diff --git a/.gitea/workflows/mcp-external-artifact-mirror.yaml b/.gitea/workflows/mcp-external-artifact-mirror.yaml
index 4017c8f30..1fa6be8da 100644
--- a/.gitea/workflows/mcp-external-artifact-mirror.yaml
+++ b/.gitea/workflows/mcp-external-artifact-mirror.yaml
@@ -3,7 +3,7 @@
 # This Gitea-only lane accepts executable bytes from exact npm registry URLs in
 # committed policy, verifies them, mirrors a scratch OCI data bundle to Harbor,
 # independently reads the digest back without starting a container, then writes
-# the two receipts to Gitea main with a normal non-force push. It never installs
+# the two receipts to a Gitea audit branch with a normal non-force push. It never installs
 # or starts the MCP, opens a browser, writes external content to RAG, or changes a
 # production route.
 
@@ -26,12 +26,13 @@ on:
 
 concurrency:
   group: awoooi-mcp-external-artifact-mirror
-  cancel-in-progress: false
+  cancel-in-progress: true
 
 env:
   GITEA_SOURCE_URL: http://192.168.0.110:3001/wooo/awoooi.git
   HARBOR_REGISTRY: registry.wooo.work
   POLICY_PATH: config/mcp/playwright-mcp-artifact-policy.json
+  RECEIPT_BRANCH: mcp-artifact-receipts
   CONTROLLER_RECEIPT_PATH: docs/operations/external-mcp-artifacts/playwright-mcp-0.0.78-controller.snapshot.json
   VERIFIER_RECEIPT_PATH: docs/operations/external-mcp-artifacts/playwright-mcp-0.0.78-verifier.snapshot.json
 
@@ -185,7 +186,7 @@ jobs:
           print("production_route_changed=false")
           PY
 
-      - name: Commit verified receipts to Gitea main
+      - name: Commit verified receipts to Gitea audit branch
         env:
           CD_PUSH_TOKEN: ${{ secrets.CD_PUSH_TOKEN }}
         run: |
@@ -230,14 +231,18 @@ jobs:
           git remote add gitea "${GITEA_SOURCE_URL}"
 
