Merge pull request 'fix(canvas): extend mc#1535 per-workspace MCP slug to codex/openclaw/hermes/kimi (multi-workspace class sweep)' (#1536) from fix/multi-workspace-install-snippets-class-sweep into main

Prior CI failures on this PR were infra-class (Detect changes hit
'Error: ENOSPC: no space left on device' from runner disk-full caused
by 120 zombie tasks since drained; Python Lint flaked on perf test
test_batch_fetcher_runs_submitted_rows_concurrently by 3ms under
contended runners — same test passes cleanly on main HEAD 1b0e947).
Re-firing CI on recovered runners; no code change. [no-op]

.gitea/scripts/review-check.sh

@@ -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
 

.gitea/scripts/tests/_review_check_fixture.py

@@ -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):

.gitea/scripts/tests/test_review_check.sh

@@ -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"

.gitea/workflows/ci-mcp-stdio-transport.yml

@@ -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

.gitea/workflows/ci.yml

@@ -145,10 +145,10 @@ jobs:
     # the diagnostic step with its own continue-on-error: true (line 203).
     # Flip confirmed by CI / Platform (Go) status = success on main HEAD 363905d3.
     continue-on-error: false
-    # Job-level ceiling. The go test step below runs with a per-step 30m timeout;
-    # this cap catches any step that leaks past that. Set well above 30m so
+    # Job-level ceiling. The go test step below runs with a per-step 10m timeout;
+    # this cap catches any step that leaks past that. Set well above 10m so
     # the per-step timeout is the active constraint.
-    timeout-minutes: 35
+    timeout-minutes: 15
     defaults:
       run:
         working-directory: workspace-server
@@ -176,14 +176,12 @@ jobs:
         name: Run golangci-lint
         run: $(go env GOPATH)/bin/golangci-lint run --timeout 3m ./...
       - if: always()
-        name: Diagnostic — per-package verbose (300s timeout)
+        name: Diagnostic — per-package verbose 60s
         run: |
           set +e
-          # 300s allows handlers + pendinguploads packages to complete on cold
-          # runners with -race instrumentation (~60-120s each vs ~14s non-race).
-          go test -race -v -timeout 300s ./internal/handlers/... 2>&1 | tee /tmp/test-handlers.log
+          go test -race -v -timeout 60s ./internal/handlers/... 2>&1 | tee /tmp/test-handlers.log
           handlers_exit=$?
-          go test -race -v -timeout 300s ./internal/pendinguploads/... 2>&1 | tee /tmp/test-pu.log
+          go test -race -v -timeout 60s ./internal/pendinguploads/... 2>&1 | tee /tmp/test-pu.log
           pu_exit=$?
           echo "::group::handlers exit=$handlers_exit (last 100 lines)"
           tail -100 /tmp/test-handlers.log
@@ -196,10 +194,10 @@ jobs:
       - if: always()
         name: Run tests with race detection and coverage
         # Explicit timeout: cold runner cache causes OOM kills at ~4m39s on the
-        # full ./... suite with race detection + coverage. A 30m per-step timeout
-        # lets the suite complete on cold cache (~13-25m) while failing cleanly
-        # instead of OOM-killing. The job-level timeout (35m) is a backstop.
-        run: go test -race -timeout 30m -coverprofile=coverage.out ./...
+        # full ./... suite with race detection + coverage. A 10m per-step timeout
+        # lets the suite complete on cold cache (~5-7m) while failing cleanly
+        # instead of OOM-killing. The job-level timeout (15m) is a backstop.
+        run: go test -race -timeout 10m -coverprofile=coverage.out ./...
 
       - if: always()
         name: Per-file coverage report
@@ -540,11 +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

.gitea/workflows/gitea-merge-queue.yml

@@ -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

.gitea/workflows/publish-runtime-autobump.yml

@@ -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

.gitea/workflows/qa-review.yml

@@ -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.

.gitea/workflows/secret-scan.yml

@@ -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

.gitea/workflows/security-review.yml

@@ -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.

.gitea/workflows/sop-checklist.yml

@@ -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:

.gitea/workflows/sop-tier-check.yml

@@ -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

canvas/src/app/page.tsx

@@ -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>

canvas/src/components/AuditTrailPanel.tsx

@@ -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>
     );

canvas/src/components/ConversationTraceModal.tsx

@@ -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>
               )}

canvas/src/components/EmptyState.tsx

@@ -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>

canvas/src/components/ExternalConnectModal.tsx

@@ -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">

canvas/src/components/MissingKeysModal.tsx

@@ -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>
           )}

canvas/src/components/WorkspaceUsage.tsx

@@ -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 ? (

canvas/src/components/__tests__/AuditTrailPanel.format.test.ts

@@ -1,55 +0,0 @@
-// @vitest-environment jsdom
-/**
- * Tests for formatAuditRelativeTime exported from AuditTrailPanel.
- */
-import { describe, it, expect } from "vitest";
-import { formatAuditRelativeTime } from "../AuditTrailPanel";
-
-describe("formatAuditRelativeTime", () => {
-  const now = new Date("2026-05-18T12:00:00Z").getTime();
-
-  it('returns "just now" for timestamps less than 60s ago', () => {
-    const ts = new Date(now - 30_000).toISOString(); // 30s ago
-    expect(formatAuditRelativeTime(ts, now)).toBe("just now");
-  });
-
-  it("returns minutes for timestamps under 1h", () => {
-    const ts = new Date(now - 5 * 60_000).toISOString(); // 5m ago
-    expect(formatAuditRelativeTime(ts, now)).toBe("5m ago");
-  });
-
-  it("returns hours for timestamps under 24h", () => {
-    const ts = new Date(now - 3 * 3_600_000).toISOString(); // 3h ago
-    expect(formatAuditRelativeTime(ts, now)).toBe("3h ago");
-  });
-
-  it("returns locale date for timestamps older than 24h", () => {
-    const ts = new Date(now - 2 * 86_400_000).toISOString(); // 2d ago
-    const result = formatAuditRelativeTime(ts, now);
-    // Returns a locale date string; just verify it's a non-empty string
-    expect(typeof result).toBe("string");
-    expect(result.length).toBeGreaterThan(0);
-    expect(result).not.toBe("just now");
-    expect(result).not.toMatch(/m ago$/);
-    expect(result).not.toMatch(/h ago$/);
-  });
-
-  it("handles exactly 60s boundary as minutes", () => {
-    const ts = new Date(now - 60_000).toISOString(); // exactly 1m ago
-    expect(formatAuditRelativeTime(ts, now)).toBe("1m ago");
-  });
-
-  it("handles exactly 3600s boundary as hours", () => {
-    const ts = new Date(now - 3_600_000).toISOString(); // exactly 1h ago
-    expect(formatAuditRelativeTime(ts, now)).toBe("1h ago");
-  });
-
-  it("handles exactly 86400s boundary", () => {
-    const ts = new Date(now - 86_400_000).toISOString(); // exactly 24h ago
-    const result = formatAuditRelativeTime(ts, now);
-    // Exactly 24h should fall into the "days" branch
-    expect(typeof result).toBe("string");
-    expect(result).not.toMatch(/m ago$/);
-    expect(result).not.toMatch(/h ago$/);
-  });
-});

canvas/src/components/__tests__/ExternalConnectModal.test.tsx

@@ -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", () => {

canvas/src/components/__tests__/MemoryInspectorPanel.helpers.test.ts

@@ -1,82 +0,0 @@
-// @vitest-environment jsdom
-/**
- * Tests for exported helpers from MemoryInspectorPanel:
- * isPluginUnavailableError, formatTTL.
- */
-import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
-import { isPluginUnavailableError, formatTTL } from "../MemoryInspectorPanel";
-
-describe("isPluginUnavailableError", () => {
-  it("returns true when error message contains MEMORY_PLUGIN_URL", () => {
-    const err = new Error("MEMORY_PLUGIN_URL is not configured");
-    expect(isPluginUnavailableError(err)).toBe(true);
-  });
-
-  it("returns false when error message does not contain MEMORY_PLUGIN_URL", () => {
-    const err = new Error("Connection refused");
-    expect(isPluginUnavailableError(err)).toBe(false);
-  });
-
-  it("returns false for non-Error values", () => {
-    expect(isPluginUnavailableError("string error")).toBe(false);
-    expect(isPluginUnavailableError(null)).toBe(false);
-    expect(isPluginUnavailableError(undefined)).toBe(false);
-    expect(isPluginUnavailableError({})).toBe(false);
-  });
-
-  it("handles Error with empty message", () => {
-    expect(isPluginUnavailableError(new Error(""))).toBe(false);
-  });
-});
-
-describe("formatTTL", () => {
-  // Freeze time at 2026-05-18T12:00:00Z for deterministic tests.
-  beforeEach(() => {
-    vi.useFakeTimers();
-    vi.setSystemTime(new Date("2026-05-18T12:00:00Z"));
-  });
-  afterEach(() => {
-    vi.useRealTimers();
-  });
-
-  it("returns empty string for null", () => {
-    expect(formatTTL(null)).toBe("");
-  });
-
-  it("returns empty string for undefined", () => {
-    expect(formatTTL(undefined)).toBe("");
-  });
-
-  it("returns empty string for empty string", () => {
-    expect(formatTTL("")).toBe("");
-  });
-
-  it("returns 'expired' for past timestamps", () => {
-    const past = new Date(Date.now() - 60_000).toISOString();
-    expect(formatTTL(past)).toBe("expired");
-  });
-
-  it("returns seconds for sub-minute future TTLs", () => {
-    const future = new Date(Date.now() + 30_000).toISOString();
-    expect(formatTTL(future)).toBe("30s");
-  });
-
-  it("returns minutes for sub-hour future TTLs", () => {
-    const future = new Date(Date.now() + 5 * 60_000).toISOString();
-    expect(formatTTL(future)).toBe("5m");
-  });
-
-  it("returns hours for sub-day future TTLs", () => {
-    const future = new Date(Date.now() + 3 * 3_600_000).toISOString();
-    expect(formatTTL(future)).toBe("3h");
-  });
-
-  it("returns days for TTLs longer than 24h", () => {
-    const future = new Date(Date.now() + 2 * 86_400_000).toISOString();
-    expect(formatTTL(future)).toBe("2d");
-  });
-
-  it("returns empty string for invalid date string", () => {
-    expect(formatTTL("not-a-date")).toBe("");
-  });
-});

canvas/src/components/mobile/MobileCanvas.tsx

@@ -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>

canvas/src/components/mobile/MobileChat.tsx

@@ -242,8 +242,6 @@ export function MobileChat({
 
   useChatSocket(agentId, {
     onAgentMessage: appendMessageDeduped,
-    // Fan-out user's own outbound message to all sessions (issue #228).
-    onUserMessage: appendMessageDeduped,
     onSendComplete: releaseSendGuards,
   });
 
@@ -358,6 +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,
@@ -404,6 +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,
@@ -434,6 +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",
@@ -477,7 +478,7 @@ export function MobileChat({
         }}
       >
         {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>
         )}
@@ -497,6 +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,
@@ -512,7 +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>
         )}
@@ -666,6 +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",
@@ -706,6 +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,
@@ -727,6 +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/
@@ -750,13 +756,11 @@ export function MobileChat({
               border: "none",
               outline: "none",
               background: "transparent",
-              // 16px floor: iOS Safari/WebKit auto-zooms the viewport on
-              // focus when a focused field's font-size is < 16px. Anything
-              // below this re-introduces the tap-to-zoom layout jump on the
-              // mobile PWA. Do NOT lower this without also adding a
-              // maximum-scale/user-scalable viewport lock — and that lock
-              // breaks pinch-to-zoom accessibility, so 16px here is the
-              // correct trade.
+              // 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,
@@ -773,12 +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

canvas/src/components/mobile/MobileComms.tsx

@@ -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>
         ) : (

canvas/src/components/mobile/MobileDetail.tsx

@@ -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,

canvas/src/components/mobile/MobileHome.tsx

@@ -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>

canvas/src/components/mobile/MobileMe.tsx

@@ -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>

canvas/src/components/mobile/MobileSpawn.tsx

@@ -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,

canvas/src/components/mobile/__tests__/MobileChat.test.tsx

@@ -264,18 +264,18 @@ describe("MobileChat — composer", () => {
     expect(sendBtn.disabled).toBe(true);
   });
 
-  // iOS Safari/WebKit auto-zooms the viewport on focus when a focused
-  // <input>/<textarea> has an effective font-size below 16px. On the
-  // mobile PWA this made the whole layout scale up the moment the user
-  // tapped into the chat box. Keeping the composer font ≥16px is the
-  // root-cause fix — it suppresses the focus-zoom WITHOUT disabling
-  // pinch-to-zoom (which a maximum-scale/user-scalable viewport hack
-  // would have done at the cost of accessibility).
-  it("composer textarea font-size is >= 16px (prevents iOS focus-zoom)", () => {
+  // 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;
-    const fontSizePx = parseFloat(textarea.style.fontSize);
-    expect(fontSizePx).toBeGreaterThanOrEqual(16);
+    expect(textarea).toBeTruthy();
+    const fs = Number.parseFloat(textarea.style.fontSize);
+    expect(Number.isFinite(fs)).toBe(true);
+    expect(fs).toBeGreaterThanOrEqual(16);
   });
 });
 

canvas/src/components/mobile/__tests__/MobileSpawn.test.tsx

@@ -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()} />);

canvas/src/components/mobile/components.tsx

@@ -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",

canvas/src/components/settings/OrgTokensTab.tsx

@@ -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>

canvas/src/components/settings/TokensTab.tsx

@@ -140,14 +140,14 @@ function WorkspaceTokensTab({ 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>
@@ -192,7 +192,7 @@ function WorkspaceTokensTab({ 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>

canvas/src/components/tabs/ActivityTab.tsx

@@ -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 && (

canvas/src/components/tabs/ChannelsTab.tsx

@@ -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>
       )}

canvas/src/components/tabs/ChatTab.tsx

@@ -143,12 +143,6 @@ function MyChatPanel({ workspaceId, data }: Props) {
         releaseSendGuards();
       }
     },
-    // Fan-out of user's own outbound message to all sessions (issue #228).
-    // Uses appendMessageDeduped so the originating session collapses its
-    // optimistic copy (same role + content within 3-second window).
-    onUserMessage: (msg) => {
-      history.setMessages((prev) => appendMessageDeduped(prev, msg));
-    },
     onActivityLog: (entry) => {
       if (!sending) return;
       setActivityLog((prev) => appendActivityLine(prev, entry));

canvas/src/components/tabs/ConfigTab.tsx

@@ -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">

canvas/src/components/tabs/DetailsTab.tsx

@@ -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>
         )}

canvas/src/components/tabs/EventsTab.tsx

@@ -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>
       )}

canvas/src/components/tabs/ExternalConnectionSection.tsx

@@ -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>
       )}

canvas/src/components/tabs/FilesTab.tsx

@@ -266,7 +266,7 @@ function PlatformOwnedFilesTab({
         // 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>
@@ -280,7 +280,7 @@ function PlatformOwnedFilesTab({
       )}
 
       {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>

canvas/src/components/tabs/ScheduleTab.tsx

@@ -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"

canvas/src/components/tabs/TracesTab.tsx

@@ -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>
       )}

canvas/src/components/tabs/__tests__/ConfigTab.abilities.test.tsx

@@ -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,
+    });
+  });
+});

canvas/src/components/tabs/chat/AgentCommsPanel.tsx

@@ -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"

canvas/src/components/tabs/chat/AttachmentViews.tsx

@@ -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>

canvas/src/components/tabs/chat/__tests__/a2aErrorHint.test.ts

@@ -64,66 +64,4 @@ describe("inferA2AErrorHint", () => {
     expect(hint).toMatch(/Claude Code SDK/);
     expect(hint).not.toMatch(/proxy timeout/);
   });
-
-  // ---- P1 #348: poll-mode timeout-class detection ----
-
-  it("routes poll-mode budget exhaustion to its specific actionable hint", () => {
-    // a2a_tools_delegation.py emits this exact shape after the 600s
-    // budget. The user must NOT be told to restart — the work is
-    // still in flight on the platform side.
-    const hint = inferA2AErrorHint(
-      "polling timeout after 600s (delegation_id=abc, last_status=processing); the platform is still working on it — call check_task_status('abc') to retrieve later",
-    );
-    expect(hint).toMatch(/Do NOT restart/);
-    expect(hint).toMatch(/check_task_status/);
-  });
-
-  it("matches the check_task_status hint clue even without the 'polling timeout' phrase", () => {
-    const hint = inferA2AErrorHint(
-      "platform busy — call check_task_status('xyz')",
-    );
-    expect(hint).toMatch(/check_task_status/);
-  });
-
-  it("poll-mode hint wins over the generic timeout bucket", () => {
-    // The string contains both "polling timeout after" and "timeout"
-    // — the more-specific poll-mode hint must win so users don't get
-    // the generic "restart" advice for a still-in-flight task.
-    const hint = inferA2AErrorHint("polling timeout after 600s ...");
-    expect(hint).toMatch(/Do NOT restart/);
-    expect(hint).not.toMatch(/restart the workspace if this repeats/);
-  });
-
-  // ---- P1 #348: codex-aware specialization ----
-
-  it("specialises the empty-detail hint for codex callees", () => {
-    // Per feedback_surface_actionable_failure_reason_to_user: opaque
-    // restart prompts are the anti-pattern. With peerKind=codex the
-    // hint explicitly de-recommends restart.
-    const hint = inferA2AErrorHint("", { peerKind: "codex" });
-    expect(hint).toMatch(/codex/);
-    expect(hint).toMatch(/check its Activity tab/i);
-    expect(hint).not.toMatch(/A workspace restart is the safe first move/);
-  });
-
-  it("specialises generic-timeout hint for codex callees", () => {
-    const hint = inferA2AErrorHint("ReadTimeout", { peerKind: "codex" });
-    expect(hint).toMatch(/codex/);
-    expect(hint).toMatch(/600s/);
-  });
-
-  it("falls back to the non-codex generic timeout hint when no peerKind given", () => {
-    const hint = inferA2AErrorHint("ReadTimeout");
-    expect(hint).toMatch(/proxy timeout/);
-    expect(hint).not.toMatch(/600s sync-proxy/);
-  });
-
-  it("preserves existing empty-detail wording when no peer context provided", () => {
-    const hint = inferA2AErrorHint("");
-    expect(hint).toMatch(/no error detail/);
-    // Updated wording: must NOT be the bare "restart is the safe
-    // first move" line — that violates surface-actionable-reason.
-    expect(hint).not.toMatch(/safe first move/);
-    expect(hint).toMatch(/Activity tab/);
-  });
 });

canvas/src/components/tabs/chat/__tests__/message-parser.test.ts

@@ -248,88 +248,6 @@ describe("extractResponseText", () => {
   });
 });
 
