule"
+	}
+	return "molecule-" + slug
+}
+
+// slugifyForMcpName lowercases, replaces non-[a-z0-9] runs with a single
+// '-', trims leading/trailing '-', and truncates to maxLen. Returns ""
+// if nothing usable remains. Pure helper; no allocations beyond the
+// builder.
+func slugifyForMcpName(s string, maxLen int) string {
+	var b strings.Builder
+	b.Grow(len(s))
+	lastHyphen := true // suppress leading hyphens
+	for _, r := range s {
+		switch {
+		case r >= 'A' && r <= 'Z':
+			b.WriteRune(r + ('a' - 'A'))
+			lastHyphen = false
+		case (r >= 'a' && r <= 'z') || (r >= '0' && r <= '9'):
+			b.WriteRune(r)
+			lastHyphen = false
+		default:
+			if !lastHyphen {
+				b.WriteByte('-')
+				lastHyphen = true
+			}
+		}
+	}
+	out := strings.TrimRight(b.String(), "-")
+	if len(out) > maxLen {
+		out = strings.TrimRight(out[:maxLen], "-")
+	}
+	return out
+}
+
 // externalCurlTemplate — zero-dependency register snippet. Placeholders:
 //   - {{PLATFORM_URL}}, {{WORKSPACE_ID}}   — filled server-side
 //   - $WORKSPACE_AUTH_TOKEN                — env var, operator sets
@@ -216,6 +304,14 @@ const externalUniversalMcpTemplate = `# Universal MCP — standalone register +
 # for any MCP-aware runtime (Claude Code, hermes, codex, etc.).
 # Pair with the Claude Code or Python SDK tab if your runtime needs
 # inbound A2A delivery (canvas messages → agent conversation turns).
+#
+# Multi-workspace: MCP supports many servers per Claude Code session.
+# This snippet uses a workspace-specific server name ({{MCP_SERVER_NAME}})
+# so installing for a second workspace ADDS another entry instead of
+# overwriting the first — run the snippet from each workspace's modal
+# in turn and ` + "`claude mcp list`" + ` will show all of them. If two
+# workspaces have the same name, slugs collide and the second install
+# overwrites the first; rename one workspace to disambiguate.

 # Requires Python >= 3.11. On 3.10 or older pip says
 # "Could not find a version that satisfies the requirement
@@ -224,11 +320,14 @@ const externalUniversalMcpTemplate = `# Universal MCP — standalone register +
 # Upgrade the interpreter (brew install python@3.12 / apt install
 # python3.12 / etc.) or use a 3.11+ venv.

-# 1. Install the workspace runtime wheel:
+# 1. Install the workspace runtime wheel (once per machine — safe to
+#    re-run; subsequent workspaces share the same wheel):
 pip install molecule-ai-workspace-runtime