           for attempt in 1 2 3; do
-            git fetch --no-tags --depth=100 gitea main
-            if ! git merge --no-edit gitea/main; then
-              git merge --abort || true
-              echo "BLOCKER receipt_writeback_merge_conflict"
-              exit 1
+            if git ls-remote --exit-code --heads gitea \
+              "refs/heads/${RECEIPT_BRANCH}" >/dev/null 2>&1; then
+              git fetch --no-tags --depth=100 gitea "${RECEIPT_BRANCH}"
+              if ! git merge --no-edit FETCH_HEAD; then
+                git merge --abort || true
+                echo "BLOCKER receipt_writeback_merge_conflict"
+                exit 1
+              fi
             fi
-            if git push gitea HEAD:main; then
-              echo "receipt_writeback=verified_normal_push"
+            if git push gitea "HEAD:refs/heads/${RECEIPT_BRANCH}"; then
+              echo "receipt_writeback=verified_audit_branch_normal_push"
+              echo "receipt_branch=${RECEIPT_BRANCH}"
               echo "receipt_commit_sha=$(git rev-parse HEAD)"
               exit 0
             fi
diff --git a/apps/api/src/jobs/asset_capability_reconciliation_job.py b/apps/api/src/jobs/asset_capability_reconciliation_job.py
index 8073fbed3..c13545927 100644
--- a/apps/api/src/jobs/asset_capability_reconciliation_job.py
+++ b/apps/api/src/jobs/asset_capability_reconciliation_job.py
@@ -34,7 +34,12 @@ logger = structlog.get_logger(__name__)
 _FIRST_DELAY_SECONDS = 90
 _LOOP_BACKOFF_SECONDS = 30 * 60
 _MAX_RECONCILIATION_CANDIDATES = 20_000
-_RETRYABLE_RESULT_STATUSES = {"blocked_safe_no_write", "degraded_no_write"}
+_RECONCILIATION_LIVE_READBACK_TIMEOUT_SECONDS = 30.0
+_RETRYABLE_RESULT_STATUSES = {
+    "blocked_safe_no_write",
+    "degraded_cache_refresh",
+    "degraded_no_write",
+}
 _DAILY_TRIGGER_HOUR_TAIPEI = 3
 _DAILY_TRIGGER_MINUTE_TAIPEI = 30
 _TAIPEI = ZoneInfo("Asia/Taipei")
@@ -232,6 +237,7 @@ async def reconcile_once(
     started = time.monotonic()
     matrix = await build_asset_capability_matrix_with_live_readback(
         project_id=project_id,
+        timeout_seconds=_RECONCILIATION_LIVE_READBACK_TIMEOUT_SECONDS,
         include_live_only_assets=True,
     )
     if matrix.get("live_readback_status") != "ready":
@@ -264,6 +270,24 @@ async def reconcile_once(
         candidates,
         project_id=project_id,
     )
+    public_matrix = await build_asset_capability_matrix_with_live_readback(
+        project_id=project_id,
+        timeout_seconds=_RECONCILIATION_LIVE_READBACK_TIMEOUT_SECONDS,
+        include_live_only_assets=False,
+        bypass_public_cache=True,
+    )
+    public_cache_refreshed = bool(
+        public_matrix.get("live_readback_status") == "ready"
+        and public_matrix.get("closed") is True
+    )
+    receipt["public_cache_refreshed"] = public_cache_refreshed
+    if not public_cache_refreshed:
+        receipt["status"] = "degraded_cache_refresh"
+        logger.warning(
+            "asset_capability_reconciliation_public_cache_refresh_degraded",
+            live_readback_status=public_matrix.get("live_readback_status"),
+            closed=public_matrix.get("closed"),
+        )
     receipt["duration_ms"] = int((time.monotonic() - started) * 1000)
     logger.info(
         "asset_capability_reconciliation_completed",
diff --git a/apps/api/src/services/ai_automation_asset_capability_matrix.py b/apps/api/src/services/ai_automation_asset_capability_matrix.py
index 2f35dba01..30ab3ac40 100644
--- a/apps/api/src/services/ai_automation_asset_capability_matrix.py
+++ b/apps/api/src/services/ai_automation_asset_capability_matrix.py
@@ -1314,6 +1314,7 @@ async def build_asset_capability_matrix_with_live_readback(
     repo_root: Path | None = None,
     timeout_seconds: float = LIVE_READBACK_TIMEOUT_SECONDS,
     include_live_only_assets: bool = False,
+    bypass_public_cache: bool = False,
 ) -> dict[str, Any]:
     """Build a bounded matrix from live DB truth, failing closed to declared scope."""
 
@@ -1324,7 +1325,7 @@ async def build_asset_capability_matrix_with_live_readback(
     bounded_timeout_seconds = max(0.001, float(timeout_seconds))
     use_public_cache = repo_root is None and not include_live_only_assets
 
-    if use_public_cache:
+    if use_public_cache and not bypass_public_cache:
         cached_payload = await _read_public_matrix_cache(
             project_id,
             namespace=_PUBLIC_MATRIX_FRESH_CACHE_NAMESPACE,
diff --git a/apps/api/src/services/awoooi_priority_work_order_readback.py b/apps/api/src/services/awoooi_priority_work_order_readback.py
index 3b0232322..e85e5fded 100644
--- a/apps/api/src/services/awoooi_priority_work_order_readback.py
+++ b/apps/api/src/services/awoooi_priority_work_order_readback.py
@@ -798,7 +798,7 @@ _AI_AUTOMATION_PROGRAM_WORK_ITEMS: list[dict[str, Any]] = [
         "problem": "Gitea runner 仍會把 actions/checkout 等 action 從 github.com 抓入執行環境，GitHub freeze 尚未覆蓋 CI action resolution。",
         "professional_practice": "self-hosted action registry、immutable digest pinning、SBOM/provenance、network egress verification",
         "asset_scope_ids": ["products", "tools", "packages", "observability_and_security"],
-        "observed_awoooi_gitea_workflow_file_count": 13,
+        "observed_awoooi_gitea_workflow_file_count": 14,
         "observed_awoooi_external_action_reference_count": 18,
         "acceptance": [
             "所有產品的 Gitea workflow action 只由受控 Gitea mirror 或 repo-owned immutable action 供應",
diff --git a/apps/api/src/services/platform_operator_service.py b/apps/api/src/services/platform_operator_service.py
index e891da881..9f88ac96b 100644
--- a/apps/api/src/services/platform_operator_service.py
+++ b/apps/api/src/services/platform_operator_service.py
@@ -7434,6 +7434,206 @@ def _build_awooop_status_chain(
     }
 