-describe("extractAgentText", () => {
-  it("extracts text from top-level parts", () => {
-    const task = {
-      parts: [{ kind: "text", text: "Agent said hello" }],
-    };
-    expect(extractAgentText(task)).toBe("Agent said hello");
-  });
-
-  it("extracts from artifacts[0].parts when top-level parts absent", () => {
-    const task = {
-      artifacts: [
-        { parts: [{ kind: "text", text: "From artifact block" }] },
-      ],
-    };
-    expect(extractAgentText(task)).toBe("From artifact block");
-  });
-
-  it("extracts from status.message.parts as fallback", () => {
-    const task = {
-      status: {
-        message: { parts: [{ kind: "text", text: "Status text" }] },
-      },
-    };
-    expect(extractAgentText(task)).toBe("Status text");
-  });
-
-  it("prefers top-level parts over artifacts", () => {
-    const task = {
-      parts: [{ kind: "text", text: "top-level wins" }],
-      artifacts: [
-        { parts: [{ kind: "text", text: "artifact text" }] },
-      ],
-    };
-    expect(extractAgentText(task)).toBe("top-level wins");
-  });
-
-  it("prefers top-level parts over status.message", () => {
-    const task = {
-      parts: [{ kind: "text", text: "parts wins" }],
-      status: {
-        message: { parts: [{ kind: "text", text: "status text" }] },
-      },
-    };
-    expect(extractAgentText(task)).toBe("parts wins");
-  });
-
-  it("returns string identity when task itself is a string", () => {
-    expect(extractAgentText("plain string task" as unknown as Record<string, unknown>)).toBe(
-      "plain string task",
-    );
-  });
-
-  it("returns fallback when task is an empty object", () => {
-    expect(extractAgentText({})).toBe("(Could not extract response text)");
-  });
-
-  it("returns fallback when task has no extractable text", () => {
-    expect(
-      extractAgentText({ status: "running", other: "fields" }),
-    ).toBe("(Could not extract response text)");
-  });
-
-  it("tolerates malformed nested shapes without throwing", () => {
-    const task = {
-      parts: null,
-      artifacts: "not an array",
-      status: { message: 42 },
-    };
-    expect(extractAgentText(task)).toBe("(Could not extract response text)");
-  });
-
-  it("joins multiple text parts with newline", () => {
-    const task = {
-      parts: [
-        { kind: "text", text: "Line one" },
-        { kind: "text", text: "Line two" },
-      ],
-    };
-    expect(extractAgentText(task)).toBe("Line one\nLine two");
-  });
-});
-
 describe("extractTextsFromParts", () => {
   it("extracts text parts with kind=text", () => {
     const parts = [

canvas/src/components/tabs/chat/__tests__/resolveWorkspaceName.test.ts

@@ -1,102 +0,0 @@
-import { describe, it, expect, beforeEach } from "vitest";
-import { useCanvasStore } from "@/store/canvas";
-import { resolveWorkspaceName } from "../hooks/resolveWorkspaceName";
-
-beforeEach(() => {
-  // Reset store to a clean slate between tests so node lookup is deterministic.
-  useCanvasStore.setState({ nodes: [] });
-});
-
-describe("resolveWorkspaceName", () => {
-  it("returns the workspace name when a node with that ID exists", () => {
-    useCanvasStore.setState({
-      nodes: [
-        {
-          id: "ws-alpha-001",
-          type: "workspace",
-          data: { name: "Alpha Agent" },
-          position: { x: 0, y: 0 },
-        },
-      ],
-    });
-
-    expect(resolveWorkspaceName("ws-alpha-001")).toBe("Alpha Agent");
-  });
-
-  it("falls back to the first 8 chars of the ID when no matching node exists", () => {
-    expect(resolveWorkspaceName("ws-zzz-not-found")).toBe("ws-zzz-n");
-  });
-
-  it("falls back to the first 8 chars when the node exists but has no name", () => {
-    useCanvasStore.setState({
-      nodes: [
-        {
-          id: "ws-no-name",
-          type: "workspace",
-          // data.name is deliberately absent
-          data: {},
-          position: { x: 0, y: 0 },
-        },
-      ],
-    });
-
-    expect(resolveWorkspaceName("ws-no-name")).toBe("ws-no-na");
-  });
-
-  it("returns the first 8 chars for a very short ID", () => {
-    expect(resolveWorkspaceName("ab")).toBe("ab");
-  });
-
-  it("returns the first 8 chars when the ID is exactly 8 characters", () => {
-    // slice(0,8) of an 8-char string is the full string
-    const id = "12345678";
-    expect(resolveWorkspaceName(id)).toBe(id);
-  });
-
-  it("picks the right node when multiple workspaces share a prefix", () => {
-    useCanvasStore.setState({
-      nodes: [
-        {
-          id: "00000000-0000-0000-0000-000000000001",
-          type: "workspace",
-          data: { name: "Backend Agent" },
-          position: { x: 0, y: 0 },
-        },
-        {
-          id: "00000000-0000-0000-0000-000000000002",
-          type: "workspace",
-          data: { name: "Frontend Agent" },
-          position: { x: 100, y: 0 },
-        },
-      ],
-    });
-
-    expect(resolveWorkspaceName("00000000-0000-0000-0000-000000000002")).toBe(
-      "Frontend Agent"
-    );
-    expect(resolveWorkspaceName("00000000-0000-0000-0000-000000000001")).toBe(
-      "Backend Agent"
-    );
-  });
-
-  it("does not mutate store state between calls", () => {
-    useCanvasStore.setState({
-      nodes: [
-        {
-          id: "stable-id",
-          type: "workspace",
-          data: { name: "Stable Workspace" },
-          position: { x: 0, y: 0 },
-        },
-      ],
-    });
-
-    resolveWorkspaceName("stable-id");
-    resolveWorkspaceName("unknown-id");
-
-    // Store nodes must be unchanged — resolveWorkspaceName is read-only.
-    const nodes = useCanvasStore.getState().nodes;
-    expect(nodes).toHaveLength(1);
-    expect((nodes[0] as { id: string }).id).toBe("stable-id");
-  });
-});

canvas/src/components/tabs/chat/__tests__/useChatSocket.userMessage.test.ts

@@ -1,216 +0,0 @@
-// @vitest-environment jsdom
-/**
- * Tests for USER_MESSAGE event handling in useChatSocket.
- *
- * Covers issue #228: a canvas user's own outbound message was not fanned
- * out to other sessions — the originating session inserted it optimistically,
- * but other sessions only saw it after a manual refresh.
- *
- * The server now broadcasts USER_MESSAGE on canvas message/send. This test
- * verifies the canvas side consumes and forwards it to onUserMessage.
- */
-import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
-import { renderHook, act } from "@testing-library/react";
-import React from "react";
-import { useChatSocket, type UseChatSocketCallbacks } from "../hooks/useChatSocket";
-import { emitSocketEvent, _resetSocketEventListenersForTests } from "@/store/socket-events";
-import type { WSMessage } from "@/store/socket";
-
-// Silence React StrictMode double-invoke noise — we care about final state.
-const WARN = console.warn;
-beforeEach(() => { console.warn = () => {}; });
-afterEach(() => { console.warn = WARN; });
-
-beforeEach(() => {
-  _resetSocketEventListenersForTests();
-  vi.useFakeTimers();
-  vi.setSystemTime(new Date("2026-05-18T10:00:00Z"));
-});
-afterEach(() => {
-  vi.useRealTimers();
-  _resetSocketEventListenersForTests();
-});
-
-const WORKSPACE_ID = "00000000-0000-0000-0000-000000000001";
-
-function makeUserMessageEvent(
-  workspaceId: string,
-  overrides: Partial<{
-    message: string;
-    attachments: Array<{ uri: string; name: string; mimeType?: string; size?: number }>;
-    messageId: string;
-  }> = {},
-): WSMessage {
-  const { message = "Hello, agent!", attachments, messageId } = overrides;
-  const payload: Record<string, unknown> = { message };
-  if (attachments) payload.attachments = attachments;
-  if (messageId) payload.messageId = messageId;
-  return {
-    event: "USER_MESSAGE",
-    workspace_id: workspaceId,
-    timestamp: "2026-05-18T10:00:00Z",
-    payload,
-  };
-}
-
-describe("useChatSocket USER_MESSAGE handling", () => {
-  it("calls onUserMessage with a ChatMessage when USER_MESSAGE arrives for matching workspace", () => {
-    const onUserMessage = vi.fn();
-    const callbacks: UseChatSocketCallbacks = { onUserMessage };
-    const { result } = renderHook(() => useChatSocket(WORKSPACE_ID, callbacks));
-
-    act(() => {
-      emitSocketEvent(makeUserMessageEvent(WORKSPACE_ID, { message: "Hello!" }));
-    });
-
-    expect(onUserMessage).toHaveBeenCalledTimes(1);
-    const msg = onUserMessage.mock.calls[0][0];
-    expect(msg.role).toBe("user");
-    expect(msg.content).toBe("Hello!");
-    expect(typeof msg.id).toBe("string");
-    expect(msg.timestamp).toBe("2026-05-18T10:00:00.000Z");
-  });
-
-  it("calls onUserMessage with attachments extracted from the payload", () => {
-    const onUserMessage = vi.fn();
-    const callbacks: UseChatSocketCallbacks = { onUserMessage };
-    renderHook(() => useChatSocket(WORKSPACE_ID, callbacks));
-
-    act(() => {
-      emitSocketEvent(
-        makeUserMessageEvent(WORKSPACE_ID, {
-          message: "Here is the file",
-          attachments: [
-            { uri: "workspace:/uploads/report.pdf", name: "report.pdf", mimeType: "application/pdf", size: 4096 },
-          ],
-        }),
-      );
-    });
-
-    expect(onUserMessage).toHaveBeenCalledTimes(1);
-    const msg = onUserMessage.mock.calls[0][0];
-    expect(msg.role).toBe("user");
-    expect(msg.content).toBe("Here is the file");
-    expect(msg.attachments).toHaveLength(1);
-    expect(msg.attachments![0].uri).toBe("workspace:/uploads/report.pdf");
-    expect(msg.attachments![0].name).toBe("report.pdf");
-    expect(msg.attachments![0].mimeType).toBe("application/pdf");
-    expect(msg.attachments![0].size).toBe(4096);
-  });
-
-  it("does NOT call onUserMessage when workspace_id does not match", () => {
-    const onUserMessage = vi.fn();
-    const callbacks: UseChatSocketCallbacks = { onUserMessage };
-    renderHook(() => useChatSocket(WORKSPACE_ID, callbacks));
-
-    act(() => {
-      emitSocketEvent(
-        makeUserMessageEvent("00000000-0000-0000-0000-000000000099", { message: "wrong workspace" }),
-      );
-    });
-
-    expect(onUserMessage).not.toHaveBeenCalled();
-  });
-
-  it("does NOT call onUserMessage when message is empty and no attachments", () => {
-    const onUserMessage = vi.fn();
-    const callbacks: UseChatSocketCallbacks = { onUserMessage };
-    renderHook(() => useChatSocket(WORKSPACE_ID, callbacks));
-
-    act(() => {
-      emitSocketEvent(makeUserMessageEvent(WORKSPACE_ID, { message: "" }));
-    });
-
-    expect(onUserMessage).not.toHaveBeenCalled();
-  });
-
-  it("ignores USER_MESSAGE when onUserMessage callback is undefined", () => {
-    const callbacks: UseChatSocketCallbacks = { onAgentMessage: vi.fn() };
-    // Should not throw — undefined callback is guarded
-    expect(() =>
-      renderHook(() => useChatSocket(WORKSPACE_ID, callbacks)),
-    ).not.toThrow();
-
-    const { result } = renderHook(() => useChatSocket(WORKSPACE_ID, callbacks));
-    act(() => {
-      emitSocketEvent(makeUserMessageEvent(WORKSPACE_ID, { message: "Hello" }));
-    });
-    // No error thrown even without onUserMessage
-  });
-
-  it("other event types do NOT trigger onUserMessage", () => {
-    const onUserMessage = vi.fn();
-    const callbacks: UseChatSocketCallbacks = { onUserMessage };
-    renderHook(() => useChatSocket(WORKSPACE_ID, callbacks));
-
-    act(() => {
-      emitSocketEvent({
-        event: "A2A_RESPONSE",
-        workspace_id: WORKSPACE_ID,
-        timestamp: "2026-05-18T10:00:00Z",
-        payload: {},
-      });
-    });
-
-    expect(onUserMessage).not.toHaveBeenCalled();
-  });
-
-  it("re-fires onUserMessage for each USER_MESSAGE event received", () => {
-    const onUserMessage = vi.fn();
-    const callbacks: UseChatSocketCallbacks = { onUserMessage };
-    renderHook(() => useChatSocket(WORKSPACE_ID, callbacks));
-
-    act(() => {
-      emitSocketEvent(makeUserMessageEvent(WORKSPACE_ID, { message: "First message" }));
-    });
-    act(() => {
-      emitSocketEvent(makeUserMessageEvent(WORKSPACE_ID, { message: "Second message" }));
-    });
-
-    expect(onUserMessage).toHaveBeenCalledTimes(2);
-    expect(onUserMessage.mock.calls[0][0].content).toBe("First message");
-    expect(onUserMessage.mock.calls[1][0].content).toBe("Second message");
-  });
-
-  it("handles USER_MESSAGE with messageId in payload", () => {
-    const onUserMessage = vi.fn();
-    const callbacks: UseChatSocketCallbacks = { onUserMessage };
-    renderHook(() => useChatSocket(WORKSPACE_ID, callbacks));
-
-    act(() => {
-      emitSocketEvent(
-        makeUserMessageEvent(WORKSPACE_ID, { message: "With ID", messageId: "msg-id-abc" }),
-      );
-    });
-
-    expect(onUserMessage).toHaveBeenCalledTimes(1);
-    const msg = onUserMessage.mock.calls[0][0];
-    expect(msg.content).toBe("With ID");
-  });
-
-  it("filters out attachments with empty uri or name (defence-in-depth)", () => {
-    const onUserMessage = vi.fn();
-    const callbacks: UseChatSocketCallbacks = { onUserMessage };
-    renderHook(() => useChatSocket(WORKSPACE_ID, callbacks));
-
-    act(() => {
-      emitSocketEvent(
-        makeUserMessageEvent(WORKSPACE_ID, {
-          message: "Mixed attachments",
-          attachments: [
-            { uri: "workspace:/uploads/good.pdf", name: "good.pdf" },
-            { uri: "", name: "bad.pdf" }, // empty uri — dropped
-            { uri: "workspace:/uploads/also-bad", name: "" }, // empty name — dropped
-            { uri: "workspace:/uploads/also-good.txt", name: "also-good.txt" },
-          ],
-        }),
-      );
-    });
-
-    expect(onUserMessage).toHaveBeenCalledTimes(1);
-    const msg = onUserMessage.mock.calls[0][0];
-    expect(msg.attachments).toHaveLength(2);
-    expect(msg.attachments![0].name).toBe("good.pdf");
-    expect(msg.attachments![1].name).toBe("also-good.txt");
-  });
-});

canvas/src/components/tabs/chat/a2aErrorHint.ts

@@ -10,37 +10,10 @@
  * had already drifted (Activity tab gained `not found`/`offline`
  * cases AgentCommsPanel never picked up) — this module is the merged
  * superset and the only place hint text should change.
- *
- * Optional `context.peerKind` lets callers signal "the callee was a
- * codex-runtime task" so the timeout-class hints can be more specific
- * about expected long completion times (PM-coordinating-Researcher is
- * the canonical case where the 600s sync-proxy budget is too tight).
  */
-export interface A2AErrorContext {
-  /** Runtime of the callee, when known. e.g. "codex", "claude-code". */
-  peerKind?: string;
-}
-
-export function inferA2AErrorHint(
-  detail: string,
-  context?: A2AErrorContext,
-): string {
+export function inferA2AErrorHint(detail: string): string {
   const t = detail.toLowerCase();
 
-  // Poll-mode budget exhaustion (a2a_tools_delegation.py emits
-  // "polling timeout after Ns ... call check_task_status(...) to
-  // retrieve later"). This is NOT a delivery failure — the work is
-  // still in flight on the platform side. Route to a specific hint
-  // BEFORE the generic timeout bucket so the user gets the actionable
-  // "wait + check_task_status" guidance instead of the misleading
-  // "restart the workspace" anti-pattern.
-  if (
-    t.includes("polling timeout after") ||
-    t.includes("call check_task_status")
-  ) {
-    return "The 600s sync-polling budget expired but the platform is still working on the delegation. Do NOT restart — the work isn't lost. Wait, then call check_task_status with the delegation_id to retrieve the result. If the callee is a long-running codex task, this is expected.";
-  }
-
   // "control request timeout" is the specific Claude Code SDK init
   // wedge symptom. Pattern on the full phrase, not bare "initialize"
   // — a user task containing "failed to initialize database" would
@@ -54,13 +27,6 @@ export function inferA2AErrorHint(
     t.includes("deadline exceeded") ||
     t.includes("timeout")
   ) {
-    // For codex callees, a 600s sync-proxy timeout is the EXPECTED
-    // shape when the task is genuinely long-running. Calling out the
-    // workspace-restart anti-pattern explicitly per
-    // `feedback_surface_actionable_failure_reason_to_user`.
-    if ((context?.peerKind || "").toLowerCase() === "codex") {
-      return "The codex remote agent didn't respond within the 600s sync-proxy timeout. Codex tasks can legitimately run longer than this — check the callee's Activity tab; the work may still be progressing. Restart only if the container is genuinely stuck (no activity for several minutes).";
-    }
     return "The remote agent didn't respond within the proxy timeout. It may be busy with a long task, or the runtime is stuck — restart the workspace if this repeats.";
   }
   if (
@@ -82,16 +48,7 @@ export function inferA2AErrorHint(
     return "The remote workspace can't be reached — it may be stopped, removed, or outside the access control list. Verify the peer is online before retrying.";
   }
   if (detail === "") {
-    // Per `feedback_surface_actionable_failure_reason_to_user`: a bare
-    // "restart the workspace" prompt is the anti-pattern when the
-    // underlying failure was a silent timeout against a long-running
-    // remote (codex Researcher being coordinated by PM is the
-    // canonical case). If the caller knows the peer is codex, route
-    // to the more specific hint that explicitly de-recommends restart.
-    if ((context?.peerKind || "").toLowerCase() === "codex") {
-      return "The codex remote agent returned no error detail — most often the 600s sync-proxy budget expired before the task finished. The work may still be progressing on the callee side; check its Activity tab before restarting.";
-    }
-    return "The remote agent returned no error detail (the underlying httpx exception had an empty message — typically a connection-reset or silent timeout). Check the callee's Activity tab to see if work is still in flight before restarting.";
+    return "The remote agent returned no error detail (the underlying httpx exception had an empty message — typically a connection-reset or silent timeout). A workspace restart is the safe first move.";
   }
   return "The remote agent reported a delivery failure. Check the workspace logs or try restarting.";
 }

canvas/src/components/tabs/chat/hooks/__tests__/useChatSend.test.tsx

@@ -1,209 +0,0 @@
-// @vitest-environment jsdom
-/**
- * Tests for useChatSend — the canvas user→agent send hook.
- *
- * Behavioural focus: the poll-mode ("queued") path. When the target
- * workspace is an external / MCP-registered agent (delivery_mode=poll,
- * e.g. an operator laptop running the molecule MCP channel), the
- * platform's POST /workspaces/:id/a2a returns a synthetic
- * {status:"queued", delivery_mode:"poll"} envelope IMMEDIATELY with no
- * reply — the real reply arrives later over the AGENT_MESSAGE
- * WebSocket push.
- *
- * Pre-fix the hook treated that synthetic envelope as a terminal
- * response and called releaseSendGuards() → `sending` went false the
- * instant the POST returned → the "agent is working" indicator
- * vanished and the external turn looked dead. This suite pins the
- * fixed contract:
- *
- *   - a real reply still clears `sending` (regression guard)
- *   - a poll "queued" envelope KEEPS `sending` true (no terminal
- *     clear) so the existing thinking indicator persists
- *   - the eventual reply path (releaseSendGuards, the same call the
- *     AGENT_MESSAGE WS push makes via useChatSocket) clears it
- *   - an offline poll agent that never replies eventually surfaces an
- *     honest error instead of an infinite spinner
- *
- * Plus pure-function coverage for the poll-envelope detector.
- *
- * Root cause: workspace-server a2a_proxy.go:402 poll-mode
- * short-circuit returns {status:"queued"} synchronously.
- */
-import {
-  describe,
-  it,
-  expect,
-  vi,
-  beforeEach,
-  afterEach,
-  type Mock,
-} from "vitest";
-import { act, renderHook, cleanup } from "@testing-library/react";
-
-const { mockApiPost } = vi.hoisted(() => ({ mockApiPost: vi.fn() }));
-
-vi.mock("@/lib/api", () => ({
-  api: { post: mockApiPost },
-}));
-
-vi.mock("../uploads", () => ({
-  uploadChatFiles: vi.fn(),
-}));
-
-// Import AFTER mocks.
-import {
-  useChatSend,
-  isPollQueuedResponse,
-  extractReplyText,
-  POLL_QUEUED_REPLY_TIMEOUT_MS,
-} from "../useChatSend";
-
-const flush = () => act(async () => { await Promise.resolve(); });
-
-describe("isPollQueuedResponse", () => {
-  it("is true only for the synthetic poll-mode queued envelope", () => {
-    expect(isPollQueuedResponse({ status: "queued", delivery_mode: "poll" })).toBe(true);
-  });
-
-  it("is false for a real agent reply", () => {
-    expect(
-      isPollQueuedResponse({ result: { parts: [{ kind: "text", text: "hi" }] } }),
-    ).toBe(false);
-  });
-
-  it("is false for null / undefined / partial shapes", () => {
-    expect(isPollQueuedResponse(null)).toBe(false);
-    expect(isPollQueuedResponse(undefined)).toBe(false);
-    // status=queued without delivery_mode=poll is NOT the poll envelope
-    // — don't accidentally swallow a real reply that happens to carry
-    // an unrelated status field.
-    expect(isPollQueuedResponse({ status: "queued" })).toBe(false);
-    expect(isPollQueuedResponse({ delivery_mode: "poll" })).toBe(false);
-  });
-});
-
-describe("extractReplyText (regression guard — unchanged by fix)", () => {
-  it("collects text parts from result", () => {
-    expect(
-      extractReplyText({ result: { parts: [{ kind: "text", text: "hello" }] } }),
-    ).toBe("hello");
-  });
-  it("returns empty for the poll-queued envelope", () => {
-    expect(extractReplyText({ status: "queued", delivery_mode: "poll" })).toBe("");
-  });
-});
-
-describe("useChatSend — poll-mode in-progress state", () => {
-  beforeEach(() => {
-    vi.useFakeTimers();
-    mockApiPost.mockReset();
-  });
-  afterEach(() => {
-    vi.runOnlyPendingTimers();
-    vi.useRealTimers();
-    cleanup();
-  });
-
-  const setup = () => {
-    const onUserMessage = vi.fn();
-    const onAgentMessage = vi.fn();
-    const { result } = renderHook(() =>
-      useChatSend("ws-ext-1", {
-        getHistoryMessages: () => [],
-        onUserMessage,
-        onAgentMessage,
-      }),
-    );
-    return { result, onUserMessage, onAgentMessage };
-  };
-
-  it("a real reply clears `sending` (regression guard)", async () => {
-    mockApiPost.mockResolvedValue({
-      result: { parts: [{ kind: "text", text: "real reply" }] },
-    });
-    const { result, onAgentMessage } = setup();
-
-    await act(async () => {
-      void result.current.sendMessage("hi");
-    });
-    await flush();
-
-    expect(onAgentMessage).toHaveBeenCalledTimes(1);
-    expect(result.current.sending).toBe(false);
-  });
-
-  it("keeps `sending` true on a poll 'queued' envelope (no terminal clear)", async () => {
-    mockApiPost.mockResolvedValue({ status: "queued", delivery_mode: "poll" });
-    const { result, onAgentMessage } = setup();
-
-    await act(async () => {
-      void result.current.sendMessage("hi external agent");
-    });
-    await flush();
-
-    // The POST resolved, but it was only a queued ack — the indicator
-    // must stay up and no agent bubble should be rendered yet.
-    expect(result.current.sending).toBe(true);
-    expect(onAgentMessage).not.toHaveBeenCalled();
-    expect(result.current.error).toBeNull();
-  });
-
-  it("releaseSendGuards (the AGENT_MESSAGE-push path) clears the poll in-progress state", async () => {
-    mockApiPost.mockResolvedValue({ status: "queued", delivery_mode: "poll" });
-    const { result } = setup();
-
-    await act(async () => {
-      void result.current.sendMessage("hi");
-    });
-    await flush();
-    expect(result.current.sending).toBe(true);
-
-    // Simulate the terminal AGENT_MESSAGE WebSocket push arriving:
-    // useChatSocket's onAgentMessage / onSendComplete call
-    // releaseSendGuards. That must clear the in-progress state AND the
-    // safety timer (asserted by the next test).
-    act(() => {
-      result.current.releaseSendGuards();
-    });
-    expect(result.current.sending).toBe(false);
-  });
-
-  it("surfaces an honest error if a poll agent never replies (safety timeout)", async () => {
-    mockApiPost.mockResolvedValue({ status: "queued", delivery_mode: "poll" });
-    const { result } = setup();
-
-    await act(async () => {
-      void result.current.sendMessage("hi");
-    });
-    await flush();
-    expect(result.current.sending).toBe(true);
-
-    act(() => {
-      vi.advanceTimersByTime(POLL_QUEUED_REPLY_TIMEOUT_MS + 1000);
-    });
-
-    expect(result.current.sending).toBe(false);
-    expect(result.current.error).toMatch(/queued/i);
-  });
-
-  it("does NOT fire the safety error when the reply arrives before timeout", async () => {
-    mockApiPost.mockResolvedValue({ status: "queued", delivery_mode: "poll" });
-    const { result } = setup();
-
-    await act(async () => {
-      void result.current.sendMessage("hi");
-    });
-    await flush();
-
-    // Reply arrives (releaseSendGuards) well before the timeout.
-    act(() => {
-      result.current.releaseSendGuards();
-    });
-    act(() => {
-      vi.advanceTimersByTime(POLL_QUEUED_REPLY_TIMEOUT_MS + 1000);
-    });
-
-    expect(result.current.error).toBeNull();
-    expect(result.current.sending).toBe(false);
-  });
-});

canvas/src/components/tabs/chat/hooks/useChatHistory.ts

@@ -2,7 +2,6 @@
 
 import { useCallback, useEffect, useRef, useState } from "react";
 import { api } from "@/lib/api";
-import { subscribeSocketResume } from "@/store/socket-events";
 import { type ChatMessage, appendMessageDeduped as appendMessageDedupedFn } from "../types";
 
 const INITIAL_HISTORY_LIMIT = 10;
@@ -83,23 +82,6 @@ export function useChatHistory(
     loadInitial();
   }, [loadInitial]);
 