 # 2. Wire molecule-mcp into your agent's MCP config. Claude Code:
-claude mcp add molecule -s user -- env \
+#    NOTE the server name is workspace-specific ("{{MCP_SERVER_NAME}}") so
+#    multiple molecule workspaces co-exist in one Claude Code session.
+claude mcp add {{MCP_SERVER_NAME}} -s user -- env \
  WORKSPACE_ID={{WORKSPACE_ID}} \
  PLATFORM_URL={{PLATFORM_URL}} \
  MOLECULE_WORKSPACE_TOKEN="<paste from create response>" \
@@ -249,8 +348,11 @@ claude mcp add molecule -s user -- env \
 #   Documentation: https://doc.moleculesai.app/docs/guides/mcp-server-setup
 #   Common errors:
 #     • "Tools not appearing in your agent" — run ` + "`claude mcp list`" + ` (or
-#       your runtime's equivalent) and confirm the molecule entry. If
-#       missing, re-run the ` + "`claude mcp add`" + ` line above.
+#       your runtime's equivalent) and confirm the {{MCP_SERVER_NAME}} entry.
+#       If missing, re-run the ` + "`claude mcp add`" + ` line above.
+#     • "Connecting a second workspace overwrote the first" — re-check that
+#       the server name in the line above is {{MCP_SERVER_NAME}} (not a bare
+#       "molecule"); each workspace's modal generates a distinct name.
 #     • "ConnectionRefused / DNS error on first call" — PLATFORM_URL must
 #       include the scheme (https://) and have NO trailing slash. Verify
 #       with: curl ${PLATFORM_URL}/healthz
@@ -331,6 +433,13 @@ const externalHermesChannelTemplate = `# Hermes channel — bridges this workspa
 # hermes-agent session. No tunnel/public URL needed (long-poll based,
 # same shape as the Claude Code channel).
 #
+# Multi-workspace: each workspace's plugin_platforms entry is keyed by a
+# workspace-specific slug ("{{MCP_SERVER_NAME}}") so two molecule
+# workspaces can coexist in one hermes config — YAML rejects duplicate
+# mapping keys, so re-using the same "molecule:" key for a second
+# workspace would silently overwrite the first. Re-running this snippet
+# for another workspace ADDS a sibling entry instead.
+#
 # Prereq: a hermes-agent install on the target machine. Latest builds
 # (post #17751) ship the platform-plugin API natively; older ones are
 # also supported via the plugin's dual-mode fallback.
@@ -345,13 +454,17 @@ export MOLECULE_PLATFORM_URL={{PLATFORM_URL}}
 export MOLECULE_WORKSPACE_TOKEN="<paste from create response>"

 # 3. Edit ~/.hermes/config.yaml — under your existing top-level
-#    gateway: block, add a plugin_platforms entry:
+#    gateway: block, add a plugin_platforms entry. The platform key
+#    ({{MCP_SERVER_NAME}}) is workspace-specific so multiple molecule
+#    workspaces coexist; re-using the same key for a second workspace
+#    would silently overwrite the first (YAML duplicate-key collapse):
 #
 #      gateway:
 #        # ...your existing gateway settings...
 #        plugin_platforms:
-#          molecule:
+#          {{MCP_SERVER_NAME}}:
 #            enabled: true
+#            workspace_id: {{WORKSPACE_ID}}
 #
 #    If you don't yet have a gateway: block, create one with just
 #    that plugin_platforms entry. Don't append blindly — YAML
@@ -404,6 +517,14 @@ hermes gateway --replace
 const externalCodexTemplate = `# Codex external setup — outbound tools (MCP) + inbound push (bridge).
 # For operators whose external agent is a codex CLI (@openai/codex)
 # session.
+#
+# Multi-workspace: the TOML table name is workspace-specific
+# ("{{MCP_SERVER_NAME}}") so two molecule workspaces can coexist in one
+# ~/.codex/config.toml — TOML rejects duplicate
+# [mcp_servers.<name>] tables, so re-using a bare "molecule" name for a
+# second workspace would either break codex parsing or silently
+# overwrite the first. Re-running this snippet for another workspace
+# ADDS a sibling table instead.

 # 1. Install codex CLI, the workspace runtime, and the bridge daemon:
 npm install -g @openai/codex@latest
@@ -412,23 +533,21 @@ pip install codex-channel-molecule

 # 2. Wire the molecule MCP server into codex's config.toml — this is
 #    the OUTBOUND path (codex calls list_peers / delegate_task /
-#    send_message_to_user / commit_memory).
-#
-#    Don't append blindly — TOML rejects duplicate
-#    [mcp_servers.molecule] tables, so re-running on an existing
-#    config will break codex parsing. If [mcp_servers.molecule]
-#    already exists (e.g. you set this up before), replace the
-#    existing block instead of appending.
+#    send_message_to_user / commit_memory). The table name
+#    ({{MCP_SERVER_NAME}}) is workspace-specific; re-running the
+#    snippet for a DIFFERENT workspace appends a sibling table without
+#    touching the first. Re-running for the SAME workspace produces
+#    the same name, so replace the existing block instead of appending.