 
+def _run_error_envelope(value: Any) -> dict[str, Any]:
+    if isinstance(value, dict):
+        return value
+    if not value:
+        return {}
+    try:
+        parsed = json.loads(str(value))
+    except (TypeError, ValueError, json.JSONDecodeError):
+        return {}
+    return parsed if isinstance(parsed, dict) else {}
+
+
+def _terminal_timestamp(value: Any) -> str | None:
+    if isinstance(value, datetime):
+        return value.isoformat()
+    if value is None:
+        return None
+    rendered = str(value).strip()
+    return rendered or None
+
+
+def _apply_run_terminal_truth_precedence(
+    *,
+    run: AwoooPRunState,
+    status_chain: dict[str, Any] | None,
+) -> dict[str, Any]:
+    """Keep a failed run receipt from being painted green by incident history."""
+
+    chain = dict(status_chain or {})
+    if (
+        str(getattr(run, "state", "")) != "failed"
+        or str(getattr(run, "error_code", ""))
+        != "E-AGENT99-OUTCOME-TIMEOUT"
+    ):
+        return chain
+
+    envelope = _run_error_envelope(getattr(run, "error_detail", None))
+    if envelope.get("outcome_timeout_no_write_terminal") is not True:
+        return chain
+
+    verifier = (
+        envelope.get("verifier")
+        if isinstance(envelope.get("verifier"), dict)
+        else {}
+    )
+    terminal_receipt = (
+        verifier.get("receipt")
+        if isinstance(verifier.get("receipt"), dict)
+        else {}
+    )
+    safe_next_action = str(
+        envelope.get("safe_next_action")
+        or "await_new_source_recurrence_after_agent99_relay_recovery"
+    )
+    historical_incident_context = {
+        "source": chain.get("source"),
+        "source_id": chain.get("source_id"),
+        "incident_ids": chain.get("incident_ids") or [],
+        "current_stage": chain.get("current_stage"),
+        "stage_status": chain.get("stage_status"),
+        "verdict": chain.get("verdict"),
+        "repair_state": chain.get("repair_state"),
+        "verification": chain.get("verification"),
+        "operator_outcome": chain.get("operator_outcome"),
+        "evidence": chain.get("evidence"),
+        "writes": chain.get("writes"),
+    }
+    run_terminal_precedence = {
+        "active": True,
+        "reason": "higher_priority_run_terminal_receipt",
+        "run_id": str(getattr(run, "run_id", "")),
+        "run_state": "failed",
+        "error_code": "E-AGENT99-OUTCOME-TIMEOUT",
+        "outcome_timeout_no_write_terminal": True,
+        "timeout_at": (
+            terminal_receipt.get("timeout_at")
+            or _terminal_timestamp(getattr(run, "timeout_at", None))
+        ),
+        "terminal_at": (
+            terminal_receipt.get("terminal_at")
+            or _terminal_timestamp(getattr(run, "completed_at", None))
+        ),
+        "terminalizer_runtime_write_performed": bool(
+            envelope.get("terminalizer_runtime_write_performed") is True
+        ),
+        "outcome_runtime_write_status": str(
+            envelope.get("outcome_runtime_write_status")
+            or "unknown_without_authenticated_outcome"
+        ),
+        "transport_replayed": bool(envelope.get("transport_replayed") is True),
+        "incident_resolution_authorized": bool(
+            envelope.get("incident_resolution_authorized") is True
+        ),
+        "runtime_closure_verified": bool(
+            envelope.get("runtime_closure_verified") is True
+        ),
+        "safe_next_action": safe_next_action,
+    }
+
+    return {
+        **chain,
+        "source": "run_terminal_receipt>historical_incident_context",
+        "current_stage": "outcome_verifier_failed",
+        "stage_status": "failed",
+        "verdict": "no_runtime_closure_authenticated_outcome_timeout",
+        "repair_state": "failed_no_write_terminal",
+        "verification": "failed_authenticated_outcome_timeout",
+        "needs_human": False,
+        "next_step": safe_next_action,
+        "operator_outcome": {