-  // Back-fill on socket resume. The singleton WS emits this when it
-  // recovers from a down period (ordinary drop, or — the case this
-  // fixes — a mobile-browser background-suspend that silently killed
-  // the socket while the page was frozen). While the socket was dead
-  // every AGENT_MESSAGE / A2A_RESPONSE for this thread was missed, and
-  // the store's rehydrate() only re-pulls /workspaces status, not chat.
-  // Re-running loadInitial() re-fetches the latest persisted history —
-  // exactly what a navigate-away-and-back (remount) does today, but
-  // without the user having to do it. Shared by desktop ChatTab and
-  // MobileChat (both consume this hook), so the realtime path stays
-  // unified across surfaces rather than forked for mobile.
-  useEffect(() => {
-    return subscribeSocketResume(() => {
-      loadInitial();
-    });
-  }, [loadInitial]);
-
   const loadOlder = useCallback(async () => {
     if (inflightRef.current || !hasMoreRef.current) return;
     const oldest = oldestMessageRef.current;

canvas/src/components/tabs/chat/hooks/useChatSend.ts

@@ -1,6 +1,6 @@
 "use client";
 
-import { useCallback, useEffect, useRef, useState } from "react";
+import { useCallback, useRef, useState } from "react";
 import { api } from "@/lib/api";
 import { uploadChatFiles } from "../uploads";
 import { createMessage, type ChatMessage, type ChatAttachment } from "../types";
@@ -22,42 +22,8 @@ interface A2AResponse {
     parts?: A2APart[];
     artifacts?: Array<{ parts: A2APart[] }>;
   };
-  /** Synthetic poll-mode envelope. The platform returns this
-   *  immediately (HTTP 200) when the target workspace is registered
-   *  delivery_mode=poll — an external / MCP-registered agent with no
-   *  public URL (e.g. an operator's laptop running the molecule MCP
-   *  channel). The request has only been QUEUED into activity_logs;
-   *  the agent will pick it up on its next poll and the real reply
-   *  arrives asynchronously over the AGENT_MESSAGE WebSocket push
-   *  (consumed by useChatSocket). See workspace-server
-   *  a2a_proxy.go:402 (poll-mode short-circuit) and
-   *  a2a_proxy_helpers.go:516 (logA2AReceiveQueued). */
-  status?: string;
-  delivery_mode?: string;
 }
 
-/** True when `resp` is the platform's synthetic poll-mode "queued"
- *  envelope rather than a real agent reply. For these the send is
- *  acknowledged-but-pending: the user's message landed and the agent
- *  is working, but there is no reply yet — the terminal AGENT_MESSAGE
- *  push will arrive later over the WebSocket. Treating this as a
- *  terminal response (the pre-fix behaviour) cleared the "agent is
- *  working" indicator the instant the POST returned, so an external
- *  workspace turn looked dead even though work had not started. */
-export function isPollQueuedResponse(resp: A2AResponse | null | undefined): boolean {
-  return !!resp && resp.status === "queued" && resp.delivery_mode === "poll";
-}
-
-/** Hard ceiling on how long the "agent is working" indicator stays up
- *  for a poll-mode turn with no reply. The terminal AGENT_MESSAGE push
- *  normally clears it well before this. The cap exists so a poll-mode
- *  workspace that is offline / never consumes its queue doesn't pin a
- *  spinner forever — at which point we surface an honest, actionable
- *  error instead of an opaque dead spinner. Generous because poll
- *  agents (an operator laptop) can legitimately take minutes to wake,
- *  poll, and respond; the goal is "eventually honest", not fail-fast. */
-export const POLL_QUEUED_REPLY_TIMEOUT_MS = 15 * 60 * 1000;
-
 export function extractReplyText(resp: A2AResponse): string {
   const collect = (parts: A2APart[] | undefined): string => {
     if (!parts) return "";
@@ -93,29 +59,14 @@ export function useChatSend(workspaceId: string, options: UseChatSendOptions) {
   const sendInFlightRef = useRef(false);
   const sendingFromAPIRef = useRef(false);
   const sendTokenRef = useRef(0);
-  // Safety-net timer armed only for poll-mode ("queued") turns: the
-  // POST returns immediately with no reply, so the normal
-  // POST-resolves-→-clear-spinner path can't drive the indicator. The
-  // terminal AGENT_MESSAGE WebSocket push clears it via
-  // releaseSendGuards (which also clears this timer); the timer is the
-  // backstop for an offline poll agent that never consumes its queue.
-  const pollTimeoutRef = useRef<ReturnType<typeof setTimeout> | null>(null);
   const optionsRef = useRef(options);
   optionsRef.current = options;
 
-  const clearPollTimeout = useCallback(() => {
-    if (pollTimeoutRef.current !== null) {
-      clearTimeout(pollTimeoutRef.current);
-      pollTimeoutRef.current = null;
-    }
-  }, []);
-
   const releaseSendGuards = useCallback(() => {
-    clearPollTimeout();
     setSending(false);
     sendingFromAPIRef.current = false;
     sendInFlightRef.current = false;
-  }, [clearPollTimeout]);
+  }, []);
 
   const clearError = useCallback(() => setError(null), []);
 
@@ -195,33 +146,6 @@ export function useChatSend(workspaceId: string, options: UseChatSendOptions) {
             sendInFlightRef.current = false;
             return;
           }
-          // Poll-mode ("queued") turn: the message landed and the
-          // external/MCP agent will pick it up on its next poll, but
-          // there is NO reply in this response. Pre-fix this fell
-          // through to releaseSendGuards() below and the "agent is
-          // working" indicator vanished the instant the POST returned —
-          // an external-workspace turn looked dead even though work had
-          // not started. Instead, keep `sending` true so the existing
-          // thinking indicator (the same one internal agents use)
-          // persists as a "received — agent is working" state; the
-          // terminal AGENT_MESSAGE WebSocket push (consumed by
-          // useChatSocket → onAgentMessage / onSendComplete →
-          // releaseSendGuards) clears it when the real reply arrives,
-          // exactly the path an internal async reply already uses.
-          if (isPollQueuedResponse(resp)) {
-            clearPollTimeout();
-            pollTimeoutRef.current = setTimeout(() => {
-              if (sendTokenRef.current !== myToken) return;
-              if (!sendingFromAPIRef.current) return;
-              releaseSendGuards();
-              setError(
-                "No response yet from this agent — it may be offline or " +
-                  "busy. Your message was delivered and is queued; the " +
-                  "reply will appear here if the agent picks it up.",
-              );
-            }, POLL_QUEUED_REPLY_TIMEOUT_MS);
-            return;
-          }
           const replyText = extractReplyText(resp);
           const replyFiles = extractFilesFromTask(
             (resp?.result ?? {}) as Record<string, unknown>,
@@ -243,15 +167,9 @@ export function useChatSend(workspaceId: string, options: UseChatSendOptions) {
           setError("Failed to send message — agent may be unreachable");
         });
     },
-    [workspaceId, sending, uploading, clearPollTimeout],
+    [workspaceId, sending, uploading],
   );
 
-  // Drop the poll-mode safety timer on unmount / workspace switch so a
-  // stale timeout can't fire setError against a panel the user has
-  // already navigated away from. sendTokenRef guards correctness if it
-  // ever did fire; this just avoids the wasted timer + setState churn.
-  useEffect(() => clearPollTimeout, [clearPollTimeout]);
-
   return {
     sending,
     uploading,

canvas/src/components/tabs/chat/hooks/useChatSocket.ts

@@ -7,10 +7,6 @@ import { createMessage, type ChatMessage } from "../types";
 
 export interface UseChatSocketCallbacks {
   onAgentMessage?: (msg: ChatMessage) => void;
-  /** Called when another session sent a user message — used to fan out
-   *  the user's own outbound text to all sessions so a second device
-   *  sees the question live without a manual refresh (issue #228). */
-  onUserMessage?: (msg: ChatMessage) => void;
   onActivityLog?: (entry: string) => void;
   onSendComplete?: () => void;
   onSendError?: (error: string) => void;
@@ -47,33 +43,6 @@ export function useChatSocket(
 
   useSocketEvent((msg) => {
     try {
-      if (msg.event === "USER_MESSAGE" && msg.workspace_id === workspaceId) {
-        const p = msg.payload || {};
-        const message = typeof p.message === "string" ? p.message : "";
-        const rawAttachments = p.attachments;
-        const attachments =
-          Array.isArray(rawAttachments)
-            ? (rawAttachments as Array<{ uri?: unknown; name?: unknown; mimeType?: unknown; size?: unknown }>)
-                .filter(
-                  (a) =>
-                    typeof a?.uri === "string" && a.uri.length > 0 &&
-                    typeof a?.name === "string" && a.name.length > 0,
-                )
-                .map((a) => ({
-                  uri: a.uri as string,
-                  name: a.name as string,
-                  mimeType: typeof a.mimeType === "string" ? a.mimeType : undefined,
-                  size: typeof a.size === "number" ? a.size : undefined,
-                }))
-            : undefined;
-        if (message || (attachments && attachments.length > 0)) {
-          callbacksRef.current.onUserMessage?.(
-            createMessage("user", message, attachments),
-          );
-        }
-        return;
-      }
-
       if (msg.event === "ACTIVITY_LOGGED") {
         if (msg.workspace_id !== workspaceId) return;
 
@@ -98,21 +67,9 @@ export function useChatSocket(
             const own = (targetId || msg.workspace_id) === workspaceId;
             if (own) {
               callbacksRef.current.onSendComplete?.();
-              // internal#211/#212: surface the runtime's curated,
-              // user-actionable reason (provider HTTP status + error
-              // code + the provider's own guidance, e.g. a 403 "org
-              // disabled · use an API key / ask your admin"). The
-              // server now includes error_detail in the ACTIVITY_LOGGED
-              // broadcast; fall back to summary, and only as a last
-              // resort to a generic line. The old hardcoded
-              // "Agent error (Exception) — see workspace logs for
-              // details." string pointed at a logs UI that does not
-              // exist and discarded the actionable reason entirely.
-              const detail =
-                (p.error_detail as string) ||
-                (p.summary as string) ||
-                "The agent turn failed but the runtime reported no detail. Retry once; if it repeats the workspace runtime may need a restart.";
-              callbacksRef.current.onSendError?.(detail);
+              callbacksRef.current.onSendError?.(
+                "Agent error (Exception) — see workspace logs for details.",
+              );
             }
           }
         } else if (type === "a2a_send") {

canvas/src/components/tabs/config/secrets-section.tsx

@@ -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}

canvas/src/components/ui/TestConnectionButton.tsx

@@ -99,7 +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>
   );

canvas/src/hooks/__tests__/useKeyboardShortcut.test.tsx

@@ -1,166 +0,0 @@
-// @vitest-environment jsdom
-/**
- * Tests for useKeyboardShortcut.
- *
- * Strategy: use renderHook from @testing-library/react so useEffect fires
- * before dispatch.  We spy on window.addEventListener to capture the registered
- * handler.  Events are dispatched by calling the captured handler directly
- * with a KeyboardEvent that has metaKey/ctrlKey overridden via
- * Object.defineProperty (jsdom's built-in modifier-key event is unreliable).
- */
-import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
-import { cleanup, act, renderHook } from "@testing-library/react";
-import { useState, useCallback } from "react";
-import { useKeyboardShortcut } from "../use-keyboard-shortcut";
-
-afterEach(cleanup);
-
-// Capture the most-recently registered keydown handler so tests can dispatch through it.
-let registeredHandler: ((e: KeyboardEvent) => void) | null = null;
-
-const addSpy = vi.spyOn(window, "addEventListener").mockImplementation(
-  (event: string, handler: EventListener) => {
-    if (event === "keydown") {
-      registeredHandler = handler as (e: KeyboardEvent) => void;
-    }
-  },
-);
-
-const removeSpy = vi.spyOn(window, "removeEventListener").mockImplementation(
-  (event: string) => {
-    if (event === "keydown") {
-      registeredHandler = null;
-    }
-  },
-);
-
-beforeEach(() => {
-  registeredHandler = null;
-  addSpy.mockClear();
-  removeSpy.mockClear();
-});
-
-/**
- * Dispatch a keydown event through the captured handler.
- * Wrapped in act() so React flushes any state updates synchronously.
- * Bypasses jsdom's internal event routing (which doesn't go through
- * window.EventTarget.prototype.addEventListener for fireEvent dispatch).
- */
-function dispatchKeydown(
-  key: string,
-  { meta = false, ctrl = false }: { meta?: boolean; ctrl?: boolean } = {},
-) {
-  act(() => {
-    const e = new KeyboardEvent("keydown", { key, bubbles: true });
-    Object.defineProperty(e, "metaKey", { value: meta });
-    Object.defineProperty(e, "ctrlKey", { value: ctrl });
-    registeredHandler?.(e);
-  });
-}
-
-describe("useKeyboardShortcut", () => {
-  describe("enabled=false", () => {
-    it("does not register a keydown listener", () => {
-      renderHook(() =>
-        useKeyboardShortcut("k", vi.fn(), { enabled: false }),
-      );
-      expect(addSpy).not.toHaveBeenCalledWith("keydown", expect.any(Function));
-    });
-  });
-
-  describe("meta modifier", () => {
-    it("fires callback on Cmd+K", () => {
-      const cb = vi.fn();
-      renderHook(() => useKeyboardShortcut("k", cb, { meta: true }));
-      dispatchKeydown("k", { meta: true });
-      expect(cb).toHaveBeenCalledTimes(1);
-    });
-
-    it("does NOT fire on Ctrl+K when only meta=true", () => {
-      const cb = vi.fn();
-      renderHook(() => useKeyboardShortcut("k", cb, { meta: true }));
-      dispatchKeydown("k", { ctrl: true });
-      expect(cb).not.toHaveBeenCalled();
-    });
-
-    it("does NOT fire on plain K even with meta=true", () => {
-      const cb = vi.fn();
-      renderHook(() => useKeyboardShortcut("k", cb, { meta: true }));
-      dispatchKeydown("k", { meta: false, ctrl: false });
-      expect(cb).not.toHaveBeenCalled();
-    });
-  });
-
-  describe("ctrl modifier", () => {
-    it("fires callback on Ctrl+K", () => {
-      const cb = vi.fn();
-      renderHook(() => useKeyboardShortcut("k", cb, { ctrl: true }));
-      dispatchKeydown("k", { ctrl: true });
-      expect(cb).toHaveBeenCalledTimes(1);
-    });
-
-    it("does NOT fire on Cmd+K when only ctrl=true", () => {
-      const cb = vi.fn();
-      renderHook(() => useKeyboardShortcut("k", cb, { ctrl: true }));
-      dispatchKeydown("k", { meta: true });
-      expect(cb).not.toHaveBeenCalled();
-    });
-  });
-
-  describe("no-modifier guard", () => {
-    it("does not fire when no modifier is held", () => {
-      const cb = vi.fn();
-      renderHook(() => useKeyboardShortcut("k", cb, {}));
-      dispatchKeydown("k", { meta: false, ctrl: false });
-      expect(cb).not.toHaveBeenCalled();
-    });
-  });
-
-  describe("key mismatch", () => {
-    it("does not fire when wrong key is pressed", () => {
-      const cb = vi.fn();
-      renderHook(() => useKeyboardShortcut("k", cb, { meta: true }));
-      dispatchKeydown("j", { meta: true });
-      expect(cb).not.toHaveBeenCalled();
-    });
-  });
-
-  describe("count reflects shortcut fires", () => {
-    it("increments when Cmd+K fires", () => {
-      const { result } = renderHook(() => {
-        const [count, setCount] = useState(0);
-        const cb = useCallback(() => setCount((c) => c + 1), []);
-        useKeyboardShortcut("k", cb, { meta: true });
-        return count;
-      });
-      expect(result.current).toBe(0);
-      dispatchKeydown("k", { meta: true });
-      expect(result.current).toBe(1);
-      dispatchKeydown("k", { meta: true });
-      expect(result.current).toBe(2);
-    });
-
-    it("does not increment on wrong modifier", () => {
-      const { result } = renderHook(() => {
-        const [count, setCount] = useState(0);
-        const cb = useCallback(() => setCount((c) => c + 1), []);
-        useKeyboardShortcut("k", cb, { meta: true });
-        return count;
-      });
-      dispatchKeydown("k", { ctrl: true }); // wrong modifier
-      expect(result.current).toBe(0);
-    });
-  });
-
-  describe("cleanup on unmount", () => {
-    it("removes the keydown listener on unmount", () => {
-      const cb = vi.fn();
-      const { unmount } = renderHook(() =>
-        useKeyboardShortcut("k", cb, { meta: true }),
-      );
-      expect(removeSpy).not.toHaveBeenCalled();
-      unmount();
-      expect(removeSpy).toHaveBeenCalledWith("keydown", expect.any(Function));
-    });
-  });
-});

canvas/src/hooks/__tests__/useSocketEvent.test.tsx

@@ -1,84 +0,0 @@
-// @vitest-environment jsdom
-/**
- * Tests for useSocketEvent.
- *
- * Covers:
- *   - subscribeSocketEvents is called on mount
- *   - Unsubscribe is called on unmount
- *   - subscribeSocketEvents is called only once (ref-based, not render-based)
- */
-import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
-import { render, cleanup } from "@testing-library/react";
-import React from "react";
-import { useSocketEvent } from "../useSocketEvent";
-
-afterEach(cleanup);
-
-// Mutable ref shared between vi.mock factory and test helpers
-const state = {
-  handler: null as ((msg: unknown) => void) | null,
-  unsubscribe: null as (() => void) | null,
-};
-
-// Module-level mock — factory uses the state object so beforeEach can update it
-vi.mock("@/store/socket-events", () => ({
-  subscribeSocketEvents: vi.fn().mockImplementation(() => {
-    if (state.unsubscribe) return state.unsubscribe;
-    const fn = vi.fn();
-    state.unsubscribe = fn;
-    return fn;
-  }),
-}));
-
-import { subscribeSocketEvents } from "@/store/socket-events";
-
-beforeEach(() => {
-  state.handler = null;
-  state.unsubscribe = null;
-  vi.mocked(subscribeSocketEvents).mockImplementation(() => {
-    const fn = vi.fn();
-    state.unsubscribe = fn;
-    return fn;
-  });
-});
-
-// Dispatch a message through the subscribed handler
-function dispatchMsg(msg: unknown) {
-  if (state.handler) {
-    state.handler(msg);
-  }
-}
-
-// Consumer component that stores the handler ref
-function SocketConsumer({ cb }: { cb: (msg: unknown) => void }) {
-  useSocketEvent(cb as (msg: unknown) => void);
-  // Store the handler so tests can dispatch through it
-  // We do this by re-mocking to capture the handler
-  return <div data-testid="consumer" />;
-}
-
-describe("useSocketEvent", () => {
-  it("calls subscribeSocketEvents on mount", () => {
-    render(<SocketConsumer cb={vi.fn()} />);
-    expect(subscribeSocketEvents).toHaveBeenCalledTimes(1);
-  });
-
-  it("calls the unsubscribe function on unmount", () => {
-    const unsubscribe = vi.fn();
-    vi.mocked(subscribeSocketEvents).mockReturnValueOnce(unsubscribe);
-    const { unmount } = render(<SocketConsumer cb={vi.fn()} />);
-    unmount();
-    expect(unsubscribe).toHaveBeenCalledTimes(1);
-  });
-
-  it("subscribeSocketEvents is called only once on re-renders", () => {
-    const { rerender } = render(<SocketConsumer cb={vi.fn()} />);
-    const initial = vi.mocked(subscribeSocketEvents).mock.calls.length;
-
-    rerender(<SocketConsumer cb={vi.fn()} />);
-    rerender(<SocketConsumer cb={vi.fn()} />);
-    rerender(<SocketConsumer cb={vi.fn()} />);
-
-    expect(vi.mocked(subscribeSocketEvents).mock.calls.length).toBe(initial);
-  });
-});