 mkdir -p ~/.codex
 # (then open ~/.codex/config.toml in your editor and paste:)
 #
-# [mcp_servers.molecule]
+# [mcp_servers.{{MCP_SERVER_NAME}}]
 # command = "molecule-mcp"
 # args = []
 # startup_timeout_sec = 30
 #
-# [mcp_servers.molecule.env]
+# [mcp_servers.{{MCP_SERVER_NAME}}.env]
 # WORKSPACE_ID = "{{WORKSPACE_ID}}"
 # PLATFORM_URL = "{{PLATFORM_URL}}"
 # MOLECULE_WORKSPACE_TOKEN = "<paste from create response>"
@@ -472,11 +591,13 @@ codex
 # Need help?
 #   Documentation: https://doc.moleculesai.app/docs/guides/mcp-server-setup
 #   Common errors:
-#     • [mcp_servers.molecule] not loaded — codex must be ≥ 0.57.
+#     • [mcp_servers.{{MCP_SERVER_NAME}}] not loaded — codex must be ≥ 0.57.
 #       Check with ` + "`codex --version`" + `; upgrade via npm install -g @openai/codex@latest.
-#     • TOML parse error after re-running setup — TOML rejects duplicate
-#       [mcp_servers.molecule] tables. Open ~/.codex/config.toml and
-#       remove the old block before pasting the new one.
+#     • TOML parse error after re-running setup for the SAME workspace —
+#       TOML rejects duplicate [mcp_servers.<name>] tables. Open
+#       ~/.codex/config.toml and remove the old block before pasting the
+#       new one. (A second molecule workspace gets a DIFFERENT table
+#       name, so coexisting workspaces don't conflict.)
 #     • Canvas messages don't wake codex — step 3 (codex-channel-molecule
 #       bridge daemon) is required for inbound push. Check
 #       pgrep -f codex-channel-molecule and tail ~/.codex-channel-molecule/daemon.log.
@@ -502,23 +623,23 @@ const externalKimiTemplate = `# Kimi CLI external setup — register + heartbeat
 pip install molecule-ai-workspace-runtime

 # 2. Save credentials and the bridge script:
-mkdir -p ~/.molecule-ai/kimi-workspace
-chmod 700 ~/.molecule-ai/kimi-workspace
-cat > ~/.molecule-ai/kimi-workspace/env <<'EOF'
+mkdir -p ~/.molecule-ai/kimi-{{MCP_SERVER_NAME}}
+chmod 700 ~/.molecule-ai/kimi-{{MCP_SERVER_NAME}}
+cat > ~/.molecule-ai/kimi-{{MCP_SERVER_NAME}}/env <<'EOF'
 WORKSPACE_ID={{WORKSPACE_ID}}
 PLATFORM_URL={{PLATFORM_URL}}
 MOLECULE_WORKSPACE_TOKEN=<paste from create response>
 EOF
-chmod 600 ~/.molecule-ai/kimi-workspace/env
+chmod 600 ~/.molecule-ai/kimi-{{MCP_SERVER_NAME}}/env

-cat > ~/.molecule-ai/kimi-workspace/kimi_bridge.py <<'PYEOF'
+cat > ~/.molecule-ai/kimi-{{MCP_SERVER_NAME}}/kimi_bridge.py <<'PYEOF'
 #!/usr/bin/env python3
 """Kimi bridge — keeps workspace online and polls for canvas messages."""
 import json, logging, time
 from pathlib import Path
 import httpx

-ENV = Path.home() / ".molecule-ai" / "kimi-workspace" / "env"
+ENV = Path.home() / ".molecule-ai" / "kimi-{{MCP_SERVER_NAME}}" / "env"
 HEARTBEAT_INTERVAL = 20
 POLL_INTERVAL = 5

@@ -608,10 +729,10 @@ def main():
 if __name__ == "__main__":
    main()
 PYEOF
-chmod +x ~/.molecule-ai/kimi-workspace/kimi_bridge.py
+chmod +x ~/.molecule-ai/kimi-{{MCP_SERVER_NAME}}/kimi_bridge.py

 # 3. Start the bridge (run in a persistent terminal or via launchd):
-python3 ~/.molecule-ai/kimi-workspace/kimi_bridge.py
+python3 ~/.molecule-ai/kimi-{{MCP_SERVER_NAME}}/kimi_bridge.py