+            "schema_version": "operator_outcome_run_terminal_v1",
+            "state": "failed_no_write_terminal",
+            "severity": "degraded",
+            "summary_zh": (
+                "Agent99 驗證結果逾時；本 Run 已失敗收斂，"
+                "未宣稱 runtime 修復或事件結案。"
+            ),
+            "needs_human": False,
+            "human_action_required": False,
+            "human_action_reason": None,
+            "next_action": safe_next_action,
+            "execution_result": {
+                "approval_status": "not_applicable",
+                "completion_status": "failed_no_write_terminal",
+                "command_status": "not_replayed",
+                "repair_status": "not_verified",
+                "failure_status": "authenticated_outcome_timeout",
+                "terminal": True,
+                "summary_zh": (
+                    "未取得 authenticated outcome；terminalizer 未執行"
+                    " target runtime write、未重送 transport、未關閉 incident。"
+                ),
+            },
+            "notification": {
+                "mode": "no_notification",
+                "channels": [],
+                "telegram": "not_sent",
+                "awooop": "readback_only",
+            },
+        },
+        "automation_handoff": None,
+        "blockers": ["authenticated_outcome_timeout_no_write"],
+        "evidence": {
+            "auto_repair_records": 0,
+            "operation_records": 0,
+            "mcp_gateway_total": 0,
+            "knowledge_entries": 0,
+            "remediation_total": 0,
+            "remediation_state": "run_terminal_receipt",
+            "latest_route": "--",
+            "latest_mode": None,
+            "latest_at": run_terminal_precedence["terminal_at"],
+            "latest_preview": "authenticated_outcome_timeout_no_write",
+            "ansible_dry_run_only": False,
+        },
+        "writes": {"incident": False, "auto_repair": False},
+        "mcp": {
+            "gateway": {
+                "total": 0,
+                "success": 0,
+                "failed": 0,
+                "blocked": 0,
+                "first_class_total": 0,
+                "legacy_bridge_total": 0,
+                "policy_enforced_total": 0,
+                "stage": "not_applicable_run_terminal",
+                "stage_status": "failed",
+            },
+            "legacy": {"total": 0, "success": 0, "failed": 0},
+            "top_tools": [],
+        },
+        "execution": {
+            "operation_total": 0,
+            "latest_operation_type": "outcome_timeout_no_write_terminal",
+            "latest_status": "failed",
+            "latest_actor": "agent99_controlled_dispatch_reconciler",
+            "latest_action": "record_no_write_terminal",
+            "latest_executor": "none",
+            "playbook_ids": [],
+            "playbook_paths": [],
+            "ansible": {
+                "considered": False,
+                "record_total": 0,
+                "candidate_count": 0,
+                "not_used_reason": "run_terminal_no_runtime_write",
+                "check_mode_total": 0,
+                "apply_total": 0,
+                "rollback_total": 0,
+                "applied": False,
+                "controlled_apply": False,
+            },
+        },
+        "repair_candidate_promotion": (
+            _empty_repair_candidate_promotion_projection()
+        ),
+        "run_terminal_precedence": run_terminal_precedence,
+        "historical_incident_context": historical_incident_context,
+    }
+
+
 def _empty_repair_candidate_promotion_projection() -> dict[str, Any]:
     return {
         "schema_version": "repair_candidate_promotion_projection_v1",
@@ -8254,6 +8454,10 @@ async def get_run_detail(
         project_id=run.project_id,
         remediation_history=remediation_history,
     )
+    awooop_status_chain = _apply_run_terminal_truth_precedence(
+        run=run,
+        status_chain=awooop_status_chain,
+    )
 