canvas/src/hooks/__tests__/useWorkspaceName.test.tsx

@@ -1,98 +0,0 @@
-// @vitest-environment jsdom
-/**
- * Tests for useWorkspaceName.
- *
- * Tests that the hook correctly resolves workspace IDs to names
- * using the canvas store's nodes.
- */
-import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
-import { renderHook, cleanup } from "@testing-library/react";
-import React from "react";
-import { useWorkspaceName } from "../useWorkspaceName";
-
-afterEach(cleanup);
-
-const mockNodes = [
-  { id: "ws-1", data: { name: "Alpha Workspace" } },
-  { id: "ws-2", data: { name: "Beta Workspace" } },
-  { id: "ws-3", data: {} }, // node without name
-  { id: "ws-4", data: { name: "" } }, // empty name
-] as const;
-
-// Stable reference so useCallback deps are stable across re-renders
-const stableNodes = [...mockNodes];
-
-vi.mock("@/store/canvas", () => ({
-  useCanvasStore: Object.assign(
-    vi.fn((selector?: (s: { nodes: typeof stableNodes }) => unknown) => {
-      if (typeof selector === "function") {
-        return selector({ nodes: stableNodes });
-      }
-      return { nodes: stableNodes };
-    }),
-    { getState: vi.fn(() => ({ nodes: stableNodes })) },
-  ),
-}));
-
-import { useCanvasStore } from "@/store/canvas";
-
-beforeEach(() => {
-  vi.mocked(useCanvasStore).mockClear();
-});
-
-describe("useWorkspaceName", () => {
-  it("returns the workspace name for a known ID", () => {
-    const { result } = renderHook(() => {
-      const resolve = useWorkspaceName();
-      return resolve("ws-1");
-    });
-    expect(result.current).toBe("Alpha Workspace");
-  });
-
-  it("returns the workspace name for another known ID", () => {
-    const { result } = renderHook(() => {
-      const resolve = useWorkspaceName();
-      return resolve("ws-2");
-    });
-    expect(result.current).toBe("Beta Workspace");
-  });
-
-  it("returns empty string for null", () => {
-    const { result } = renderHook(() => {
-      const resolve = useWorkspaceName();
-      return resolve(null);
-    });
-    expect(result.current).toBe("");
-  });
-
-  it("falls back to first 8 chars of ID when node has no name", () => {
-    const { result } = renderHook(() => {
-      const resolve = useWorkspaceName();
-      return resolve("ws-3");
-    });
-    expect(result.current).toBe("ws-3".slice(0, 8));
-  });
-
-  it("falls back to first 8 chars of ID when name is empty string", () => {
-    const { result } = renderHook(() => {
-      const resolve = useWorkspaceName();
-      return resolve("ws-4");
-    });
-    expect(result.current).toBe("ws-4".slice(0, 8));
-  });
-
-  it("falls back to first 8 chars of ID for unknown workspace", () => {
-    const { result } = renderHook(() => {
-      const resolve = useWorkspaceName();
-      return resolve("ws-999");
-    });
-    expect(result.current).toBe("ws-999".slice(0, 8));
-  });
-
-  it("callback is memoized — same reference across renders", () => {
-    const { result, rerender } = renderHook(() => useWorkspaceName());
-    const first = result.current;
-    rerender();
-    expect(result.current).toBe(first);
-  });
-});

canvas/src/lib/__tests__/cssVar.test.ts

@@ -1,32 +1,67 @@
 // @vitest-environment jsdom
+/**
+ * Tests for cssVar — maps ColorToken to a CSS variable string.
+ *
+ * Exists for the rare case where an inline style="" or SVG fill needs
+ * a token value rather than a Tailwind class. The returned var(--color-foo)
+ * string follows the live theme without re-renders.
+ */
 import { describe, it, expect } from "vitest";
-import { cssVar, type ColorToken } from "../theme";
+import { cssVar } from "../theme";
+import type { ColorToken } from "../theme";
 
 describe("cssVar", () => {
-  const tokens: ColorToken[] = [
-    "surface", "surface-elevated", "surface-sunken", "surface-card",
-    "line", "line-soft", "ink", "ink-mid", "ink-soft",
-    "accent", "accent-strong", "warm", "good", "bad",
-    "bg", "bg-elev", "bg-card", "line-strong",
-    "ink-mute", "ink-dim", "accent-dim", "plasma", "warn",
-  ];
+  it("returns 'var(--color-surface)' for 'surface'", () => {
+    expect(cssVar("surface")).toBe("var(--color-surface)");
+  });
 
-  it("returns a CSS variable string for every colour token", () => {
-    for (const token of tokens) {
-      expect(cssVar(token)).toBe(`var(--color-${token})`);
+  it("returns 'var(--color-ink)' for 'ink'", () => {
+    expect(cssVar("ink")).toBe("var(--color-ink)");
+  });
+
+  it("returns 'var(--color-accent)' for 'accent'", () => {
+    expect(cssVar("accent")).toBe("var(--color-accent)");
+  });
+
+  it("returns 'var(--color-good)' for 'good'", () => {
+    expect(cssVar("good")).toBe("var(--color-good)");
+  });
+
+  it("returns 'var(--color-bad)' for 'bad'", () => {
+    expect(cssVar("bad")).toBe("var(--color-bad)");
+  });
+
+  it("returns 'var(--color-warn)' for 'warn'", () => {
+    expect(cssVar("warn")).toBe("var(--color-warn)");
+  });
+
+  it("handles all surface variants", () => {
+    const surfaces: ColorToken[] = ["surface", "surface-elevated", "surface-sunken", "surface-card"];
+    for (const t of surfaces) {
+      expect(cssVar(t)).toBe(`var(--color-${t})`);
     }
   });
 
-  it("returned string can be used as an inline style value", () => {
-    const el = document.createElement("div");
-    el.style.color = cssVar("ink");
-    el.style.backgroundColor = cssVar("surface");
-    expect(el.style.color).toBe("var(--color-ink)");
-    expect(el.style.backgroundColor).toBe("var(--color-surface)");
+  it("handles all ink variants", () => {
+    const inks: ColorToken[] = ["ink", "ink-mid", "ink-soft", "ink-mute", "ink-dim"];
+    for (const t of inks) {
+      expect(cssVar(t)).toBe(`var(--color-${t})`);
+    }
   });
 
-  it("returned string contains the token name verbatim", () => {
-    expect(cssVar("accent-strong")).toContain("accent-strong");
-    expect(cssVar("ink-dim")).toContain("ink-dim");
+  it("handles always-dark tokens", () => {
+    const dark: ColorToken[] = ["bg", "bg-elev", "bg-card", "line-strong", "accent-dim", "plasma"];
+    for (const t of dark) {
+      expect(cssVar(t)).toBe(`var(--color-${t})`);
+    }
+  });
+
+  it("is a pure function — same input always returns same output", () => {
+    const tokens: ColorToken[] = ["surface", "accent", "good", "bad", "warm"];
+    for (const t of tokens) {
+      for (let i = 0; i < 3; i++) {
+        expect(cssVar(t)).toBe(`var(--color-${t})`);
+      }
+    }
   });
 });

canvas/src/lib/__tests__/theme-provider.test.tsx

@@ -1,134 +0,0 @@
-// @vitest-environment jsdom
-/**
- * Tests for ThemeProvider and useTheme.
- *
- * Uses renderHook so useEffect fires before assertions.
- * matchMedia is stubbed via Object.defineProperty in beforeEach.
- */
-import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
-import { render, renderHook, cleanup, act } from "@testing-library/react";
-import React from "react";
-import { ThemeProvider, useTheme } from "../theme-provider";
-
-afterEach(cleanup);
-
-function makeMatcher(prefersDark: boolean) {
-  return {
-    matches: prefersDark,
-    media: "(prefers-color-scheme: dark)",
-    onchange: null,
-    addListener: vi.fn(),
-    removeListener: vi.fn(),
-    addEventListener: vi.fn(),
-    removeEventListener: vi.fn(),
-    dispatchEvent: vi.fn(),
-  };
-}
-
-beforeEach(() => {
-  Object.defineProperty(window, "matchMedia", {
-    writable: true,
-    configurable: true,
-    value: vi.fn().mockImplementation(() => makeMatcher(false)),
-  });
-});
-
-afterEach(() => {
-  vi.restoreAllMocks();
-});
-
-describe("useTheme", () => {
-  it("returns noopTheme when no provider is in the tree", () => {
-    const { result } = renderHook(() => useTheme());
-    expect(result.current).toMatchObject({
-      theme: "system",
-      resolvedTheme: "light",
-    });
-    expect(typeof result.current.setTheme).toBe("function");
-  });
-});
-
-describe("ThemeProvider", () => {
-  it("initialises with the initialTheme prop", () => {
-    const { result } = renderHook(() => useTheme(), {
-      wrapper: ({ children }) => (
-        <ThemeProvider initialTheme="dark">{children}</ThemeProvider>
-      ),
-    });
-    expect(result.current).toMatchObject({
-      theme: "dark",
-      resolvedTheme: "dark",
-    });
-    expect(document.documentElement.dataset.theme).toBe("dark");
-  });
-
-  it("reflects system preference when theme=system", () => {
-    Object.defineProperty(window, "matchMedia", {
-      writable: true,
-      configurable: true,
-      value: vi.fn().mockImplementation(() => makeMatcher(true)),
-    });
-
-    const { result } = renderHook(() => useTheme(), {
-      wrapper: ({ children }) => (
-        <ThemeProvider initialTheme="system">{children}</ThemeProvider>
-      ),
-    });
-    expect(result.current).toMatchObject({
-      theme: "system",
-      resolvedTheme: "dark",
-    });
-    expect(document.documentElement.dataset.theme).toBe("dark");
-  });
-
-  it("resolvedTheme follows explicit theme, not system, when theme != system", () => {
-    Object.defineProperty(window, "matchMedia", {
-      writable: true,
-      configurable: true,
-      value: vi.fn().mockImplementation(() => makeMatcher(true)),
-    });
-
-    const { result } = renderHook(() => useTheme(), {
-      wrapper: ({ children }) => (
-        <ThemeProvider initialTheme="light">{children}</ThemeProvider>
-      ),
-    });
-    expect(result.current).toMatchObject({
-      theme: "light",
-      resolvedTheme: "light",
-    });
-    expect(document.documentElement.dataset.theme).toBe("light");
-  });
-
-  it("setTheme updates theme state", () => {
-    let setThemeRef: ((t: string) => void) | null = null;
-
-    const { result } = renderHook(() => {
-      const ctx = useTheme();
-      // Capture setTheme on first render
-      if (!setThemeRef) setThemeRef = ctx.setTheme;
-      return ctx;
-    }, {
-      wrapper: ({ children }) => (
-        <ThemeProvider initialTheme="light">{children}</ThemeProvider>
-      ),
-    });
-
-    expect(result.current.theme).toBe("light");
-
-    act(() => { setThemeRef!("dark"); });
-    expect(result.current.theme).toBe("dark");
-    expect(document.documentElement.dataset.theme).toBe("dark");
-  });
-
-  it("sets document.documentElement.dataset.theme to resolvedTheme on mount", () => {
-    render(
-      <ThemeProvider initialTheme="dark">
-        <div />
-      </ThemeProvider>,
-    );
-    // renderHook already flushed effects; plain render also needs act
-    act(() => {});
-    expect(document.documentElement.dataset.theme).toBe("dark");
-  });
-});

canvas/src/store/__tests__/socket.test.ts

@@ -21,22 +21,12 @@ vi.mock("../canvas", () => ({
 class MockWebSocket {
   static instances: MockWebSocket[] = [];
 
-  // Mirror the real WebSocket readyState constants — socket.ts's wake
-  // path reads WebSocket.OPEN / WebSocket.CONNECTING and this.ws.readyState.
-  static readonly CONNECTING = 0;
-  static readonly OPEN = 1;
-  static readonly CLOSING = 2;
-  static readonly CLOSED = 3;
-
   url: string;
   onopen: (() => void) | null = null;
   onmessage: ((event: { data: string }) => void) | null = null;
   onclose: (() => void) | null = null;
   onerror: (() => void) | null = null;
   closeCallCount = 0;
-  // Starts OPEN once triggerOpen runs; tests flip this to simulate a
-  // mobile background-suspend that left a dead/half-open socket.
-  readyState = MockWebSocket.CONNECTING;
 
   constructor(url: string) {
     this.url = url;
@@ -45,12 +35,10 @@ class MockWebSocket {
 
   close() {
     this.closeCallCount++;
-    this.readyState = MockWebSocket.CLOSED;
   }
 
   // Helpers to trigger events in tests
   triggerOpen() {
-    this.readyState = MockWebSocket.OPEN;
     this.onopen?.();
   }
 
@@ -71,51 +59,11 @@ class MockWebSocket {
   }
 }
 
-// ---------------------------------------------------------------------------
-// Minimal DOM stub (vitest environment is 'node' — no window/document).
-// socket.ts's wake-recovery attaches visibilitychange/pageshow/online/
-// focus listeners; under node it self-no-ops via a typeof guard, so to
-// exercise the path we inject just enough of window/document here, the
-// same way WebSocket is stubbed above. Kept tiny on purpose — a single
-// listener registry keyed by event name, plus a settable
-// visibilityState.
-// ---------------------------------------------------------------------------
-
-interface FakeTarget {
-  _l: Record<string, Array<() => void>>;
-  addEventListener: (type: string, fn: () => void) => void;
-  removeEventListener: (type: string, fn: () => void) => void;
-  dispatch: (type: string) => void;
-}
-
-function makeFakeTarget(): FakeTarget {
-  const l: Record<string, Array<() => void>> = {};
-  return {
-    _l: l,
-    addEventListener(type, fn) {
-      (l[type] ||= []).push(fn);
-    },
-    removeEventListener(type, fn) {
-      l[type] = (l[type] || []).filter((f) => f !== fn);
-    },
-    dispatch(type) {
-      for (const fn of l[type] || []) fn();
-    },
-  };
-}
-
-const fakeWindow = makeFakeTarget();
-const fakeDocument = Object.assign(makeFakeTarget(), {
-  visibilityState: "visible" as string,
-});
-(globalThis as unknown as Record<string, unknown>).window = fakeWindow;
-(globalThis as unknown as Record<string, unknown>).document = fakeDocument;
-
 // Install mock WebSocket globally before importing socket module
 (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";
 
 // ---------------------------------------------------------------------------
@@ -380,153 +328,6 @@ describe("WebSocket onerror", () => {
   });
 });
 
-// ---------------------------------------------------------------------------
-// Wake recovery — mobile background-suspend regression (mobile chat not
-// updating in real time until refresh). Simulates: connect → open →
-// the OS freezes the page and silently kills the WS WITHOUT firing
-// onclose → user returns (visibilitychange / pageshow / online /
-// focus) → assert the dead socket is replaced AND, on the new socket's
-// open, the resume signal fires so chat history back-fills the missed
-// AGENT_MESSAGE / A2A_RESPONSE events.
-// ---------------------------------------------------------------------------
-
-import {
-  subscribeSocketResume,
-  _resetSocketResumeListenersForTests,
-} from "../socket-events";
-
-describe("wake recovery (mobile background-suspend)", () => {
-  beforeEach(() => {
-    _resetSocketResumeListenersForTests();
-    fakeDocument.visibilityState = "visible";
-  });
-
-  function suspendKill(ws: MockWebSocket) {
-    // Mobile background-suspend: the OS tore the transport down but the
-    // page was frozen so onclose never ran. The socket object survives
-    // with a CLOSED readyState and no reconnect was scheduled.
-    ws.readyState = MockWebSocket.CLOSED;
-  }
-
-  it("reconnects on visibilitychange when the socket was silently killed", () => {
-    connectSocket();
-    const ws = getLastWS();
-    ws.triggerOpen();
-    expect(MockWebSocket.instances).toHaveLength(1);
-
-    suspendKill(ws);
-    fakeDocument.dispatch("visibilitychange");
-
-    // A fresh socket must have been created — the stale one is not
-    // reused.
-    expect(MockWebSocket.instances.length).toBeGreaterThan(1);
-  });
-
-  it("does NOT reconnect on visibilitychange while the socket is still healthy", () => {
-    connectSocket();
-    const ws = getLastWS();
-    ws.triggerOpen();
-    expect(MockWebSocket.instances).toHaveLength(1);
-
-    // Healthy OPEN socket + a spurious visibilitychange (e.g. quick tab
-    // peek that never actually suspended) → no churn.
-    fakeDocument.dispatch("visibilitychange");
-    expect(MockWebSocket.instances).toHaveLength(1);
-  });
-
-  it("ignores visibilitychange when the page is hidden (the hide transition)", () => {
-    connectSocket();
-    const ws = getLastWS();
-    ws.triggerOpen();
-    suspendKill(ws);
-
-    fakeDocument.visibilityState = "hidden";
-    fakeDocument.dispatch("visibilitychange");
-    // Hidden → must not reconnect (would defeat the purpose; we only
-    // re-arm when the user is actually looking at the page again).
-    expect(MockWebSocket.instances).toHaveLength(1);
-  });
-
-  it.each(["pageshow", "online", "focus"])(
-    "reconnects on window '%s' after a silent kill",
-    (evt) => {
-      connectSocket();
-      const ws = getLastWS();
-      ws.triggerOpen();
-      suspendKill(ws);
-
-      fakeWindow.dispatch(evt);
-      expect(MockWebSocket.instances.length).toBeGreaterThan(1);
-    },
-  );
-
-  it("emits the resume signal once the recovered socket re-opens (so chat back-fills missed messages)", () => {
-    const onResume = vi.fn();
-    const unsub = subscribeSocketResume(onResume);
-
-    connectSocket();
-    const ws1 = getLastWS();
-    ws1.triggerOpen();
-    // First open must NOT fire resume — the mount-time chat-history
-    // fetch already covers the initial load.
-    expect(onResume).not.toHaveBeenCalled();
-
-    // Background-suspend silently kills the socket, then the user
-    // returns.
-    suspendKill(ws1);
-    fakeDocument.dispatch("visibilitychange");
-
-    // The wake handler force-reconnected; the new socket completing its
-    // handshake is what signals "we recovered from a gap — re-fetch".
-    const ws2 = getLastWS();
-    expect(ws2).not.toBe(ws1);
-    ws2.triggerOpen();
-
-    expect(onResume).toHaveBeenCalledTimes(1);
-    unsub();
-  });
-
-  it("does not emit resume on the very first connect", () => {
-    const onResume = vi.fn();
-    const unsub = subscribeSocketResume(onResume);
-    connectSocket();
-    getLastWS().triggerOpen();
-    expect(onResume).not.toHaveBeenCalled();
-    unsub();
-  });
-
-  it("emits resume after an ordinary onclose-driven reconnect too (desktop path unchanged)", () => {
-    const onResume = vi.fn();
-    const unsub = subscribeSocketResume(onResume);
-
-    connectSocket();
-    const ws1 = getLastWS();
-    ws1.triggerOpen();
-    // Ordinary network drop — onclose fires normally.
-    ws1.triggerClose();
-    vi.advanceTimersByTime(1100); // past the 1s backoff
-    const ws2 = getLastWS();
-    expect(ws2).not.toBe(ws1);
-    ws2.triggerOpen();
-
-    expect(onResume).toHaveBeenCalledTimes(1);
-    unsub();
-  });
-
-  it("detaches wake listeners on disconnect (no reconnect after teardown)", () => {
-    connectSocket();
-    const ws = getLastWS();
-    ws.triggerOpen();
-    disconnectSocket();
-
-    const countAfterDisconnect = MockWebSocket.instances.length;
-    // A wake event after teardown must be inert.
-    fakeDocument.dispatch("visibilitychange");
-    fakeWindow.dispatch("focus");
-    expect(MockWebSocket.instances.length).toBe(countAfterDisconnect);
-  });
-});
-
 // ---------------------------------------------------------------------------
 // Health check (startHealthCheck / stopHealthCheck via onopen / disconnect)
 // ---------------------------------------------------------------------------
@@ -615,3 +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);
+  });
+});

canvas/src/store/socket-events.ts

@@ -61,53 +61,3 @@ export function subscribeSocketEvents(listener: Listener): () => void {
 export function _resetSocketEventListenersForTests(): void {
   listeners.clear();
 }
-
-// ---------------------------------------------------------------------------
-// Socket-resume signal
-// ---------------------------------------------------------------------------
-//
-// Fired by the ReconnectingSocket when the WS comes back up AFTER having
-// been down (drop, or a mobile-browser background-suspend that silently
-// killed the socket while the page was frozen). Distinct from the raw
-// event bus above: while the socket was dead the page missed every
-// AGENT_MESSAGE / A2A_RESPONSE, and the store's rehydrate() only re-pulls
-// /workspaces status — it does NOT back-fill chat messages. Components
-// that render a live message thread (desktop ChatTab + MobileChat, both
-// via useChatHistory) subscribe here to re-fetch their history on resume
-// so missed agent replies appear without the user having to navigate
-// away+back or hard-refresh. Shared by desktop and mobile — the recovery
-// is in the singleton socket, not forked per-surface.
-
-type ResumeListener = () => void;
-
-const resumeListeners = new Set<ResumeListener>();
-
-/** Notify every resume subscriber that the socket just recovered from a
- *  down period. Called by ReconnectingSocket.onopen, but only when the
- *  open follows a prior loss (not the very first connect — the initial
- *  mount-time history fetch already covers that). */
-export function emitSocketResume(): void {
-  for (const listener of resumeListeners) {
-    try {
-      listener();
-    } catch (err) {
-      if (typeof console !== "undefined") {
-        console.error("socket-resume listener threw:", err);
-      }
-    }
-  }
-}
-
-/** Register a resume subscriber. Returns an unsubscribe function the
- *  caller must invoke from its effect cleanup. */
-export function subscribeSocketResume(listener: ResumeListener): () => void {
-  resumeListeners.add(listener);
-  return () => {
-    resumeListeners.delete(listener);
-  };
-}
-
-/** Test-only: drop all resume subscribers. */
-export function _resetSocketResumeListenersForTests(): void {
-  resumeListeners.clear();
-}

canvas/src/store/socket.ts

@@ -1,6 +1,6 @@
 import { useCanvasStore } from "./canvas";
 import { deriveWsBaseUrl } from "@/lib/ws-url";
-import { emitSocketEvent, emitSocketResume } from "./socket-events";
+import { emitSocketEvent } from "./socket-events";
 