 # What the script does:
 #   • Registers the workspace in poll mode (no public URL needed)
@@ -622,7 +743,7 @@ python3 ~/.molecule-ai/kimi-workspace/kimi_bridge.py
 # To change the reply logic, edit the send_reply() call inside the loop.
 # To send a one-off reply from another terminal:
 #   curl -fsS -X POST "{{PLATFORM_URL}}/workspaces/{{WORKSPACE_ID}}/notify" \
-#     -H "Authorization: Bearer $(cat ~/.molecule-ai/kimi-workspace/env | grep TOKEN | cut -d= -f2)" \
+#     -H "Authorization: Bearer $(cat ~/.molecule-ai/kimi-{{MCP_SERVER_NAME}}/env | grep TOKEN | cut -d= -f2)" \
 #     -H "Content-Type: application/json" \
 #     -d '{"message":"Hello from Kimi"}'
 #
@@ -644,6 +765,13 @@ const externalOpenClawTemplate = `# OpenClaw MCP config — outbound tool path.
 # sessions.steer push path; an external setup would need the same
 # bridge daemon the template uses. For inbound delivery on an
 # external machine today, pair with the Python SDK tab.
+#
+# Multi-workspace: each workspace registers under a workspace-specific
+# MCP server name ("{{MCP_SERVER_NAME}}"). openclaw keys MCP servers by
+# name in its config (~/.openclaw/mcp/<name>.json), so re-running with
+# a bare "molecule" name would overwrite the prior workspace's entry.
+# Re-run this snippet for another workspace to ADD a sibling entry
+# instead.

 # 1. Install openclaw CLI + the workspace runtime wheel:
 #    The version pin (>=0.1.999) ensures the "molecule-mcp" console
@@ -674,7 +802,7 @@ pip install "molecule-ai-workspace-runtime>=0.1.999"
 # workspace as awaiting_agent (OFFLINE) within 60-90s even while
 # tools work.
 WORKSPACE_TOKEN="<paste from create response>"
-openclaw mcp set molecule "$(cat <<EOF
+openclaw mcp set {{MCP_SERVER_NAME}} "$(cat <<EOF
 {
  "command": "molecule-mcp",
  "args": [],
@@ -704,6 +832,6 @@ openclaw agent --message "list my peers"
 #     • Gateway not starting — tail ~/.openclaw/gateway.log. The loopback
 #       bind requires :18789 to be free; check with ` + "`lsof -iTCP:18789`" + `.
 #     • ` + "`openclaw mcp set`" + ` rejected — the heredoc generates JSON;
-#       verify with ` + "`jq < ~/.openclaw/mcp/molecule.json`" + ` and re-run
+#       verify with ` + "`jq < ~/.openclaw/mcp/{{MCP_SERVER_NAME}}.json`" + ` and re-run
 #       ` + "`openclaw mcp set`" + ` if the file is malformed.
 `
@@ -52,7 +52,7 @@ func (h *WorkspaceHandler) RotateExternalCredentials(c *gin.Context) {
 	}
 	ctx := c.Request.Context()

-	runtime, err := lookupWorkspaceRuntime(ctx, db.DB, id)
+	runtime, name, err := lookupWorkspaceRuntimeAndName(ctx, db.DB, id)
 	if errors.Is(err, sql.ErrNoRows) {
 		c.JSON(http.StatusNotFound, gin.H{"error": "workspace not found"})
 		return
@@ -108,7 +108,7 @@ func (h *WorkspaceHandler) RotateExternalCredentials(c *gin.Context) {

 	platformURL := externalPlatformURL(c)
 	c.JSON(http.StatusOK, gin.H{
-		"connection": BuildExternalConnectionPayload(platformURL, id, tok),
+		"connection": BuildExternalConnectionPayload(platformURL, id, name, tok),
 	})
 }

@@ -129,7 +129,7 @@ func (h *WorkspaceHandler) GetExternalConnection(c *gin.Context) {
 	}
 	ctx := c.Request.Context()