     timeline: list[dict[str, Any]] = [
         _timeline_item(
diff --git a/apps/api/tests/test_ai_automation_asset_capability_matrix.py b/apps/api/tests/test_ai_automation_asset_capability_matrix.py
index ad2680394..ac817d4c5 100644
--- a/apps/api/tests/test_ai_automation_asset_capability_matrix.py
+++ b/apps/api/tests/test_ai_automation_asset_capability_matrix.py
@@ -389,6 +389,91 @@ def test_reconciliation_persistence_batches_candidate_queries_and_inserts() -> N
     assert "AND output ->> 'closed' = 'true'" in source
 
 
+def test_reconcile_once_uses_bounded_worker_live_readback_timeout(
+    monkeypatch: Any,
+) -> None:
+    captured: dict[str, object] = {}
+
+    async def _fake_live_readback(**kwargs: object) -> dict[str, object]:
+        captured.update(kwargs)
+        return {"live_readback_status": "degraded"}
+
+    monkeypatch.setattr(
+        reconciliation_job,
+        "build_asset_capability_matrix_with_live_readback",
+        _fake_live_readback,
+    )
+
+    result = asyncio.run(
+        reconciliation_job.reconcile_once(
+            triggered_by="controlled_replay",
+            project_id="awoooi",
+        )
+    )
+
+    assert captured == {
+        "project_id": "awoooi",
+        "timeout_seconds": 30.0,
+        "include_live_only_assets": True,
+    }
+    assert result["status"] == "degraded_no_write"
+
+
+def test_reconcile_once_publishes_closed_public_matrix_cache(
+    monkeypatch: Any,
+) -> None:
+    live_readback_calls: list[dict[str, object]] = []
+
+    async def _fake_live_readback(**kwargs: object) -> dict[str, object]:
+        live_readback_calls.append(dict(kwargs))
+        if kwargs["include_live_only_assets"] is True:
+            return {"live_readback_status": "ready"}
+        return {"live_readback_status": "ready", "closed": True}
+
+    async def _fake_persist(
+        _matrix: object,
+        _candidates: object,
+        *,
+        project_id: str,
+    ) -> dict[str, object]:
+        assert project_id == "awoooi"
+        return {"closed": True, "candidate_count": 0}
+
+    monkeypatch.setattr(
+        reconciliation_job,
+        "build_asset_capability_matrix_with_live_readback",
+        _fake_live_readback,
+    )
+    monkeypatch.setattr(
+        reconciliation_job,
+        "build_reconciliation_candidates",
+        lambda _matrix: [],
+    )
+    monkeypatch.setattr(reconciliation_job, "_persist_reconciliation", _fake_persist)
+
+    result = asyncio.run(
+        reconciliation_job.reconcile_once(
+            triggered_by="controlled_replay",
+            project_id="awoooi",
+        )
+    )
+
+    assert live_readback_calls == [
+        {
+            "project_id": "awoooi",
+            "timeout_seconds": 30.0,
+            "include_live_only_assets": True,
+        },
+        {
+            "project_id": "awoooi",
+            "timeout_seconds": 30.0,
+            "include_live_only_assets": False,
+            "bypass_public_cache": True,
+        },
+    ]
+    assert result["public_cache_refreshed"] is True
+
+
 def test_reconciliation_loop_retries_degraded_result_before_daily_wait(
     monkeypatch: Any,
 ) -> None:
@@ -678,6 +763,37 @@ def test_public_matrix_success_warms_cache_for_next_request(
     assert live_read_count == 1
 