 // If explicit WS_URL is set, use it as-is (may include custom path).
 // Otherwise derive base + append /ws.
@@ -98,107 +98,9 @@ class ReconnectingSocket {
   // caller can fire-and-forget without coordinating.
   private rehydrateInFlight: Promise<void> | null = null;
   private rehydrateDedup = new RehydrateDedup(REHYDRATE_DEDUP_WINDOW_MS);
-  // True once any onopen has fired. Gates the resume signal so the very
-  // first connect doesn't fire it (the mount-time chat-history fetch
-  // already covers the initial load — a resume here would be a wasted
-  // duplicate). Set on the first successful open and stays true.
-  private everConnected = false;
-  // True between a loss (onclose / wake-detected stale socket) and the
-  // next successful onopen. Only when this is set does onopen emit the
-  // resume signal — i.e. we recovered from a real gap during which
-  // AGENT_MESSAGE / A2A_RESPONSE events may have been missed.
-  private wasDown = false;
-  // Bound wake handler. iOS Safari / Chrome-mobile freeze the page and
-  // its timers when the tab is backgrounded or the device locks, and
-  // tear the WS down WITHOUT reliably firing onclose before the freeze.
-  // On thaw nothing re-arms: onclose never ran so no reconnect was
-  // scheduled, and the health-check / fallback-poll intervals were
-  // suspended. The socket is silently dead until a manual refresh. This
-  // handler force-reconnects on any wake signal when the socket isn't
-  // healthy. Stored so disconnect() can detach the listeners.
-  private onWake: (() => void) | null = null;
 
   constructor(url: string) {
     this.url = url;
-    this.installWakeListeners();
-  }
-
-  /** Attach page-lifecycle listeners that force a reconnect when the
-   *  page returns to the foreground / regains connectivity and the
-   *  socket is not OPEN. Shared by desktop and mobile — desktop rarely
-   *  hits the stale-socket path (its onclose fires promptly) so this is
-   *  effectively a no-op there, while mobile depends on it because the
-   *  background-suspend kills the socket without an onclose. */
-  private installWakeListeners() {
-    if (typeof window === "undefined" || typeof document === "undefined") {
-      return;
-    }
-    const wake = () => {
-      if (this.disposed) return;
-      // Only act on a visible page — visibilitychange also fires on the
-      // hide transition, which we must ignore (closing here would defeat
-      // the point).
-      if (
-        typeof document.visibilityState === "string" &&
-        document.visibilityState !== "visible"
-      ) {
-        return;
-      }
-      // Healthy socket → nothing to do. A stale/half-open socket on
-      // mobile reports CLOSED or CLOSING (the OS tore the transport
-      // down); CONNECTING is also unhealthy from the user's POV but a
-      // reconnect attempt is already in flight, so leave it.
-      const live =
-        this.ws !== null &&
-        (this.ws.readyState === WebSocket.OPEN ||
-          this.ws.readyState === WebSocket.CONNECTING);
-      if (live) return;
-      // Tear down any zombie and reconnect immediately. Mark wasDown so
-      // the subsequent onopen emits the resume signal and chat threads
-      // back-fill the messages missed while frozen.
-      this.wasDown = true;
-      this.forceReconnect();
-    };
-    this.onWake = wake;
-    document.addEventListener("visibilitychange", wake);
-    window.addEventListener("pageshow", wake);
-    window.addEventListener("online", wake);
-    window.addEventListener("focus", wake);
-  }
-
-  private removeWakeListeners() {
-    if (!this.onWake) return;
-    if (typeof window !== "undefined" && typeof document !== "undefined") {
-      document.removeEventListener("visibilitychange", this.onWake);
-      window.removeEventListener("pageshow", this.onWake);
-      window.removeEventListener("online", this.onWake);
-      window.removeEventListener("focus", this.onWake);
-    }
-    this.onWake = null;
-  }
-
-  /** Detach the current (presumed dead/stale) socket without routing
-   *  through its onclose, cancel any pending backoff timer, and
-   *  reconnect now. Used by the wake path: the browser already killed
-   *  the transport, so the exponential backoff that onclose would have
-   *  scheduled is both absent and undesirable — the user is looking at
-   *  the page and wants it live immediately. */
-  private forceReconnect() {
-    if (this.disposed) return;
-    if (this.reconnectTimer) {
-      clearTimeout(this.reconnectTimer);
-      this.reconnectTimer = null;
-    }
-    if (this.ws) {
-      this.ws.onopen = null;
-      this.ws.onmessage = null;
-      this.ws.onclose = null;
-      this.ws.onerror = null;
-      try { this.ws.close(); } catch { /* noop */ }
-      this.ws = null;
-    }
-    this.attempt = 0;
-    this.connect();
   }
 
   connect() {
@@ -230,18 +132,6 @@ class ReconnectingSocket {
       this.stopFallbackPoll();
       this.rehydrate();
       this.startHealthCheck();
-      // If this open follows a real loss (drop, or a mobile background-
-      // suspend that the wake handler recovered from), signal resume so
-      // live message threads re-fetch the AGENT_MESSAGE / A2A_RESPONSE
-      // history they missed while the socket was dead — rehydrate()
-      // above only refreshes /workspaces status, not chat. Gate on
-      // everConnected so the very first open (covered by the mount-time
-      // history fetch) doesn't fire a redundant resume.
-      if (this.everConnected && this.wasDown) {
-        emitSocketResume();
-      }
-      this.everConnected = true;
-      this.wasDown = false;
     };
 
     ws.onmessage = (event) => {
@@ -267,11 +157,6 @@ class ReconnectingSocket {
       // corresponds to the WS we just tore down (prevents a stale
       // onclose from a zombie socket from re-arming the loop).
       if (this.disposed || this.ws !== ws) return;
-      // We had a live socket and lost it — mark down so the next onopen
-      // emits the resume signal and chat threads back-fill missed
-      // messages. (The wake path also sets this; setting it here covers
-      // the ordinary network-drop case.)
-      this.wasDown = true;
       this.stopHealthCheck();
       useCanvasStore.getState().setWsStatus("connecting");
       this.startFallbackPoll();
@@ -362,7 +247,6 @@ class ReconnectingSocket {
 
   disconnect() {
     this.disposed = true;
-    this.removeWakeListeners();
     this.stopHealthCheck();
     this.stopFallbackPoll();
     if (this.reconnectTimer) {
@@ -384,6 +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 {
@@ -422,11 +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() {
@@ -434,4 +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();
 }

canvas/src/styles/settings-panel.css

@@ -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; }
 

review-pr3029-pr3033.md

@@ -1,130 +0,0 @@
-# 5-Axis Review: PR #3029 (fix #2989) + PR #3033 (docs refresh)
-
-**Reviewer:** Kimi / Engineer-A  
-**Date:** 2026-05-31  
-**Scope:** Local review (CR2 auth-down, filling review gap per PM dispatch)
-
----
-
-## PR #3029 — CP orphan sweeper + registry prefix abstraction
-
-### Correctness ✅ (with 1 semantic conflict to resolve)
-
-**cp_orphan_sweeper.go** — The deprovision split-write race fix is sound:
-- SELECT `status='removed' AND instance_id IS NOT NULL AND instance_id != ''` correctly targets leaked EC2s.
-- Stop → clear instance_id is idempotent; on Stop failure the row stays targeted for retry.
-- `ORDER BY updated_at DESC` + `LIMIT $1` + `UPDATE updated_at = now()` creates fair round-robin drain across cycles.
-- `supervised.RunWithRecover` wiring in `cmd/server/main.go` mirrors the Docker sweeper pattern.
-
-**provisioner/registry.go** — Clean env-driven prefix abstraction:
-- `RegistryPrefix()` respects `MOLECULE_IMAGE_REGISTRY` override; falls back to GHCR OSS default.
-- `RuntimeImage()` returns `""` for unknown runtimes, forcing explicit fallback at call sites.
-- `computeRuntimeImages()` runs at init; captures prefix active at boot.
-
-** provisioner.go migration** — Hardcoded map → `computeRuntimeImages()` is a safe refactor; no behavioral change for OSS default.
-
-**admin_workspace_images.go** — `TemplateImageRef()` now uses `provisioner.RegistryPrefix()`; keeps admin ops and provisioner pulls consistent.
-
-### Security ✅
-
-- Sweeper SQL has no user-input surface; parameters are internal LIMIT constant and DB-generated IDs.
-- `RegistryPrefix()` reads env only; comment correctly notes it is deploy-time trusted (operator-set, not user-supplied).
-- No new secrets, auth tokens, or credential exposure.
-
-### Performance ✅
-
-- 60s tick / 30s deadline / LIMIT 100 is conservative and safe.
-- Sequential Stop calls share the 30s parent context; with typical CP DELETE latency (<1s), 100 orphans finish well within budget.
-- If CP is degraded, deadline expires, UPDATEs don't fire, and next cycle retries — no stampede.
-
-### Style / Readability ✅
-
-- Excellent docstrings; the `#2989` race narrative is clearly documented for future maintainers.
-- `CPOrphanReaper` interface is minimal and testable.
-- Nil-reaper and nil-DB guards follow existing patterns.
-- One minor nit: `cpSweepOnce` could return `[]string` of processed IDs to make post-hoc assertions easier, but the fake-reaper test pattern works fine as-is.
-
-### Tests ✅ (excellent coverage)
-
-| Scenario | Covered |
-|---|---|
-| Happy path: Stop succeeds, instance_id cleared | ✅ |
-| Stop fails, instance_id retained for retry | ✅ |
-| Empty result set (steady state) | ✅ |
-| Multiple orphans, partial failure, others proceed | ✅ |
-| DB query error (transient) | ✅ |
-| UPDATE error after Stop success (logs, continues) | ✅ |
-| Nil db.DB (defensive boot safety) | ✅ |
-| Nil reaper (disabled, no goroutine leak) | ✅ |
-| Boot sweep + tick cadence + ctx cancel | ✅ |
-| Registry prefix default / env override / empty env | ✅ |
-| Runtime image format for all known runtimes | ✅ |
-| Unknown runtime returns `""` | ✅ |
-| Registry override applies to ALL runtimes | ✅ |
-| Alphabetical order pin | ✅ |
-
-**All tests pass:**
-```
-ok  github.com/.../internal/registry     0.107s  (9/9 CP sweeper tests)
-ok  github.com/.../internal/provisioner  0.009s  (7/7 registry tests)
-```
-
-### ⚠️ BLOCKER: Semantic conflict with PR #3033
-
-`registry.go` adds `"codex"` to `knownRuntimes`, making **9** production runtimes:
-```go
-knownRuntimes = []string{
-    "autogen", "claude-code", "codex", "crewai", "deepagents",
-    "gemini-cli", "hermes", "langgraph", "openclaw",
-}
-```
-
-PR #3033 updates the README to claim **eight** production runtimes and explicitly lists:
-> Claude Code, Hermes, Gemini CLI, LangGraph, DeepAgents, CrewAI, AutoGen, OpenClaw
-
-`codex` is absent from the README compatibility table, the "What Ships In main" section, and the architecture diagram list. After both PRs merge, the code will support 9 runtimes but the docs will claim 8 — a public-facing drift.
-
-**Fix path:** Add `codex` to the README runtime list in PR #3033 (or a fast-follow) so the count and table stay accurate. `codex` already exists in `manifest.json` and has a template repo, so it is legitimate to list as "shipping on main."
-
----
-
-## PR #3033 — Docs refresh (README + branding assets)
-
-### Correctness ✅ (with 1 semantic drift pending)
-
-- Terminology standardization ("adapters" → "runtimes") is correct and consistent with platform usage.
-- Deploy buttons updated from `molecule-monorepo` → `molecule-core`.
-- Canvas v4, Memory v2, SaaS surface, RFC #2967 mentions are all factually accurate.
-- **Missing:** `codex` runtime (see blocker above).
-
-### Security ✅
-
-- SVG assets are static branding; no scripts, no external references beyond the existing `<style>` media query.
-- No auth or credential surface touched.
-
-### Performance N/A
-
-- Docs-only; no runtime impact.
-
-### Style / Readability ✅
-
-- warm-paper theme description is concise and helpful.
-- Architecture diagram update (Docker → EC2 + SSM, KMS, SaaS CP) is accurate.
-- Quick Start clone URL fixed.
-
-### Tests N/A
-
-- No code changes; no test delta.
-
----
-
-## Summary
-
-| PR | Verdict | Action needed |
-|---|---|---|
-| #3029 | **Approve with nit** | Merge-ready after confirming #3033 (or follow-up) adds `codex` to README runtime list. |
-| #3033 | **Approve with blocker** | Add `codex` to the 8-runtimes list (making 9) and to the compatibility table before merge. |
-
-**Risk if both merge as-is:** Public docs understate runtime count by 1; operators reading README may think `codex` is not supported when the provisioner already knows about it.
-
-**Recommended merge order:** #3029 first (adds runtime support), then #3033 with `codex` line added (docs catch up).

scripts/build_runtime_package.py

@@ -58,11 +58,11 @@ TOP_LEVEL_MODULES = {
     "a2a_response",
     "a2a_tools",
     "a2a_tools_delegation",
+    "a2a_tools_identity",
     "a2a_tools_inbox",
     "a2a_tools_memory",
     "a2a_tools_messaging",
     "a2a_tools_rbac",
-    "a2a_tools_identity",
     "adapter_base",
     "agent",
     "agents_md",
@@ -311,8 +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
 
@@ -336,8 +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
 
@@ -375,8 +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)),

workspace-server/cmd/server/cp_config.go

@@ -60,8 +60,7 @@ func refreshEnvFromCP() error {
 	req.Header.Set("Authorization", "Bearer "+adminToken)
 	req.Header.Set("X-Molecule-Org-Id", orgID)
 
-	client := &http.Client{Timeout: 10 * time.Second}
-	resp, err := client.Do(req)
+	resp, err := http.DefaultClient.Do(req)
 	if err != nil {
 		return fmt.Errorf("do request: %w", err)
 	}

workspace-server/cmd/t4-contract-dump/main.go

@@ -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)
+	}
+}

workspace-server/internal/bundle/importer.go

@@ -3,7 +3,6 @@ package bundle
 import (
 	"context"
 	"fmt"
-	"log"
 	"strings"
 
 	"github.com/Molecule-AI/molecule-monorepo/platform/internal/db"
@@ -87,20 +86,13 @@ func Import(
 			// PluginsPath set by caller if available
 		}
 		go func() {
-			defer func() {
-				if r := recover(); r != nil {
-					log.Printf("bundle/importer: PANIC during provision start for %s: %v", wsID, r)
-				}
-			}()
 			provCtx, cancel := context.WithTimeout(context.Background(), provisioner.ProvisionTimeout)
 			defer cancel()
 			url, err := prov.Start(provCtx, cfg)
 			if err != nil {
 				markFailed(provCtx, wsID, broadcaster, err)
 			} else if url != "" {
-				if _, dbErr := db.DB.ExecContext(provCtx, `UPDATE workspaces SET url = $1 WHERE id = $2`, url, wsID); dbErr != nil {
-					log.Printf("bundle import: failed to update workspace URL for %s: %v", wsID, dbErr)
-				}
+				db.DB.ExecContext(provCtx, `UPDATE workspaces SET url = $1 WHERE id = $2`, url, wsID)
 			}
 		}()
 	}
@@ -147,16 +139,12 @@ func markFailed(ctx context.Context, wsID string, broadcaster *events.Broadcaste
 	// markProvisionFailed in workspace-server/internal/handlers/
 	// workspace_provision_shared.go.
 	msg := err.Error()
-	if _, dbErr := db.DB.ExecContext(ctx,
+	db.DB.ExecContext(ctx,
 		`UPDATE workspaces SET status = $1, last_sample_error = $2, updated_at = now() WHERE id = $3`,
-		models.StatusFailed, msg, wsID); dbErr != nil {
-		log.Printf("bundle import: failed to mark workspace %s failed: %v", wsID, dbErr)
-	}
-	if bcErr := broadcaster.RecordAndBroadcast(ctx, string(events.EventWorkspaceProvisionFailed), wsID, map[string]interface{}{
+		models.StatusFailed, msg, wsID)
+	broadcaster.RecordAndBroadcast(ctx, string(events.EventWorkspaceProvisionFailed), wsID, map[string]interface{}{
 		"error": msg,
-	}); bcErr != nil {
-		log.Printf("bundle import: failed to broadcast provision failed for %s: %v", wsID, bcErr)
-	}
+	})
 }
 
 func nilIfEmpty(s string) interface{} {

workspace-server/internal/channels/manager.go

@@ -375,25 +375,21 @@ func (m *Manager) HandleInbound(ctx context.Context, ch ChannelRow, msg *Inbound
 
 	// Update stats in DB
 	if db.DB != nil {
-		if _, err := db.DB.ExecContext(ctx, `
+		db.DB.ExecContext(ctx, `
 			UPDATE workspace_channels
 			SET last_message_at = now(), message_count = message_count + 1, updated_at = now()
 			WHERE id = $1
-		`, ch.ID); err != nil {
-			log.Printf("Channels: failed to update inbound stats for channel %s: %v", ch.ID, err)
-		}
+		`, ch.ID)
 	}
 
 	// Broadcast event
 	if m.broadcaster != nil {
-		if err := m.broadcaster.RecordAndBroadcast(ctx, string(events.EventChannelMessage), ch.WorkspaceID, map[string]interface{}{
+		m.broadcaster.RecordAndBroadcast(ctx, string(events.EventChannelMessage), ch.WorkspaceID, map[string]interface{}{
 			"channel_id":   ch.ID,
 			"channel_type": ch.ChannelType,
 			"username":     msg.Username,
 			"direction":    "inbound",
-		}); err != nil {
-			log.Printf("Channels: failed to broadcast inbound event: %v", err)
-		}
+		})
 	}
 
 	return nil
@@ -424,23 +420,19 @@ func (m *Manager) SendOutbound(ctx context.Context, channelID string, text strin
 	}
 
 	if db.DB != nil {
-		if _, err := db.DB.ExecContext(ctx, `
+		db.DB.ExecContext(ctx, `
 			UPDATE workspace_channels
 			SET last_message_at = now(), message_count = message_count + 1, updated_at = now()
 			WHERE id = $1
-		`, channelID); err != nil {
-			log.Printf("Channels: failed to update outbound stats for channel %s: %v", channelID, err)
-		}
+		`, channelID)
 	}
 
 	if m.broadcaster != nil {
-		if err := m.broadcaster.RecordAndBroadcast(ctx, string(events.EventChannelMessage), ch.WorkspaceID, map[string]interface{}{
+		m.broadcaster.RecordAndBroadcast(ctx, string(events.EventChannelMessage), ch.WorkspaceID, map[string]interface{}{
 			"channel_id":   ch.ID,
 			"channel_type": ch.ChannelType,
 			"direction":    "outbound",
-		}); err != nil {
-			log.Printf("Channels: failed to broadcast outbound event: %v", err)
-		}
+		})
 	}
 
 	return nil
@@ -506,10 +498,7 @@ func (m *Manager) FetchWorkspaceChannelContext(ctx context.Context, workspaceID
 		return ""
 	}
 	var config map[string]interface{}
-	if err := json.Unmarshal(configJSON, &config); err != nil {
-		log.Printf("Channels: failed to unmarshal channel config: %v", err)
-		return ""
-	}
+	json.Unmarshal(configJSON, &config)
 	if err := DecryptSensitiveFields(config); err != nil {
 		return ""
 	}
@@ -566,12 +555,8 @@ func (m *Manager) loadChannel(ctx context.Context, channelID string) (ChannelRow
 	if err != nil {
 		return ch, fmt.Errorf("channel %s not found: %w", channelID, err)
 	}
-	if err := json.Unmarshal(configJSON, &ch.Config); err != nil {
-		return ch, fmt.Errorf("unmarshal channel %s config: %w", channelID, err)
-	}
-	if err := json.Unmarshal(allowedJSON, &ch.AllowedUsers); err != nil {
-		return ch, fmt.Errorf("unmarshal channel %s allowed_users: %w", channelID, err)
-	}
+	json.Unmarshal(configJSON, &ch.Config)
+	json.Unmarshal(allowedJSON, &ch.AllowedUsers)
 	// #319: decrypt bot_token / webhook_secret — SendOutbound and adapter
 	// methods downstream read them as plaintext strings.
 	if err := DecryptSensitiveFields(ch.Config); err != nil {

workspace-server/internal/channels/telegram.go

@@ -482,12 +482,10 @@ func (t *TelegramAdapter) StartPolling(ctx context.Context, config map[string]in
 				if apiErr.Code == 429 {
 					retryAfter := time.Duration(apiErr.RetryAfter) * time.Second
 					log.Printf("Channels: Telegram poll rate-limited, sleeping %s", retryAfter)
-					timer := time.NewTimer(retryAfter)
 					select {
 					case <-ctx.Done():
-						timer.Stop()
 						return nil
-					case <-timer.C:
+					case <-time.After(retryAfter):
 						continue
 					}
 				}
@@ -497,12 +495,10 @@ func (t *TelegramAdapter) StartPolling(ctx context.Context, config map[string]in
 				}
 			}
 			log.Printf("Channels: Telegram poll error: %v", err)
-			timer := time.NewTimer(telegramPollInterval)
 			select {
 			case <-ctx.Done():
-				timer.Stop()
 				return nil
-			case <-timer.C:
+			case <-time.After(telegramPollInterval):
 				continue
 			}
 		}
@@ -517,9 +513,7 @@ func (t *TelegramAdapter) StartPolling(ctx context.Context, config map[string]in
 