-	runtime, err := lookupWorkspaceRuntime(ctx, db.DB, id)
+	runtime, name, err := lookupWorkspaceRuntimeAndName(ctx, db.DB, id)
 	if errors.Is(err, sql.ErrNoRows) {
 		c.JSON(http.StatusNotFound, gin.H{"error": "workspace not found"})
 		return
@@ -149,16 +149,20 @@ func (h *WorkspaceHandler) GetExternalConnection(c *gin.Context) {

 	platformURL := externalPlatformURL(c)
 	c.JSON(http.StatusOK, gin.H{
-		"connection": BuildExternalConnectionPayload(platformURL, id, ""),
+		"connection": BuildExternalConnectionPayload(platformURL, id, name, ""),
 	})
 }

-// lookupWorkspaceRuntime returns the workspace's runtime field. Wrapped
-// for readability + so tests can mock the single SELECT.
-func lookupWorkspaceRuntime(ctx context.Context, handle *sql.DB, id string) (string, error) {
-	var runtime string
-	err := handle.QueryRowContext(ctx, `
-		SELECT COALESCE(runtime, '') FROM workspaces WHERE id = $1
-	`, id).Scan(&runtime)
-	return runtime, err
+// lookupWorkspaceRuntimeAndName returns runtime + name in one round-trip.
+// Wrapped for readability + so tests can mock the single SELECT.
+// Used by rotate / re-show paths: runtime gates the external-only check;
+// name feeds the per-workspace MCP server slug in BuildExternalConnectionPayload
+// (so the Universal MCP snippet uses a stable per-workspace name instead
+// of overwriting prior `claude mcp add molecule` entries).
+// Returns sql.ErrNoRows when the workspace doesn't exist.
+func lookupWorkspaceRuntimeAndName(ctx context.Context, handle *sql.DB, id string) (runtime, name string, err error) {
+	err = handle.QueryRowContext(ctx, `
+		SELECT COALESCE(runtime, ''), COALESCE(name, '') FROM workspaces WHERE id = $1
+	`, id).Scan(&runtime, &name)
+	return runtime, name, err
 }
@@ -35,9 +35,9 @@ func TestRotateExternalCredentials_HappyPath(t *testing.T) {
 	wh := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())

 	// 1. Runtime lookup
-	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\) FROM workspaces WHERE id = \$1`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\), COALESCE\(name, ''\) FROM workspaces WHERE id = \$1`).
 		WithArgs("ws-ext").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("external"))
+		WillReturnRows(sqlmock.NewRows([]string{"runtime", "name"}).AddRow("external", "test-ws"))

 	// 2. Revoke all live tokens
 	mock.ExpectExec(`UPDATE workspace_auth_tokens`).
@@ -98,9 +98,9 @@ func TestRotateExternalCredentials_RejectsNonExternal(t *testing.T) {
 	setupTestRedis(t)
 	wh := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())

-	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\) FROM workspaces WHERE id = \$1`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\), COALESCE\(name, ''\) FROM workspaces WHERE id = \$1`).
 		WithArgs("ws-hermes").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("hermes"))
+		WillReturnRows(sqlmock.NewRows([]string{"runtime", "name"}).AddRow("hermes", "test-ws"))

 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)
@@ -129,9 +129,9 @@ func TestRotateExternalCredentials_NotFound(t *testing.T) {
 	setupTestRedis(t)
 	wh := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())

-	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\) FROM workspaces WHERE id = \$1`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\), COALESCE\(name, ''\) FROM workspaces WHERE id = \$1`).
 		WithArgs("ws-missing").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"})) // no rows
+		WillReturnRows(sqlmock.NewRows([]string{"runtime", "name"})) // no rows

 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)
@@ -172,9 +172,9 @@ func TestGetExternalConnection_HappyPathReturnsBlankToken(t *testing.T) {
 	setupTestRedis(t)
 	wh := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())

-	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\) FROM workspaces WHERE id = \$1`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\), COALESCE\(name, ''\) FROM workspaces WHERE id = \$1`).
 		WithArgs("ws-ext").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("external"))
+		WillReturnRows(sqlmock.NewRows([]string{"runtime", "name"}).AddRow("external", "test-ws"))