 
+def test_public_matrix_cache_bypass_forces_shared_refresh(monkeypatch: Any) -> None:
+    stored_payloads: list[dict[str, Any]] = []
+
+    async def _unexpected_cache_read(*_args: Any, **_kwargs: Any) -> dict[str, Any]:
+        raise AssertionError("cache bypass must not read a prior projection")
+
+    async def _live_readback(
+        _project_id: str,
+    ) -> tuple[dict[str, Any], list[dict[str, Any]], dict[str, Any]]:
+        return {}, [], {}
+
+    async def _store_cache(_project_id: str, payload: dict[str, Any]) -> None:
+        stored_payloads.append(payload)
+
+    monkeypatch.setattr(
+        matrix_service, "_read_public_matrix_cache", _unexpected_cache_read
+    )
+    monkeypatch.setattr(matrix_service, "_load_live_rows", _live_readback)
+    monkeypatch.setattr(matrix_service, "_store_public_matrix_caches", _store_cache)
+
+    payload = asyncio.run(
+        build_asset_capability_matrix_with_live_readback(
+            bypass_public_cache=True,
+        )
+    )
+
+    assert payload["live_readback_status"] == "ready"
+    assert len(stored_payloads) == 1
+    assert stored_payloads[0]["matrix_fingerprint"] == payload["matrix_fingerprint"]
+
+
 def test_public_matrix_timeout_uses_last_ready_cache(monkeypatch: Any) -> None:
     async def _slow_live_readback(
         _project_id: str,
diff --git a/apps/api/tests/test_awooop_operator_timeline_labels.py b/apps/api/tests/test_awooop_operator_timeline_labels.py
index 33f6a1bc9..cf4bbab07 100644
--- a/apps/api/tests/test_awooop_operator_timeline_labels.py
+++ b/apps/api/tests/test_awooop_operator_timeline_labels.py
@@ -1,5 +1,6 @@
 import asyncio
 import inspect
+import json
 from datetime import datetime
 from decimal import Decimal
 from types import SimpleNamespace
@@ -32,6 +33,7 @@ from src.services.platform_operator_service import (
     _ai_route_lane_state,
     _ai_route_policy_order,
     _ai_route_repair_evidence_item,
+    _apply_run_terminal_truth_precedence,
     _build_awooop_status_chain,
     _callback_reply_audit_summary_from_row,
     _callback_reply_event_item,
@@ -2940,6 +2942,128 @@ def test_awooop_status_chain_marks_verified_repair() -> None:
     assert chain["source_refs"]["refs"]["signoz_alerts"] == ["signoz:abc"]
 