 				// Acknowledge the button press (removes loading spinner)
 				ackCfg := tgbotapi.NewCallback(cb.ID, "Received")
-				if _, err := bot.Send(ackCfg); err != nil {
-					log.Printf("telegram: failed to send callback ack: %v", err)
-				}
+				bot.Send(ackCfg)
 
 				// Update the message to show what was clicked
 				decision := "approved"
@@ -531,9 +525,7 @@ func (t *TelegramAdapter) StartPolling(ctx context.Context, config map[string]in
 					cb.Message.MessageID,
 					cb.Message.Text+"\n\n✅ CEO "+decision,
 				)
-				if _, err := bot.Send(editMsg); err != nil {
-					log.Printf("telegram: failed to send edit message: %v", err)
-				}
+				bot.Send(editMsg)
 
 				// Route the decision as an inbound message to the agent
 				inbound := &InboundMessage{

workspace-server/internal/events/types.go

@@ -41,9 +41,8 @@ type EventType string
 // scan-friendly as it grows.
 const (
 	// Chat / agent messaging — surfaces in canvas chat panels.
-	EventAgentMessage   EventType = "AGENT_MESSAGE"
-	EventA2AResponse   EventType = "A2A_RESPONSE"
-	EventUserMessage    EventType = "USER_MESSAGE"
+	EventAgentMessage EventType = "AGENT_MESSAGE"
+	EventA2AResponse  EventType = "A2A_RESPONSE"
 	EventActivityLogged EventType = "ACTIVITY_LOGGED"
 	EventChannelMessage EventType = "CHANNEL_MESSAGE"
 
@@ -96,7 +95,6 @@ const (
 var AllEventTypes = []EventType{
 	EventA2AResponse,
 	EventActivityLogged,
-	EventUserMessage,
 	EventAgentAssigned,
 	EventAgentCardUpdated,
 	EventAgentMessage,

workspace-server/internal/events/types_test.go

@@ -41,7 +41,6 @@ func TestAllEventTypes_IsSnapshot(t *testing.T) {
 		"DELEGATION_STATUS",
 		"EXTERNAL_CREDENTIALS_ROTATED",
 		"TASK_UPDATED",
-		"USER_MESSAGE",
 		"WORKSPACE_AWAITING_AGENT",
 		"WORKSPACE_DEGRADED",
 		"WORKSPACE_HEARTBEAT",

workspace-server/internal/handlers/a2a_proxy.go

@@ -932,12 +932,7 @@ func applyIdleTimeout(parent context.Context, b *events.Broadcaster, workspaceID
 	ctx, cancel := context.WithCancel(parent)
 	sub, unsub := b.SubscribeSSE(workspaceID)
 	go func() {
-		defer func() {
-			if r := recover(); r != nil {
-				log.Printf("a2a_proxy: PANIC in SSE idle watcher for %s: %v", workspaceID, r)
-			}
-			unsub()
-		}()
+		defer unsub()
 		timer := time.NewTimer(idle)
 		defer timer.Stop()
 		for {

workspace-server/internal/handlers/a2a_proxy_helpers.go

@@ -11,7 +11,6 @@ import (
 	"log"
 	"net/http"
 	"strconv"
-	"strings"
 	"time"
 
 	"github.com/Molecule-AI/molecule-monorepo/platform/internal/db"
@@ -345,19 +344,6 @@ func (h *WorkspaceHandler) logA2ASuccess(ctx context.Context, workspaceID, calle
 			"duration_ms":   durationMs,
 		})
 	}
-
-	// #228: fan user's own outbound message to all sessions of the workspace.
-	// When a canvas user sends a message (callerID == "" and method == "message/send"),
-	// the originating session already inserted it optimistically in useChatSend.
-	// Other sessions see nothing until a manual refresh — this broadcast closes
-	// that gap. The originating session collapses its optimistic copy via the
-	// 3-second appendMessageDeduped window (same role + content = deduped).
-	if callerID == "" && a2aMethod == "message/send" && statusCode < 400 {
-		userPayload := extractCanvasUserMessage(body)
-		if userPayload != nil {
-			h.broadcaster.BroadcastOnly(workspaceID, string(events.EventUserMessage), userPayload)
-		}
-	}
 }
 
 func nilIfEmpty(s string) *string {
@@ -407,110 +393,6 @@ func validateCallerToken(ctx context.Context, c *gin.Context, callerID string) e
 // matching (the wsauth errors are typed for the invalid case).
 var errInvalidCallerToken = errors.New("missing caller auth token")
 
-// extractCanvasUserMessage parses an A2A JSON-RPC request body and extracts
-// the user-authored text and attachments from a canvas-initiated message/send.
-// Returns nil when the body is not a canvas user message (empty, malformed,
-// or not a message/send from canvas). The returned payload is safe to pass
-// directly to BroadcastOnly — nil fields are omitted from JSON.
-func extractCanvasUserMessage(body []byte) map[string]interface{} {
-	if len(body) == 0 {
-		return nil
-	}
-	var top map[string]json.RawMessage
-	if err := json.Unmarshal(body, &top); err != nil {
-		return nil
-	}
-	// Only handle message/send from canvas
-	var method string
-	if err := json.Unmarshal(top["method"], &method); err != nil || method != "message/send" {
-		return nil
-	}
-	params, ok := top["params"]
-	if !ok {
-		return nil
-	}
-	var paramsMap map[string]json.RawMessage
-	if err := json.Unmarshal(params, &paramsMap); err != nil {
-		return nil
-	}
-	msgRaw, ok := paramsMap["message"]
-	if !ok {
-		return nil
-	}
-	var msg map[string]json.RawMessage
-	if err := json.Unmarshal(msgRaw, &msg); err != nil {
-		return nil
-	}
-
-	// role field: only broadcast user-role messages (canvas users)
-	var role string
-	if err := json.Unmarshal(msg["role"], &role); err != nil || role != "user" {
-		return nil
-	}
-
-	result := make(map[string]interface{})
-
-	// Extract messageId if present
-	var mid string
-	if err := json.Unmarshal(msg["messageId"], &mid); err == nil && mid != "" {
-		result["messageId"] = mid
-	}
-
-	// Extract text from parts — accumulate all text parts into a single string
-	var parts []json.RawMessage
-	if err := json.Unmarshal(msg["parts"], &parts); err == nil {
-		var texts []string
-		var fileAttachments []map[string]interface{}
-		for _, pRaw := range parts {
-			var p map[string]json.RawMessage
-			if err := json.Unmarshal(pRaw, &p); err != nil {
-				continue
-			}
-			var t string
-			if err := json.Unmarshal(p["text"], &t); err == nil && t != "" {
-				texts = append(texts, t)
-			}
-			var fileRaw json.RawMessage
-			if err := json.Unmarshal(p["file"], &fileRaw); err == nil && fileRaw != nil {
-				var f map[string]json.RawMessage
-				if err := json.Unmarshal(fileRaw, &f); err == nil {
-					att := make(map[string]interface{})
-					var s string
-					if err := json.Unmarshal(f["uri"], &s); err == nil {
-						att["uri"] = s
-					}
-					if err := json.Unmarshal(f["name"], &s); err == nil {
-						att["name"] = s
-					}
-					if err := json.Unmarshal(f["mimeType"], &s); err == nil {
-						att["mimeType"] = s
-					}
-					var n float64
-					if err := json.Unmarshal(f["size"], &n); err == nil {
-						att["size"] = n
-					}
-					if len(att) > 0 {
-						fileAttachments = append(fileAttachments, att)
-					}
-				}
-			}
-		}
-		if len(texts) > 0 {
-			// Join with newlines — user may have sent multiple text parts
-			result["message"] = strings.Join(texts, "\n")
-		}
-		if len(fileAttachments) > 0 {
-			result["attachments"] = fileAttachments
-		}
-	}
-
-	// Drop empty payloads
-	if len(result) == 0 {
-		return nil
-	}
-	return result
-}
-
 // extractToolTrace pulls metadata.tool_trace from an A2A JSON-RPC response.
 // Returns nil when absent or malformed — callers can pass it straight through.
 func extractToolTrace(respBody []byte) json.RawMessage {

workspace-server/internal/handlers/a2a_proxy_helpers_user_message_test.go

@@ -1,262 +0,0 @@
-package handlers
-
-import (
-	"encoding/json"
-	"testing"
-)
-
-// TestExtractCanvasUserMessage_TextOnly covers the primary path: a canvas user
-// sends a plain text message with no attachments.
-func TestExtractCanvasUserMessage_TextOnly(t *testing.T) {
-	body := []byte(`{
-		"method": "message/send",
-		"params": {
-			"message": {
-				"role": "user",
-				"messageId": "msg-abc-123",
-				"parts": [
-					{"kind": "text", "text": "Hello, agent!"}
-				]
-			}
-		}
-	}`)
-	got := extractCanvasUserMessage(body)
-	if got == nil {
-		t.Fatal("expected non-nil payload for text message")
-	}
-	if got["message"] != "Hello, agent!" {
-		t.Errorf("message = %v, want %q", got["message"], "Hello, agent!")
-	}
-	mid, ok := got["messageId"].(string)
-	if !ok || mid != "msg-abc-123" {
-		t.Errorf("messageId = %v, want %q", got["messageId"], "msg-abc-123")
-	}
-	_, hasAttachments := got["attachments"]
-	if hasAttachments {
-		t.Errorf("unexpected attachments: %v", got["attachments"])
-	}
-}
-
-// TestExtractCanvasUserMessage_FileOnly covers a user message with a file but no text.
-func TestExtractCanvasUserMessage_FileOnly(t *testing.T) {
-	body := []byte(`{
-		"method": "message/send",
-		"params": {
-			"message": {
-				"role": "user",
-				"messageId": "msg-file-456",
-				"parts": [
-					{
-						"kind": "file",
-						"file": {
-							"name": "report.pdf",
-							"uri": "workspace:/uploads/report.pdf",
-							"mimeType": "application/pdf",
-							"size": 4096
-						}
-					}
-				]
-			}
-		}
-	}`)
-	got := extractCanvasUserMessage(body)
-	if got == nil {
-		t.Fatal("expected non-nil payload for file-only message")
-	}
-	if got["message"] != nil {
-		t.Errorf("unexpected message text: %v", got["message"])
-	}
-	attachments, ok := got["attachments"].([]map[string]interface{})
-	if !ok || len(attachments) != 1 {
-		t.Fatalf("attachments = %v, want 1-element array", got["attachments"])
-	}
-	att := attachments[0]
-	if att["uri"] != "workspace:/uploads/report.pdf" {
-		t.Errorf("uri = %v, want %q", att["uri"], "workspace:/uploads/report.pdf")
-	}
-	if att["name"] != "report.pdf" {
-		t.Errorf("name = %v, want %q", att["name"], "report.pdf")
-	}
-	if att["mimeType"] != "application/pdf" {
-		t.Errorf("mimeType = %v, want %q", att["mimeType"], "application/pdf")
-	}
-}
-
-// TestExtractCanvasUserMessage_TextAndFile covers a user message with both text and a file.
-func TestExtractCanvasUserMessage_TextAndFile(t *testing.T) {
-	body := []byte(`{
-		"method": "message/send",
-		"params": {
-			"message": {
-				"role": "user",
-				"parts": [
-					{"kind": "text", "text": "Here is the file:"},
-					{"kind": "text", "text": "see below"},
-					{
-						"kind": "file",
-						"file": {
-							"name": "data.csv",
-							"uri": "workspace:/exports/data.csv",
-							"mimeType": "text/csv",
-							"size": 8192
-						}
-					}
-				]
-			}
-		}
-	}`)
-	got := extractCanvasUserMessage(body)
-	if got == nil {
-		t.Fatal("expected non-nil payload")
-	}
-	// Two text parts are joined with newline
-	if got["message"] != "Here is the file:\nsee below" {
-		t.Errorf("message = %v, want %q", got["message"], "Here is the file:\nsee below")
-	}
-	attachments, ok := got["attachments"].([]map[string]interface{})
-	if !ok || len(attachments) != 1 {
-		t.Fatalf("attachments = %v, want 1-element array", got["attachments"])
-	}
-}
-
-// TestExtractCanvasUserMessage_Malformed covers malformed JSON.
-func TestExtractCanvasUserMessage_Malformed(t *testing.T) {
-	cases := []struct {
-		name string
-		body []byte
-	}{
-		{"not JSON", []byte(`{not valid`)},
-		{"wrong type top-level", []byte(`123`)},
-		{"missing params", []byte(`{"method":"message/send"}`)},
-		{"params not object", []byte(`{"method":"message/send","params":123}`)},
-		{"missing message", []byte(`{"method":"message/send","params":{}}`)},
-		{"message not object", []byte(`{"method":"message/send","params":{"message":123}}`)},
-		{"role missing", []byte(`{"method":"message/send","params":{"message":{"parts":[]}}}`)},
-	}
-	for _, tc := range cases {
-		t.Run(tc.name, func(t *testing.T) {
-			if got := extractCanvasUserMessage(tc.body); got != nil {
-				t.Errorf("expected nil for %s, got %v", tc.name, got)
-			}
-		})
-	}
-}
-
-// TestExtractCanvasUserMessage_NotUserRole covers agent/workspace callers
-// whose role is not "user" — these should not be broadcast as USER_MESSAGE.
-func TestExtractCanvasUserMessage_NotUserRole(t *testing.T) {
-	cases := []struct {
-		name string
-		body []byte
-	}{
-		{
-			"agent role",
-			[]byte(`{"method":"message/send","params":{"message":{"role":"agent","parts":[{"kind":"text","text":"hello"}]}}}`),
-		},
-		{
-			"assistant role",
-			[]byte(`{"method":"message/send","params":{"message":{"role":"assistant","parts":[{"kind":"text","text":"hello"}]}}}`),
-		},
-		{
-			"empty role",
-			[]byte(`{"method":"message/send","params":{"message":{"role":"","parts":[{"kind":"text","text":"hello"}]}}}`),
-		},
-	}
-	for _, tc := range cases {
-		t.Run(tc.name, func(t *testing.T) {
-			if got := extractCanvasUserMessage(tc.body); got != nil {
-				t.Errorf("expected nil for role=%s, got %v", tc.name, got)
-			}
-		})
-	}
-}
-
-// TestExtractCanvasUserMessage_NotMessageSend covers non-message/send methods.
-func TestExtractCanvasUserMessage_NotMessageSend(t *testing.T) {
-	cases := []struct {
-		name   string
-		method string
-	}{
-		{"tasks/send", "tasks/send"},
-		{"initialize", "initialize"},
-		{"ping", "ping"},
-	}
-	for _, tc := range cases {
-		t.Run(tc.name, func(t *testing.T) {
-			body, _ := json.Marshal(map[string]interface{}{
-				"method": tc.method,
-				"params": map[string]interface{}{
-					"message": map[string]interface{}{
-						"role":  "user",
-						"parts": []map[string]interface{}{{"kind": "text", "text": "hello"}},
-					},
-				},
-			})
-			if got := extractCanvasUserMessage(body); got != nil {
-				t.Errorf("expected nil for method=%q, got %v", tc.method, got)
-			}
-		})
-	}
-}
-
-// TestExtractCanvasUserMessage_BlankOrEmpty covers text with only whitespace
-// and empty parts arrays.
-func TestExtractCanvasUserMessage_BlankOrEmpty(t *testing.T) {
-	cases := []struct {
-		name string
-		body []byte
-	}{
-		{
-			"empty text part",
-			[]byte(`{"method":"message/send","params":{"message":{"role":"user","parts":[{"kind":"text","text":""}]}}}`),
-		},
-		{
-			"empty parts array",
-			[]byte(`{"method":"message/send","params":{"message":{"role":"user","parts":[]}}}`),
-		},
-		{
-			"whitespace-only text — still included as valid content",
-			[]byte(`{"method":"message/send","params":{"message":{"role":"user","parts":[{"kind":"text","text":"   "}]}}}`),
-		},
-	}
-	for _, tc := range cases {
-		t.Run(tc.name, func(t *testing.T) {
-			got := extractCanvasUserMessage(tc.body)
-			if tc.name == "whitespace-only text — still included as valid content" {
-				// Whitespace-only text is valid content — preserve it as-is.
-				// Canvas dedup collapses identical copies; whitespace is not stripped.
-				if got == nil {
-					t.Error("expected non-nil for whitespace-only text")
-				} else if got["message"] != "   " {
-					t.Errorf("message = %q, want %q", got["message"], "   ")
-				}
-				return
-			}
-			if got != nil {
-				t.Errorf("expected nil for %s, got %v", tc.name, got)
-			}
-		})
-	}
-}
-
-// TestExtractCanvasUserMessage_Unicode covers non-ASCII text.
-func TestExtractCanvasUserMessage_Unicode(t *testing.T) {
-	body := []byte(`{
-		"method": "message/send",
-		"params": {
-			"message": {
-				"role": "user",
-				"parts": [
-					{"kind": "text", "text": "こんにちは世界 🌍 日本語"}
-				]
-			}
-		}
-	}`)
-	got := extractCanvasUserMessage(body)
-	if got == nil {
-		t.Fatal("expected non-nil payload for unicode message")
-	}
-	if got["message"] != "こんにちは世界 🌍 日本語" {
-		t.Errorf("message = %v, want %q", got["message"], "こんにちは世界 🌍 日本語")
-	}
-}

workspace-server/internal/handlers/activity.go

@@ -691,19 +691,6 @@ func logActivityExec(ctx context.Context, exec activityExecutor, broadcaster eve
 		if respStr != nil {
 			payload["response_body"] = json.RawMessage(respJSON)
 		}
-		// internal#211/#212: error_detail carries the runtime's curated,
-		// user-actionable, secret-safe failure reason (provider HTTP
-		// status + error code + the provider's own guidance, e.g. a 403
-		// "org disabled · use an API key / ask your admin"). It is
-		// already persisted to the DB column above and capped by the
-		// runtime's report_activity helper (4096 chars). Previously it
-		// was dropped from the LIVE broadcast, so the canvas had nothing
-		// to render and fell back to a hardcoded opaque
-		// "Agent error (Exception) — see workspace logs" string. Include
-		// it so the chat bubble shows the real reason in real time.
-		if params.ErrorDetail != nil && *params.ErrorDetail != "" {
-			payload["error_detail"] = *params.ErrorDetail
-		}
 	}
 
 	return func() {

workspace-server/internal/handlers/agent_git_identity.go

@@ -17,6 +17,17 @@ 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 +61,34 @@ 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)
 }
 
 // slugifyForEmail collapses a workspace name to a safe email localpart:

workspace-server/internal/handlers/agent_git_identity_test.go

@@ -75,6 +75,53 @@ 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)
+}
+
 func TestSlugifyForEmail(t *testing.T) {
 	cases := []struct {
 		in, want string

workspace-server/internal/handlers/approvals.go

@@ -51,29 +51,23 @@ func (h *ApprovalsHandler) Create(c *gin.Context) {
 		return
 	}
 
-	if err := h.broadcaster.RecordAndBroadcast(ctx, string(events.EventApprovalRequested), workspaceID, map[string]interface{}{
+	h.broadcaster.RecordAndBroadcast(ctx, string(events.EventApprovalRequested), workspaceID, map[string]interface{}{
 		"approval_id": approvalID,
 		"action":      body.Action,
 		"reason":      body.Reason,
 		"task_id":     body.TaskID,
-	}); err != nil {
-		log.Printf("approvals: failed to broadcast approval requested: %v", err)
-	}
+	})
 