 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)
@@ -211,9 +211,9 @@ func TestGetExternalConnection_RejectsNonExternal(t *testing.T) {
 	setupTestRedis(t)
 	wh := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())

-	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\) FROM workspaces WHERE id = \$1`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\), COALESCE\(name, ''\) FROM workspaces WHERE id = \$1`).
 		WithArgs("ws-claude").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("claude-code"))
+		WillReturnRows(sqlmock.NewRows([]string{"runtime", "name"}).AddRow("claude-code", "test-ws"))

 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)
@@ -233,9 +233,9 @@ func TestGetExternalConnection_NotFound(t *testing.T) {
 	setupTestRedis(t)
 	wh := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())

-	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\) FROM workspaces WHERE id = \$1`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, ''\), COALESCE\(name, ''\) FROM workspaces WHERE id = \$1`).
 		WithArgs("ws-missing").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"}))
+		WillReturnRows(sqlmock.NewRows([]string{"runtime", "name"}))

 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)
@@ -253,7 +253,7 @@ func TestGetExternalConnection_NotFound(t *testing.T) {
 // ---------- BuildExternalConnectionPayload (pure helper) ----------

 func TestBuildExternalConnectionPayload_StampsPlaceholders(t *testing.T) {
-	got := BuildExternalConnectionPayload("https://platform.test", "ws-7", "tok-abc")
+	got := BuildExternalConnectionPayload("https://platform.test", "ws-7", "my-bot", "tok-abc")