 	// Auto-escalate to parent
 	var parentID *string
-	if err := db.DB.QueryRowContext(ctx, `SELECT parent_id FROM workspaces WHERE id = $1`, workspaceID).Scan(&parentID); err != nil {
-		log.Printf("approvals: failed to lookup parent for escalation: %v", err)
-	}
+	db.DB.QueryRowContext(ctx, `SELECT parent_id FROM workspaces WHERE id = $1`, workspaceID).Scan(&parentID)
 	if parentID != nil {
-		if err := h.broadcaster.RecordAndBroadcast(ctx, string(events.EventApprovalEscalated), *parentID, map[string]interface{}{
+		h.broadcaster.RecordAndBroadcast(ctx, string(events.EventApprovalEscalated), *parentID, map[string]interface{}{
 			"approval_id":       approvalID,
 			"from_workspace_id": workspaceID,
 			"action":            body.Action,
 			"reason":            body.Reason,
-		}); err != nil {
-			log.Printf("approvals: failed to broadcast approval escalated: %v", err)
-		}
+		})
 	}
 
 	c.JSON(http.StatusCreated, gin.H{"approval_id": approvalID, "status": "pending"})
@@ -86,12 +80,10 @@ func (h *ApprovalsHandler) ListAll(c *gin.Context) {
 	ctx := c.Request.Context()
 
 	// Auto-expire stale approvals (older than 10 min)
-	if _, err := db.DB.ExecContext(ctx, `
+	db.DB.ExecContext(ctx, `
 		UPDATE approval_requests SET status = 'denied', decided_by = 'auto-expired', decided_at = now()
 		WHERE status = 'pending' AND created_at < now() - interval '10 minutes'
-	`); err != nil {
-		log.Printf("approvals: failed to auto-expire stale approvals: %v", err)
-	}
+	`)
 
 	rows, err := db.DB.QueryContext(ctx, `
 		SELECT a.id, a.workspace_id, w.name, a.action, a.reason, a.status, a.created_at
@@ -219,13 +211,11 @@ func (h *ApprovalsHandler) Decide(c *gin.Context) {
 		eventType = "APPROVAL_DENIED"
 	}
 
-	if err := h.broadcaster.RecordAndBroadcast(ctx, eventType, workspaceID, map[string]interface{}{
+	h.broadcaster.RecordAndBroadcast(ctx, eventType, workspaceID, map[string]interface{}{
 		"approval_id": approvalID,
 		"decision":    body.Decision,
 		"decided_by":  decidedBy,
-	}); err != nil {
-		log.Printf("approvals: failed to broadcast approval decision: %v", err)
-	}
+	})
 
 	c.JSON(http.StatusOK, gin.H{"status": body.Decision, "approval_id": approvalID})
 }

workspace-server/internal/handlers/channels.go

@@ -558,11 +558,6 @@ func (h *ChannelHandler) Webhook(c *gin.Context) {
 
 	// Process asynchronously — don't block the webhook response
 	go func() {
-		defer func() {
-			if r := recover(); r != nil {
-				log.Printf("Channels: PANIC in async HandleInbound for workspace %s: %v", ch.WorkspaceID[:12], r)
-			}
-		}()
 		bgCtx := context.Background()
 		if err := h.manager.HandleInbound(bgCtx, ch, msg); err != nil {
 			log.Printf("Channels: async HandleInbound error for workspace %s: %v", ch.WorkspaceID[:12], err)

workspace-server/internal/handlers/chat_files.go

@@ -107,29 +107,10 @@ func (h *ChatFilesHandler) WithPendingUploads(storage pendinguploads.Storage, br
 }
 
 // chatUploadMaxBytes caps the full multipart request body so a
-// malicious / runaway client can't OOM the proxy hop. 100 MB matches
-// the workspace-side total limit; anything larger is rejected at the
+// malicious / runaway client can't OOM the proxy hop. 50 MB matches
+// the workspace-side limit; anything larger is rejected at the
 // network boundary before forwarding.
-//
-// SSOT NOTE (issue #1520): this constant is the source of truth for
-// chat upload limits across the platform. Its value is exported to
-// the workspace container at provision time via the env var
-// CHAT_UPLOAD_MAX_TOTAL_BYTES (see
-// workspace_provision_shared.go::applyChatUploadLimits) so the
-// Python runtime cap stays in lock-step. Do NOT change this without
-// updating the per-file cap chatUploadMaxFileBytes below and
-// verifying the env-injection site is unchanged.
-const chatUploadMaxBytes = 100 * 1024 * 1024
-
-// chatUploadMaxFileBytes caps any single multipart part. Mirrors the
-// total cap by default because most chat uploads are a single file;
-// keeping per-file equal to total avoids the surprise of "my 60 MB
-// file fit under the total but got 413'd on per-file". Exported to
-// the workspace container as CHAT_UPLOAD_MAX_FILE_BYTES so the
-// Starlette parser's max_part_size matches and any single part above
-// Starlette's default 1 MiB no longer raises MultiPartException
-// (root cause of issue #1520).
-const chatUploadMaxFileBytes = 100 * 1024 * 1024
+const chatUploadMaxBytes = 50 * 1024 * 1024
 
 // resolveWorkspaceForwardCreds resolves the workspace's URL +
 // platform_inbound_secret for an /internal/* forward, applying

workspace-server/internal/handlers/chat_upload_limits_test.go

@@ -1,63 +0,0 @@
-package handlers
-
-// chat_upload_limits_test.go — pins the SSOT env-injection contract
-// for chat-upload caps (issue #1520). The Python workspace runtime
-// reads these env vars at module init; drift between the constant in
-// chat_files.go and the env-var name here silently breaks chat upload
-// fleet-wide, so the contract is asserted as a unit test in the same
-// package as the producer.
-
-import (
-	"fmt"
-	"testing"
-)
-
-// applyChatUploadLimits MUST seed both env vars to the byte-count
-// stringification of the Go-side constants. Anything else means a
-// Python-side parser cap that disagrees with the Go-side network cap,
-// which is exactly the drift that shipped #1520.
-func TestApplyChatUploadLimits_DefaultsMatchGoConstants(t *testing.T) {
-	env := map[string]string{}
-	applyChatUploadLimits(env)
-
-	wantFile := fmt.Sprintf("%d", chatUploadMaxFileBytes)
-	if got := env["CHAT_UPLOAD_MAX_FILE_BYTES"]; got != wantFile {
-		t.Errorf("CHAT_UPLOAD_MAX_FILE_BYTES = %q, want %q", got, wantFile)
-	}
-
-	wantTotal := fmt.Sprintf("%d", chatUploadMaxBytes)
-	if got := env["CHAT_UPLOAD_MAX_TOTAL_BYTES"]; got != wantTotal {
-		t.Errorf("CHAT_UPLOAD_MAX_TOTAL_BYTES = %q, want %q", got, wantTotal)
-	}
-}
-
-// Pre-existing values win. A tenant override, plugin mutator, or A/B
-// experiment that already set the env MUST be preserved — the SSOT
-// helper is a defaulting layer, not an override layer.
-func TestApplyChatUploadLimits_PreExistingValuesPreserved(t *testing.T) {
-	env := map[string]string{
-		"CHAT_UPLOAD_MAX_FILE_BYTES":  "1234",
-		"CHAT_UPLOAD_MAX_TOTAL_BYTES": "5678",
-	}
-	applyChatUploadLimits(env)
-
-	if got := env["CHAT_UPLOAD_MAX_FILE_BYTES"]; got != "1234" {
-		t.Errorf("pre-existing CHAT_UPLOAD_MAX_FILE_BYTES overwritten: got %q", got)
-	}
-	if got := env["CHAT_UPLOAD_MAX_TOTAL_BYTES"]; got != "5678" {
-		t.Errorf("pre-existing CHAT_UPLOAD_MAX_TOTAL_BYTES overwritten: got %q", got)
-	}
-}
-
-// The 100 MB minimum is the CTO-directed allowance floor (issue #1520).
-// Pin so a future "tidy up: 100 MB seems large" refactor surfaces here
-// before reverting the user-visible behaviour change.
-func TestChatUploadCaps_MinimumAllowanceFloor(t *testing.T) {
-	const floor = 100 * 1024 * 1024
-	if chatUploadMaxBytes < floor {
-		t.Errorf("chatUploadMaxBytes = %d, below #1520 floor %d", chatUploadMaxBytes, floor)
-	}
-	if chatUploadMaxFileBytes < floor {
-		t.Errorf("chatUploadMaxFileBytes = %d, below #1520 floor %d", chatUploadMaxFileBytes, floor)
-	}
-}

workspace-server/internal/handlers/delegation.go

@@ -747,14 +747,6 @@ func (h *DelegationHandler) listDelegationsFromLedger(ctx context.Context, works
 			entry["response_preview"] = textutil.TruncateBytes(resultPreview.String, 300)
 		}
 		if errorDetail.Valid && errorDetail.String != "" {
-			// Emit both keys: `error_detail` is the canonical field the
-			// Python poll-mode consumer (a2a_tools_delegation.py:184)
-			// reads from /delegations rows — without it, poll-mode
-			// silently loses the failure reason and falls through to
-			// the generic "delegation failed" string. `error` is kept
-			// for back-compat with existing UI surfaces that read the
-			// shorter name.
-			entry["error_detail"] = errorDetail.String
 			entry["error"] = errorDetail.String
 		}
 		if lastHeartbeat != nil {
@@ -816,8 +808,6 @@ func (h *DelegationHandler) listDelegationsFromActivityLogs(ctx context.Context,
 			entry["delegation_id"] = delegationID
 		}
 		if errorDetail != "" {
-			// Emit both keys per the rename: see listDelegationsFromLedger.
-			entry["error_detail"] = errorDetail
 			entry["error"] = errorDetail
 		}
 		if responseBody != "" {

workspace-server/internal/handlers/delegation_test.go

@@ -1546,71 +1546,6 @@ func TestListDelegations_LedgerEmptyFallsBackToActivityLogs(t *testing.T) {
 	}
 }
 
-// ---------- ListDelegations: activity_logs failed row emits BOTH error + error_detail ----------
-
-// Field-rename pin (P1 #348 / RFC #2829 PR-2 follow-up): the legacy
-// activity_logs fallback path must also emit `error_detail` alongside
-// the historical `error` key. Without this, poll-mode (which reads
-// `error_detail`) silently loses the failure reason when the ledger
-// is empty and the handler falls back to activity_logs.
-func TestListDelegations_ActivityLogsFailedEmitsBothErrorKeys(t *testing.T) {
-	mock := setupTestDB(t)
-	setupTestRedis(t)
-	broadcaster := newTestBroadcaster()
-	wh := NewWorkspaceHandler(broadcaster, nil, "http://localhost:8080", t.TempDir())
-	dh := NewDelegationHandler(wh, broadcaster)
-
-	// Ledger empty → fall back to activity_logs.
-	mock.ExpectQuery("SELECT d.delegation_id, d.caller_id, d.callee_id, d.task_preview").
-		WithArgs("ws-source").
-		WillReturnRows(sqlmock.NewRows([]string{
-			"delegation_id", "caller_id", "callee_id", "task_preview",
-			"status", "result_preview", "error_detail", "last_heartbeat",
-			"deadline", "created_at", "updated_at",
-		}))
-
-	now := time.Now()
-	activityRows := sqlmock.NewRows([]string{
-		"id", "activity_type", "source_id", "target_id",
-		"summary", "status", "error_detail", "response_body",
-		"delegation_id", "created_at",
-	}).AddRow(
-		"act-failed", "delegate_result", "ws-source", "ws-target",
-		"Delegation failed", "error", "codex runtime timed out", "",
-		"del-failed-002", now,
-	)
-	mock.ExpectQuery("SELECT id, activity_type").
-		WithArgs("ws-source").
-		WillReturnRows(activityRows)
-
-	w := httptest.NewRecorder()
-	c, _ := gin.CreateTestContext(w)
-	c.Params = gin.Params{{Key: "id", Value: "ws-source"}}
-	c.Request = httptest.NewRequest("GET", "/workspaces/ws-source/delegations", nil)
-
-	dh.ListDelegations(c)
-
-	if w.Code != http.StatusOK {
-		t.Fatalf("expected 200, got %d: %s", w.Code, w.Body.String())
-	}
-	var resp []map[string]interface{}
-	if err := json.Unmarshal(w.Body.Bytes(), &resp); err != nil {
-		t.Fatalf("failed to parse response: %v", err)
-	}
-	if len(resp) != 1 {
-		t.Fatalf("expected 1 row, got %d", len(resp))
-	}
-	if resp[0]["error"] != "codex runtime timed out" {
-		t.Errorf("expected `error` field set, got %v", resp[0]["error"])
-	}
-	if resp[0]["error_detail"] != "codex runtime timed out" {
-		t.Errorf("expected `error_detail` field set (poll-mode contract), got %v", resp[0]["error_detail"])
-	}
-	if err := mock.ExpectationsWereMet(); err != nil {
-		t.Errorf("unmet sqlmock expectations: %v", err)
-	}
-}
-
 // ---------- ListDelegations: both ledger and activity_logs empty → [] ----------
 
 func TestListDelegations_BothEmptyReturnsEmptyArray(t *testing.T) {
@@ -1809,15 +1744,7 @@ func TestListDelegations_LedgerFailedIncludesErrorDetail(t *testing.T) {
 		t.Errorf("expected status 'failed', got %v", resp[0]["status"])
 	}
 	if resp[0]["error"] != "Callee workspace not reachable" {
-		t.Errorf("expected error detail under `error`, got %v", resp[0]["error"])
-	}
-	// Field-rename pin (P1 #348 / RFC #2829 PR-2 follow-up): the
-	// Python poll-mode consumer in a2a_tools_delegation.py:184 reads
-	// `error_detail`, not `error`. Both keys MUST be present so polling
-	// surfaces the real failure reason instead of falling through to
-	// the generic "delegation failed" string.
-	if resp[0]["error_detail"] != "Callee workspace not reachable" {
-		t.Errorf("expected error detail under `error_detail`, got %v", resp[0]["error_detail"])
+		t.Errorf("expected error detail, got %v", resp[0]["error"])
 	}
 	if err := mock.ExpectationsWereMet(); err != nil {
 		t.Errorf("unmet sqlmock expectations: %v", err)

workspace-server/internal/handlers/discovery.go

@@ -239,7 +239,7 @@ func (h *DiscoveryHandler) Peers(c *gin.Context) {
 
 	// Siblings
 	if parentID.Valid {
-		siblings, _ := queryPeerMaps(ctx, `
+		siblings, _ := queryPeerMaps(`
 			SELECT w.id, w.name, COALESCE(w.role, ''), w.tier, w.status,
 				   COALESCE(w.agent_card, 'null'::jsonb), COALESCE(w.url, ''),
 				   w.parent_id, w.active_tasks
@@ -247,7 +247,7 @@ func (h *DiscoveryHandler) Peers(c *gin.Context) {
 			parentID.String, workspaceID)
 		peers = append(peers, siblings...)
 	} else {
-		siblings, _ := queryPeerMaps(ctx, `
+		siblings, _ := queryPeerMaps(`
 			SELECT w.id, w.name, COALESCE(w.role, ''), w.tier, w.status,
 				   COALESCE(w.agent_card, 'null'::jsonb), COALESCE(w.url, ''),
 				   w.parent_id, w.active_tasks
@@ -257,7 +257,7 @@ func (h *DiscoveryHandler) Peers(c *gin.Context) {
 	}
 
 	// Children
-	children, _ := queryPeerMaps(ctx, `
+	children, _ := queryPeerMaps(`
 		SELECT w.id, w.name, COALESCE(w.role, ''), w.tier, w.status,
 			   COALESCE(w.agent_card, 'null'::jsonb), COALESCE(w.url, ''),
 			   w.parent_id, w.active_tasks
@@ -266,7 +266,7 @@ func (h *DiscoveryHandler) Peers(c *gin.Context) {
 
 	// Parent
 	if parentID.Valid {
-		parent, _ := queryPeerMaps(ctx, `
+		parent, _ := queryPeerMaps(`
 			SELECT w.id, w.name, COALESCE(w.role, ''), w.tier, w.status,
 				   COALESCE(w.agent_card, 'null'::jsonb), COALESCE(w.url, ''),
 				   w.parent_id, w.active_tasks
@@ -303,8 +303,8 @@ func filterPeersByQuery(peers []map[string]interface{}, q string) []map[string]i
 }
 
 // queryPeerMaps returns clean JSON-serializable maps instead of Workspace structs.
-func queryPeerMaps(ctx context.Context, query string, args ...interface{}) ([]map[string]interface{}, error) {
-	rows, err := db.DB.QueryContext(ctx, query, args...)
+func queryPeerMaps(query string, args ...interface{}) ([]map[string]interface{}, error) {
+	rows, err := db.DB.Query(query, args...)
 	if err != nil {
 		log.Printf("queryPeerMaps error: %v", err)
 		return nil, err

workspace-server/internal/handlers/external_connection.go

@@ -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.
 `

workspace-server/internal/handlers/external_rotate.go

@@ -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
 }

workspace-server/internal/handlers/external_rotate_test.go

@@ -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)
+		}
+	}
+}

workspace-server/internal/handlers/mcp_tools.go

@@ -15,7 +15,6 @@ import (
 	"fmt"
 	"io"
 	"log"
-	"net"
 	"net/http"
 	"os"
 	"strings"
@@ -55,22 +54,6 @@ func updateMCPDelegationStatus(ctx context.Context, db *sql.DB, workspaceID, del
 	}
 }
 
-// ─────────────────────────────────────────────────────────────────────────────
-// mcpHTTPClient is a dedicated client for MCP bridge A2A calls.
-// Per-request deadlines are enforced via context (30 s sync, 8 s async).
-// Transport-level timeouts ensure dead workspaces fail fast instead of
-// hanging on OS default TCP timeouts (~75 s Linux).
-var mcpHTTPClient = &http.Client{
-	Transport: &http.Transport{
-		DialContext: (&net.Dialer{
-			Timeout:   5 * time.Second,
-			KeepAlive: 30 * time.Second,
-		}).DialContext,
-		ResponseHeaderTimeout: 30 * time.Second,
-		TLSHandshakeTimeout:   5 * time.Second,
-	},
-}
-
 // ─────────────────────────────────────────────────────────────────────────────
 // Tool implementations
 // ─────────────────────────────────────────────────────────────────────────────
@@ -248,7 +231,7 @@ func (h *MCPHandler) toolDelegateTask(ctx context.Context, callerID string, args
 	// so this header reflects a verified caller identity, not a spoofable value.
 	httpReq.Header.Set("X-Workspace-ID", callerID)
 
-	resp, err := mcpHTTPClient.Do(httpReq)
+	resp, err := http.DefaultClient.Do(httpReq)
 	if err != nil {
 		updateMCPDelegationStatus(ctx, h.database, callerID, delegationID, "failed", err.Error())
 		return "", fmt.Errorf("A2A call failed: %w", err)
@@ -296,11 +279,6 @@ func (h *MCPHandler) toolDelegateTaskAsync(ctx context.Context, callerID string,
 	// Fire and forget in a detached goroutine. Use a background context so
 	// the call is not cancelled when the HTTP request completes.
 	go func() {
-		defer func() {
-			if r := recover(); r != nil {
-				log.Printf("MCPHandler.delegate_task_async: PANIC for %s → %s: %v", callerID, targetID, r)
-			}
-		}()
 		bgCtx, cancel := context.WithTimeout(context.Background(), mcpAsyncCallTimeout)
 		defer cancel()
 
@@ -336,7 +314,7 @@ func (h *MCPHandler) toolDelegateTaskAsync(ctx context.Context, callerID string,
 		httpReq.Header.Set("Content-Type", "application/json")
 		httpReq.Header.Set("X-Workspace-ID", callerID)
 
-		resp, err := mcpHTTPClient.Do(httpReq)
+		resp, err := http.DefaultClient.Do(httpReq)
 		if err != nil {
 			log.Printf("MCPHandler.delegate_task_async: A2A call to %s: %v", targetID, err)
 			return

workspace-server/internal/handlers/org_helpers.go

@@ -218,6 +218,14 @@ func loadWorkspaceEnv(orgBaseDir, filesDir string) map[string]string {
 // check, or when the env file does not exist (workspaces without a role —
 // or running on hosts that don't ship the bootstrap dir — keep their old
 // behavior).
+//
+// Token-file fallback: the newer prod-team personas (agent-dev-a,
+// agent-dev-b, agent-pm) ship `token` + `universal-auth.env` only — no
+// legacy plaintext `env` file. When the env-file load produces zero rows,
+// loadPersonaTokenFile fills in GITEA_TOKEN / GITEA_USER / GITEA_USER_EMAIL
+// from the token file so the GIT_ASKPASS helper has something to emit.
+// The env-file form remains authoritative when present (it may carry
+// richer rows like GITEA_TOKEN_SCOPES / GITEA_SSH_KEY_PATH).
 func loadPersonaEnvFile(role string, out map[string]string) {
 	if !isSafeRoleName(role) {
 		if role != "" {
@@ -229,7 +237,61 @@ func loadPersonaEnvFile(role string, out map[string]string) {
 	if root == "" {
 		root = "/etc/molecule-bootstrap/personas"
 	}
+	before := len(out)
 	parseEnvFile(filepath.Join(root, role, "env"), out)
+	if len(out) == before {
+		// No env-file rows landed (file absent, or present-but-empty).
+		// Try the token-only persona shape used by the prod-team
+		// identities. Existing keys in out are preserved.
+		loadPersonaTokenFile(role, out)
+	}
+}
+
+// loadPersonaTokenFile populates GITEA_TOKEN / GITEA_USER / GITEA_USER_EMAIL
+// from a persona dir that ships only the bare `token` file — the shape used
+// by the production agent personas (agent-dev-a, agent-dev-b, agent-pm).
+// Those dirs do not carry an `env` file because their non-Gitea creds come
+// from Infisical Universal Auth at runtime (universal-auth.env), so the
+// historical loadPersonaEnvFile path silently no-ops on them.
+//
+// File layout: $MOLECULE_PERSONA_ROOT/<role>/token (mode 600, plain text).
+// The token contents become GITEA_TOKEN (whitespace-trimmed); the role
+// name becomes GITEA_USER; GITEA_USER_EMAIL is synthesised as
+// <role>@<gitIdentityEmailDomain> to match the email shape that
+// applyAgentGitIdentity uses for its slug-derived authorship addresses.
+//
+// Silent no-op when the role fails the safe-segment check, when the
+// token file does not exist, or when its contents are empty after
+// trimming. Existing keys in out are not overwritten — the caller's
+// later .env layers and any prior loadPersonaEnvFile rows always win.
+func loadPersonaTokenFile(role string, out map[string]string) {
+	if out == nil {
+		return
+	}
+	if !isSafeRoleName(role) {
+		return
+	}
+	root := os.Getenv("MOLECULE_PERSONA_ROOT")
+	if root == "" {
+		root = "/etc/molecule-bootstrap/personas"
+	}
+	data, err := os.ReadFile(filepath.Join(root, role, "token"))
+	if err != nil {
+		return
+	}
+	token := strings.TrimSpace(string(data))
+	if token == "" {
+		return
+	}
+	if _, ok := out["GITEA_TOKEN"]; !ok {
+		out["GITEA_TOKEN"] = token
+	}
+	if _, ok := out["GITEA_USER"]; !ok {
+		out["GITEA_USER"] = role
+	}
+	if _, ok := out["GITEA_USER_EMAIL"]; !ok {
+		out["GITEA_USER_EMAIL"] = role + "@" + gitIdentityEmailDomain
+	}
 }
 