 	if got["workspace_id"] != "ws-7" {
 		t.Errorf("workspace_id: %v", got["workspace_id"])
@@ -267,6 +267,18 @@ func TestBuildExternalConnectionPayload_StampsPlaceholders(t *testing.T) {
 	if got["registry_endpoint"] != "https://platform.test/registry/register" {
 		t.Errorf("registry_endpoint: %v", got["registry_endpoint"])
 	}
+	// Universal MCP snippet must contain a workspace-specific server
+	// name derived from the workspace name. Without this each new
+	// `claude mcp add` would overwrite the previous entry in the user's
+	// ~/.claude.json (servers are keyed by name) — collapsing
+	// multi-workspace use into one slot. See mcpServerNameForWorkspace.
+	mcp, _ := got["universal_mcp_snippet"].(string)
+	if !strings.Contains(mcp, "claude mcp add molecule-my-bot ") {
+		t.Errorf("universal_mcp_snippet missing per-workspace server name 'molecule-my-bot':\n%s", mcp)
+	}
+	if strings.Contains(mcp, "{{MCP_SERVER_NAME}}") {
+		t.Errorf("universal_mcp_snippet still contains literal {{MCP_SERVER_NAME}}")
+	}
 	// {{PLATFORM_URL}} + {{WORKSPACE_ID}} placeholders must be substituted
 	// out of every snippet — if any snippet still contains a literal
 	// "{{PLATFORM_URL}}" or "{{WORKSPACE_ID}}", a future template author
@@ -292,7 +304,7 @@ func TestBuildExternalConnectionPayload_TrimsTrailingSlash(t *testing.T) {
 	// being concatenated into endpoint paths — otherwise the operator
 	// gets `https://platform.test//registry/register` (double slash) which
 	// some servers reject as a redirect target.
-	got := BuildExternalConnectionPayload("https://platform.test/", "ws-7", "")
+	got := BuildExternalConnectionPayload("https://platform.test/", "ws-7", "", "")
 	if got["platform_url"] != "https://platform.test" {
 		t.Errorf("platform_url: trailing slash not trimmed; got %v", got["platform_url"])
 	}
@@ -304,8 +316,100 @@ func TestBuildExternalConnectionPayload_TrimsTrailingSlash(t *testing.T) {
 func TestBuildExternalConnectionPayload_BlankAuthTokenIsAllowed(t *testing.T) {
 	// Re-show path: auth_token="" is the contract; the modal masks the
 	// field and labels it "rotate to reveal a new token".
-	got := BuildExternalConnectionPayload("https://platform.test", "ws-7", "")
+	got := BuildExternalConnectionPayload("https://platform.test", "ws-7", "", "")
 	if got["auth_token"] != "" {
 		t.Errorf("blank token must propagate as \"\"; got %v", got["auth_token"])
 	}
 }
+
+// TestBuildExternalConnectionPayload_McpServerNameUniquePerWorkspace
+// pins the multi-workspace install contract: two distinct workspaces
+// must produce two distinct `claude mcp add` server-name lines, or
+// installing the second one will overwrite the first entry in the
+// user's ~/.claude.json (servers are keyed by name) — collapsing
+// multi-workspace use into a single per-session slot, which is the
+// "this is per-session" UX the CTO observed 2026-05-18.
+func TestBuildExternalConnectionPayload_McpServerNameUniquePerWorkspace(t *testing.T) {
+	cases := []struct {
+		name        string
+		workspaceID string
+		wsName      string
+		wantAddLine string // must appear in universal_mcp_snippet
+	}{
+		{"plain name", "id-a", "my-bot", "claude mcp add molecule-my-bot "},
+		{"name with spaces + caps", "id-b", "My Bot 1", "claude mcp add molecule-my-bot-1 "},
+		// Symbol/punctuation collapses to single hyphens and trims.
+		{"name with symbols", "id-c", "--Foo!!Bar--", "claude mcp add molecule-foo-bar "},
+		// Empty name falls back to the first 8 chars of the (de-hyphenated)
+		// workspace UUID — keeps the snippet unique per workspace even
+		// when callers (rotate/re-show pre-name-lookup) pass "".
+		{"empty name, uuid id", "12345678-aaaa-bbbb-cccc-deadbeef0000", "", "claude mcp add molecule-12345678 "},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			got := BuildExternalConnectionPayload("https://p.test", tc.workspaceID, tc.wsName, "tok")
+			mcp, _ := got["universal_mcp_snippet"].(string)
+			if !strings.Contains(mcp, tc.wantAddLine) {
+				t.Errorf("missing %q in universal_mcp_snippet:\n%s", tc.wantAddLine, mcp)
+			}
+			// Belt + suspenders: never the bare fixed `molecule` name —
+			// that was the bug. (Match with trailing space so the
+			// "molecule-…" form passes.)
+			if strings.Contains(mcp, "claude mcp add molecule ") {
+				t.Errorf("snippet regressed to fixed `claude mcp add molecule `; got:\n%s", mcp)
+			}
+		})
+	}
+}
+
+// TestBuildExternalConnectionPayload_AllRuntimeSnippetsAreWorkspaceUnique
+// extends the multi-workspace install contract to every runtime tab in
+// the modal. Each MCP-host config keyspace has the SAME equivalence
+// class as Claude Code's `claude mcp add <name>`:
+//
+//   - codex: ~/.codex/config.toml [mcp_servers.<name>] — TOML rejects
+//     duplicate table keys, so a second workspace with the same name
+//     either breaks parsing or overwrites the first table.
+//   - openclaw: ~/.openclaw/mcp/<name>.json — file is keyed by <name>,
+//     `openclaw mcp set <same-name>` overwrites.
+//   - hermes: ~/.hermes/config.yaml gateway.plugin_platforms.<key>:
+//     YAML rejects duplicate mapping keys.
+//   - kimi: ~/.molecule-ai/kimi-<slug>/ per-workspace dir — single
+//     "kimi-workspace" dir would have both workspaces' envs collide.
+//
+// All four must therefore stamp the workspace-specific
+// {{MCP_SERVER_NAME}} slug. This test catches a future template author
+// who introduces a new runtime tab without plumbing the slug.
+func TestBuildExternalConnectionPayload_AllRuntimeSnippetsAreWorkspaceUnique(t *testing.T) {
+	got := BuildExternalConnectionPayload("https://p.test", "id-a", "my-bot", "tok")
+
+	// Per-template literal that proves the slug was stamped through.
+	wantPerSnippet := map[string]string{
+		"universal_mcp_snippet": "claude mcp add molecule-my-bot ",
+		"codex_snippet":         "[mcp_servers.molecule-my-bot]",
+		"openclaw_snippet":      "openclaw mcp set molecule-my-bot ",
+		"hermes_channel_snippet": "          molecule-my-bot:",
+		"kimi_snippet":          "~/.molecule-ai/kimi-molecule-my-bot",
+	}
+	for key, needle := range wantPerSnippet {
+		v, _ := got[key].(string)
+		if !strings.Contains(v, needle) {
+			t.Errorf("%s missing per-workspace slug literal %q:\n%s", key, needle, v)
+		}
+	}
+
+	// No template should still contain the unstamped placeholder — that
+	// would mean BuildExternalConnectionPayload's stamp() didn't sweep
+	// it, which is the regression we're guarding against.
+	for _, k := range []string{
+		"curl_register_template", "python_snippet",
+		"claude_code_channel_snippet", "universal_mcp_snippet",
+		"hermes_channel_snippet", "codex_snippet", "openclaw_snippet",
+		"kimi_snippet",
+	} {
+		v, _ := got[k].(string)
+		if strings.Contains(v, "{{MCP_SERVER_NAME}}") {
+			t.Errorf("%s still contains literal {{MCP_SERVER_NAME}}", k)
+		}
+	}
+}
@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"net/http"
 	"net/http/httptest"
+	"sync"
 	"testing"
 	"time"