 // isSafeRoleName accepts a single path segment of [A-Za-z0-9_-]+. Rejects

workspace-server/internal/handlers/org_persona_env_test.go

@@ -164,3 +164,181 @@ func TestIsSafeRoleName_Acceptance(t *testing.T) {
 		}
 	}
 }
+
+// TestLoadPersonaTokenFile_TokenOnlyPersona: the prod-team personas
+// (agent-dev-a / agent-dev-b / agent-pm) ship `token` only — no `env`
+// file. loadPersonaEnvFile's fallback path must populate GITEA_TOKEN /
+// GITEA_USER / GITEA_USER_EMAIL from the token contents + role name so
+// the GIT_ASKPASS helper has something to emit.
+func TestLoadPersonaTokenFile_TokenOnlyPersona(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)
+
+	out := map[string]string{}
+	loadPersonaEnvFile("agent-dev-a", out)
+
+	want := map[string]string{
+		"GITEA_TOKEN":      "token-bytes-redacted",
+		"GITEA_USER":       "agent-dev-a",
+		"GITEA_USER_EMAIL": "agent-dev-a@" + gitIdentityEmailDomain,
+	}
+	if len(out) != len(want) {
+		t.Fatalf("got %d keys, want %d: %#v", len(out), len(want), out)
+	}
+	for k, v := range want {
+		if out[k] != v {
+			t.Errorf("out[%q] = %q; want %q", k, out[k], v)
+		}
+	}
+}
+
+// TestLoadPersonaTokenFile_EnvFileWins: when BOTH an env file and a
+// token file exist in the same persona dir, the env file is the more-
+// specific declaration and wins outright — the fallback must not fire
+// at all. This pins precedence so a persona later migrated to the
+// richer env-file form (carrying GITEA_TOKEN_SCOPES / GITEA_SSH_KEY_PATH)
+// doesn't get its token silently overridden by the fallback.
+func TestLoadPersonaTokenFile_EnvFileWins(t *testing.T) {
+	root := t.TempDir()
+	roleDir := filepath.Join(root, "agent-dev-b")
+	if err := os.MkdirAll(roleDir, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	envBody := "GITEA_USER=env-form-user\nGITEA_TOKEN=env-form-token\n" +
+		"GITEA_USER_EMAIL=env-form@example.invalid\nGITEA_TOKEN_SCOPES=write:repository\n"
+	if err := os.WriteFile(filepath.Join(roleDir, "env"), []byte(envBody), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(filepath.Join(roleDir, "token"),
+		[]byte("token-form-token\n"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("MOLECULE_PERSONA_ROOT", root)
+
+	out := map[string]string{}
+	loadPersonaEnvFile("agent-dev-b", out)
+
+	if out["GITEA_USER"] != "env-form-user" {
+		t.Errorf("env file should win for GITEA_USER; got %q", out["GITEA_USER"])
+	}
+	if out["GITEA_TOKEN"] != "env-form-token" {
+		t.Errorf("env file should win for GITEA_TOKEN; got %q", out["GITEA_TOKEN"])
+	}
+	if out["GITEA_USER_EMAIL"] != "env-form@example.invalid" {
+		t.Errorf("env file should win for GITEA_USER_EMAIL; got %q", out["GITEA_USER_EMAIL"])
+	}
+	if out["GITEA_TOKEN_SCOPES"] != "write:repository" {
+		t.Errorf("env file extras must be preserved; got GITEA_TOKEN_SCOPES=%q", out["GITEA_TOKEN_SCOPES"])
+	}
+}
+
+// TestLoadPersonaTokenFile_NeitherFile: persona dir exists but ships
+// neither env nor token — silent no-op. This is the legitimate case
+// for a partially-provisioned persona during bootstrap; callers expect
+// an empty map, no error, no log noise.
+func TestLoadPersonaTokenFile_NeitherFile(t *testing.T) {
+	root := t.TempDir()
+	roleDir := filepath.Join(root, "agent-pm")
+	if err := os.MkdirAll(roleDir, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("MOLECULE_PERSONA_ROOT", root)
+
+	out := map[string]string{}
+	loadPersonaEnvFile("agent-pm", out)
+	if len(out) != 0 {
+		t.Errorf("expected empty out when neither env nor token exists; got %#v", out)
+	}
+}
+
+// TestLoadPersonaTokenFile_EmptyToken: a token file with only
+// whitespace must be treated as absent — never emit
+// GITEA_TOKEN="" / GITEA_USER=<role> / GITEA_USER_EMAIL=<role>@... because
+// that would set GITEA_USER without a usable token, and the askpass
+// helper would then prompt with an empty password. Silent no-op is the
+// correct behavior — let downstream auth fall through to its existing
+// "no credentials available" path.
+func TestLoadPersonaTokenFile_EmptyToken(t *testing.T) {
+	root := t.TempDir()
+	roleDir := filepath.Join(root, "agent-dev-a")
+	if err := os.MkdirAll(roleDir, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	// Whitespace-only contents: spaces, tabs, newlines.
+	if err := os.WriteFile(filepath.Join(roleDir, "token"),
+		[]byte("   \t\n  \n"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	t.Setenv("MOLECULE_PERSONA_ROOT", root)
+
+	out := map[string]string{}
+	loadPersonaEnvFile("agent-dev-a", out)
+	if len(out) != 0 {
+		t.Errorf("expected empty out when token file is whitespace-only; got %#v", out)
+	}
+}
+
+// TestLoadPersonaTokenFile_TrimsWhitespace: tokens shipped from the
+// operator-host bootstrap kit may have a trailing newline (the
+// canonical `printf "%s\n" "$token" > token` shape). The fallback must
+// trim leading + trailing whitespace so the askpass helper emits the
+// raw token bytes — Gitea's PAT validator rejects tokens with embedded
+// whitespace.
+func TestLoadPersonaTokenFile_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)
+
+	out := map[string]string{}
+	loadPersonaEnvFile("agent-dev-b", out)
+	if out["GITEA_TOKEN"] != "raw-token-bytes" {
+		t.Errorf("token whitespace not trimmed; got %q", out["GITEA_TOKEN"])
+	}
+}
+
+// TestLoadPersonaTokenFile_RejectsUnsafeRole: defense-in-depth — even
+// in the fallback path, role names that fail isSafeRoleName must not
+// touch the filesystem. Mirrors TestLoadPersonaEnvFile_RejectsTraversal.
+func TestLoadPersonaTokenFile_RejectsUnsafeRole(t *testing.T) {
+	root := t.TempDir()
+	// Plant a token at /tmp/.../token so a bad traversal would reach it.
+	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", "."} {
+		out := map[string]string{}
+		loadPersonaTokenFile(bad, out)
+		if len(out) != 0 {
+			t.Errorf("role %q should have been rejected; got %#v", bad, out)
+		}
+	}
+}
+
+// TestLoadPersonaTokenFile_NilMapSafe: callers pass a fresh map in
+// practice, but defense-in-depth — a nil map must not panic.
+func TestLoadPersonaTokenFile_NilMapSafe(t *testing.T) {
+	defer func() {
+		if r := recover(); r != nil {
+			t.Fatalf("nil map caused panic: %v", r)
+		}
+	}()
+	loadPersonaTokenFile("agent-dev-a", nil)
+}

workspace-server/internal/handlers/plugins_install.go

@@ -201,11 +201,6 @@ func (h *PluginsHandler) uninstallViaDocker(ctx context.Context, c *gin.Context,
 	// Auto-restart (small delay to ensure fs writes are flushed)
 	if h.restartFunc != nil {
 		go func() {
-			defer func() {
-				if r := recover(); r != nil {
-					log.Printf("plugins_install: PANIC in delayed restart for %s: %v", workspaceID, r)
-				}
-			}()
 			time.Sleep(2 * time.Second)
 			h.restartFunc(workspaceID)
 		}()

workspace-server/internal/handlers/plugins_install_test.go

@@ -1,53 +0,0 @@
-package handlers
-
-// plugins_install_test.go — additional coverage for plugins_install.go.
-//
-// Gaps filled vs. existing test files:
-//   - plugins_install_external_test.go: Install + Uninstall 422 (external runtime) ✓ covered
-//   - plugins_test.go:                 Install 400 (missing source, invalid body, etc.) ✓ covered
-//                                    Uninstall 400 (invalid plugin name, empty name) ✓ covered
-//                                    Download auth gate ✓ covered
-//   - org_import_helpers_test.go:      countWorkspaces, envRequirementKey, sanitizeEnvMembers,
-//                                        flattenAndSortRequirements, collectOrgEnv ✓ covered
-//
-// New test added here:
-//   - Uninstall 503: container not running, no SaaS dispatch.
-//
-// NOTE: validateWorkspaceID is not called inside the Install/Uninstall handlers.
-// UUID validation is the responsibility of the WorkspaceAuth middleware, so no
-// 400 test is needed here for UUID format.
-
-import (
-	"encoding/json"
-	"net/http"
-	"net/http/httptest"
-	"testing"
-
-	"github.com/gin-gonic/gin"
-	"github.com/stretchr/testify/require"
-)
-
-// TestPluginUninstall_ContainerNotRunning_Returns503 exercises the 503 path
-// where neither a local Docker container nor a SaaS instance-id dispatch
-// resolves. The handler must return "workspace container not running" — NOT a
-// generic 500 or a misleading 422 (external-runtime) message.
-func TestPluginUninstall_ContainerNotRunning_Returns503(t *testing.T) {
-	// No docker client + no instance-id lookup → falls through to 503.
-	h := NewPluginsHandler(t.TempDir(), nil, nil)
-
-	w := httptest.NewRecorder()
-	c, _ := gin.CreateTestContext(w)
-	c.Params = gin.Params{
-		{Key: "id", Value: "550e8400-e29b-41d4-a716-446655440000"},
-		{Key: "name", Value: "some-plugin"},
-	}
-	c.Request = httptest.NewRequest("DELETE",
-		"/workspaces/550e8400-e29b-41d4-a716-446655440000/plugins/some-plugin", nil)
-
-	h.Uninstall(c)
-
-	require.Equal(t, http.StatusServiceUnavailable, w.Code)
-	var body map[string]string
-	json.Unmarshal(w.Body.Bytes(), &body)
-	require.Equal(t, "workspace container not running", body["error"])
-}

workspace-server/internal/handlers/plugins_listing_test.go

@@ -1,141 +0,0 @@
-package handlers
-
-import (
-	"encoding/json"
-	"net/http"
-	"net/http/httptest"
-	"os"
-	"path/filepath"
-	"testing"
-
-	"github.com/gin-gonic/gin"
-)
-
-func TestListRegistry_EmptyDir(t *testing.T) {
-	dir := t.TempDir()
-	h := NewPluginsHandler(dir, nil, nil)
-
-	got := h.listRegistryFiltered("")
-	if len(got) != 0 {
-		t.Errorf("expected empty list, got %d plugins", len(got))
-	}
-}
-
-func TestListRegistry_IgnoresFiles(t *testing.T) {
-	dir := t.TempDir()
-	if err := os.WriteFile(filepath.Join(dir, "not-a-plugin.txt"), []byte("x"), 0600); err != nil {
-		t.Fatal(err)
-	}
-	h := NewPluginsHandler(dir, nil, nil)
-
-	got := h.listRegistryFiltered("")
-	if len(got) != 0 {
-		t.Errorf("expected empty list (files ignored), got %d", len(got))
-	}
-}
-
-func TestListRegistry_SinglePlugin(t *testing.T) {
-	dir := t.TempDir()
-	pluginDir := filepath.Join(dir, "my-plugin")
-	if err := os.Mkdir(pluginDir, 0755); err != nil {
-		t.Fatal(err)
-	}
-	if err := os.WriteFile(filepath.Join(pluginDir, "plugin.yaml"), []byte("name: my-plugin\nversion: 1.0.0\n"), 0600); err != nil {
-		t.Fatal(err)
-	}
-	h := NewPluginsHandler(dir, nil, nil)
-
-	got := h.listRegistryFiltered("")
-	if len(got) != 1 {
-		t.Fatalf("expected 1 plugin, got %d", len(got))
-	}
-	if got[0].Name != "my-plugin" {
-		t.Errorf("expected name 'my-plugin', got %q", got[0].Name)
-	}
-}
-
-func TestListRegistry_FiltersByRuntime(t *testing.T) {
-	dir := t.TempDir()
-	for _, spec := range []struct{ name, yaml string }{
-		{"runtime-a", "name: runtime-a\nruntimes:\n  - claude-code\n"},
-		{"runtime-b", "name: runtime-b\nruntimes:\n  - hermes\n"},
-		{"universal", "name: universal\nversion: 1.0.0\n"},
-	} {
-		pd := filepath.Join(dir, spec.name)
-		if err := os.Mkdir(pd, 0755); err != nil {
-			t.Fatal(err)
-		}
-		if err := os.WriteFile(filepath.Join(pd, "plugin.yaml"), []byte(spec.yaml), 0600); err != nil {
-			t.Fatal(err)
-		}
-	}
-	h := NewPluginsHandler(dir, nil, nil)
-
-	// Filter to claude-code: runtime-a matches, universal (no runtimes field)
-	// is always included per supportsRuntime semantics.
-	got := h.listRegistryFiltered("claude-code")
-	if len(got) != 2 {
-		t.Fatalf("expected 2 (runtime-a + universal), got %d: %v", len(got), func() []string {
-			ns := make([]string, len(got))
-			for i, p := range got { ns[i] = p.Name }
-			return ns
-		}())
-	}
-}
-
-func TestListRegistry_PluginWithNoRuntimeDeclarations_AlwaysIncluded(t *testing.T) {
-	dir := t.TempDir()
-	pd := filepath.Join(dir, "universal-plugin")
-	if err := os.Mkdir(pd, 0755); err != nil {
-		t.Fatal(err)
-	}
-	if err := os.WriteFile(filepath.Join(pd, "plugin.yaml"), []byte("name: universal-plugin\nversion: 1.0.0\n"), 0600); err != nil {
-		t.Fatal(err)
-	}
-	h := NewPluginsHandler(dir, nil, nil)
-
-	// When plugin declares no runtimes, it should always be included (try-it).
-	got := h.listRegistryFiltered("any-runtime")
-	if len(got) != 1 {
-		t.Errorf("expected 1 plugin (unspecified runtime), got %d", len(got))
-	}
-}
-
-func TestListRegistry_ReadDirError_ReturnsEmpty(t *testing.T) {
-	h := NewPluginsHandler("/nonexistent/path/for/plugins", nil, nil)
-	got := h.listRegistryFiltered("")
-	if len(got) != 0 {
-		t.Errorf("expected empty list on ReadDir error, got %d", len(got))
-	}
-}
-
-func TestListRegistry_HTTPEndpoint(t *testing.T) {
-	dir := t.TempDir()
-	pd := filepath.Join(dir, "test-plugin")
-	if err := os.Mkdir(pd, 0755); err != nil {
-		t.Fatal(err)
-	}
-	if err := os.WriteFile(filepath.Join(pd, "plugin.yaml"), []byte("name: test-plugin\nversion: 2.0.0\n"), 0600); err != nil {
-		t.Fatal(err)
-	}
-	h := NewPluginsHandler(dir, nil, nil)
-
-	w := httptest.NewRecorder()
-	c, _ := gin.CreateTestContext(w)
-	c.Request = httptest.NewRequest("GET", "/plugins", nil)
-	h.ListRegistry(c)
-
-	if w.Code != http.StatusOK {
-		t.Errorf("expected 200, got %d: %s", w.Code, w.Body.String())
-	}
-	var plugins []pluginInfo
-	if err := json.Unmarshal(w.Body.Bytes(), &plugins); err != nil {
-		t.Fatalf("failed to parse JSON: %v", err)
-	}
-	if len(plugins) != 1 {
-		t.Errorf("expected 1 plugin, got %d", len(plugins))
-	}
-	if plugins[0].Name != "test-plugin" {
-		t.Errorf("expected name 'test-plugin', got %q", plugins[0].Name)
-	}
-}

workspace-server/internal/handlers/restart_context.go

@@ -133,24 +133,24 @@ func loadRestartContextData(ctx context.Context, workspaceID string) restartCont
 	// message bus.
 	keySet := map[string]struct{}{}
 	if rows, err := db.DB.QueryContext(ctx, `SELECT key FROM global_secrets`); err == nil {
-		defer rows.Close()
 		for rows.Next() {
 			var k string
 			if rows.Scan(&k) == nil {
 				keySet[k] = struct{}{}
 			}
 		}
+		rows.Close()
 	}
 	if rows, err := db.DB.QueryContext(ctx,
 		`SELECT key FROM workspace_secrets WHERE workspace_id = $1`, workspaceID,
 	); err == nil {
-		defer rows.Close()
 		for rows.Next() {
 			var k string
 			if rows.Scan(&k) == nil {
 				keySet[k] = struct{}{}
 			}
 		}
+		rows.Close()
 	}
 	for k := range keySet {
 		d.EnvKeys = append(d.EnvKeys, k)
@@ -163,8 +163,6 @@ func loadRestartContextData(ctx context.Context, workspaceID string) restartCont
 // workspace's status flips to 'online' or the deadline expires.
 // Returns true on success; callers log+drop on false.
 func waitForWorkspaceOnline(ctx context.Context, workspaceID string, timeout time.Duration) bool {
-	ticker := time.NewTicker(restartContextOnlinePollInterval)
-	defer ticker.Stop()
 	deadline := time.Now().Add(timeout)
 	for time.Now().Before(deadline) {
 		var status string
@@ -176,7 +174,7 @@ func waitForWorkspaceOnline(ctx context.Context, workspaceID string, timeout tim
 		select {
 		case <-ctx.Done():
 			return false
-		case <-ticker.C:
+		case <-time.After(restartContextOnlinePollInterval):
 		}
 	}
 	return false

workspace-server/internal/handlers/runtime_registry.go

@@ -86,40 +86,6 @@ var fallbackRuntimes = map[string]struct{}{
 	"mock": {},
 }
 
-// stripJSON5Comments removes // single-line comments from JSON5-formatted
-// data. The Integration Tester appends "// Triggered by <job>" to
-// manifest.json after cloning, which causes json.Unmarshal to fail with
-// "invalid character '/'". This strips trailing and mid-file comments
-// before parsing so Go's strict JSON parser accepts JSON5 files.
-//
-// Handles:
-//   - Standalone comment lines:    // comment
-//   - Trailing comments:          "key": "value", // comment
-//   - Comments inside strings are NOT touched ("http://example.com")
-func stripJSON5Comments(data []byte) []byte {
-	var result []byte
-	inString := false
-	i := 0
-	for i < len(data) {
-		if data[i] == '"' && (i == 0 || data[i-1] != '\\') {
-			inString = !inString
-			result = append(result, data[i])
-			i++
-			continue
-		}
-		if !inString && i+1 < len(data) && data[i] == '/' && data[i+1] == '/' {
-			// Skip to end of line
-			for i < len(data) && data[i] != '\n' {
-				i++
-			}
-			continue
-		}
-		result = append(result, data[i])
-		i++
-	}
-	return result
-}
-
 // loadRuntimesFromManifest builds the runtime allowlist from
 // manifest.json. Each workspace_templates[].name is normalized to its
 // base runtime identifier (strips the `-default` suffix templates
@@ -135,9 +101,6 @@ func loadRuntimesFromManifest(path string) (map[string]struct{}, error) {
 	if err != nil {
 		return nil, err
 	}
-	// Strip JSON5 // comments before parsing. The Integration Tester
-	// appends "// Triggered by <job>" to manifest.json after cloning.
-	data = stripJSON5Comments(data)
 	var m manifestFile
 	if err := json.Unmarshal(data, &m); err != nil {
 		return nil, err