@@ -22,8 +23,39 @@ import (
 	"github.com/redis/go-redis/v9"
 )

+// liveTestHandlers tracks every WorkspaceHandler built during the test
+// binary's lifetime so setupTestDB can drain their in-flight goAsync
+// goroutines (notably the detached RestartByID restart cycle, which
+// reads the global db.DB) BEFORE restoring db.DB. Without this drain a
+// fire-and-forget restart goroutine spawned by one test outlives that
+// test and races the db.DB swap in a later test's t.Cleanup — the
+// 0x...d548 data race on platform/internal/db.DB.
+var (
+	liveTestHandlersMu sync.Mutex
+	liveTestHandlers   []*WorkspaceHandler
+)
+
 func init() {
 	gin.SetMode(gin.TestMode)
+	newHandlerHook = func(h *WorkspaceHandler) {
+		liveTestHandlersMu.Lock()
+		liveTestHandlers = append(liveTestHandlers, h)
+		liveTestHandlersMu.Unlock()
+	}
+}
+
+// drainTestAsync waits for every tracked handler's goAsync goroutines to
+// finish. Called from setupTestDB's cleanup before db.DB is restored so
+// no detached restart/provision goroutine is mid-read of db.DB when the
+// pointer is swapped.
+func drainTestAsync() {
+	liveTestHandlersMu.Lock()
+	handlers := make([]*WorkspaceHandler, len(liveTestHandlers))
+	copy(handlers, liveTestHandlers)
+	liveTestHandlersMu.Unlock()
+	for _, h := range handlers {
+		h.waitAsyncForTest()
+	}
 }

 // setupTestDB creates a sqlmock DB and assigns it to the global db.DB.
@@ -42,7 +74,16 @@ func setupTestDB(t *testing.T) sqlmock.Sqlmock {
 	}
 	prevDB := db.DB
 	db.DB = mockDB
-	t.Cleanup(func() { db.DB = prevDB; mockDB.Close() })
+	t.Cleanup(func() {
+		// Drain detached async goroutines (e.g. goAsync(RestartByID),
+		// which reads db.DB in runRestartCycle before its provisioner
+		// gate) BEFORE swapping db.DB back. Doing the restore first
+		// would let an in-flight restart goroutine read db.DB while
+		// this line writes it — the data race this guards against.
+		drainTestAsync()
+		db.DB = prevDB
+		mockDB.Close()
+	})

 	// Disable SSRF checks for the duration of this test only. Restore
 	// the previous state via t.Cleanup so that TestIsSafeURL_* tests
--- a/Show More
+++ b/Show More