 
+def test_run_terminal_receipt_overrides_historical_incident_green_state() -> None:
+    historical_chain = {
+        "schema_version": "awooop_status_chain_v1",
+        "source": "truth_chain+adr100_history",
+        "source_id": "INC-BRR-35994B66C2807E74",
+        "incident_ids": ["INC-BRR-35994B66C2807E74"],
+        "current_stage": "execution_succeeded",
+        "stage_status": "success",
+        "verdict": "auto_repaired_verified",
+        "repair_state": "auto_repaired_verified",
+        "verification": "success",
+        "needs_human": False,
+        "next_step": "monitor_for_regression",
+        "operator_outcome": {
+            "state": "completed_verified",
+            "execution_result": {
+                "completion_status": "completed_verified",
+                "terminal": True,
+            },
+        },
+        "evidence": {
+            "auto_repair_records": 1,
+            "operation_records": 4,
+            "mcp_gateway_total": 8,
+            "knowledge_entries": 1,
+        },
+        "writes": {"incident": True, "auto_repair": True},
+        "mcp": {"gateway": {"total": 8, "success": 4}},
+        "execution": {"operation_total": 4, "latest_status": "success"},
+    }
+    run = SimpleNamespace(
+        run_id=UUID("f53d95ac-04d3-5228-ae74-9ab0ad95c013"),
+        state="failed",
+        error_code="E-AGENT99-OUTCOME-TIMEOUT",
+        timeout_at=datetime.fromisoformat("2026-07-15T00:22:32.527349"),
+        completed_at=datetime.fromisoformat("2026-07-15T01:23:15.118099"),
+        error_detail=json.dumps({
+            "outcome_timeout_no_write_terminal": True,
+            "terminalizer_runtime_write_performed": False,
+            "outcome_runtime_write_status": (
+                "unknown_without_authenticated_outcome"
+            ),
+            "transport_replayed": False,
+            "incident_resolution_authorized": False,
+            "runtime_closure_verified": False,
+            "safe_next_action": (
+                "await_new_source_recurrence_after_agent99_relay_recovery"
+            ),
+            "verifier": {
+                "receipt": {
+                    "timeout_at": "2026-07-15T00:22:32.527349",
+                    "terminal_at": "2026-07-15T01:23:15.118099",
+                }
+            },
+        }),
+    )
+
+    chain = _apply_run_terminal_truth_precedence(
+        run=run,
+        status_chain=historical_chain,
+    )
+
+    assert chain["source"] == (
+        "run_terminal_receipt>historical_incident_context"
+    )
+    assert chain["stage_status"] == "failed"
+    assert chain["repair_state"] == "failed_no_write_terminal"
+    assert chain["verification"] == "failed_authenticated_outcome_timeout"
+    assert chain["needs_human"] is False
+    assert chain["operator_outcome"]["state"] == "failed_no_write_terminal"
+    assert chain["operator_outcome"]["execution_result"][
+        "completion_status"
+    ] == "failed_no_write_terminal"
+    assert chain["writes"] == {"incident": False, "auto_repair": False}
+    assert chain["evidence"]["auto_repair_records"] == 0
+    assert chain["execution"]["operation_total"] == 0
+    assert chain["run_terminal_precedence"] == {
+        "active": True,
+        "reason": "higher_priority_run_terminal_receipt",
+        "run_id": "f53d95ac-04d3-5228-ae74-9ab0ad95c013",
+        "run_state": "failed",
+        "error_code": "E-AGENT99-OUTCOME-TIMEOUT",
+        "outcome_timeout_no_write_terminal": True,
+        "timeout_at": "2026-07-15T00:22:32.527349",
+        "terminal_at": "2026-07-15T01:23:15.118099",
+        "terminalizer_runtime_write_performed": False,
+        "outcome_runtime_write_status": (
+            "unknown_without_authenticated_outcome"
+        ),
+        "transport_replayed": False,
+        "incident_resolution_authorized": False,
+        "runtime_closure_verified": False,
+        "safe_next_action": (
+            "await_new_source_recurrence_after_agent99_relay_recovery"
+        ),
+    }
+    assert chain["historical_incident_context"]["repair_state"] == (
+        "auto_repaired_verified"
+    )
+    assert chain["historical_incident_context"]["writes"] == {
+        "incident": True,
+        "auto_repair": True,
+    }
+    assert "_apply_run_terminal_truth_precedence" in inspect.getsource(
+        platform_operator_service.get_run_detail
+    )
+
+
+def test_run_terminal_precedence_leaves_unrelated_failed_run_unchanged() -> None:
+    chain = {"repair_state": "auto_repaired_verified"}
+    run = SimpleNamespace(
+        state="failed",
+        error_code="E-OTHER",
+        error_detail="{}",
+    )
+
+    assert _apply_run_terminal_truth_precedence(
+        run=run,
+        status_chain=chain,
+    ) == chain
+
+
 def test_awooop_status_chain_surfaces_controlled_ansible_apply_proof() -> None:
     chain = _build_awooop_status_chain(
         incident_ids=["INC-20260531-D6A3C4"],
diff --git a/apps/api/tests/test_gitea_actions_secret_variable_name_inventory_api.py b/apps/api/tests/test_gitea_actions_secret_variable_name_inventory_api.py
index bdaea07ac..ba34851c5 100644
--- a/apps/api/tests/test_gitea_actions_secret_variable_name_inventory_api.py
+++ b/apps/api/tests/test_gitea_actions_secret_variable_name_inventory_api.py
@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+from pathlib import Path
+
 from fastapi import FastAPI
 from fastapi.testclient import TestClient
 
@@ -11,13 +13,20 @@ from src.services.gitea_actions_secret_variable_name_inventory import (
 
 def test_gitea_actions_secret_variable_name_inventory_loader_is_value_safe() -> None:
     payload = load_latest_gitea_actions_secret_variable_name_inventory()
+    repo_root = Path(__file__).resolve().parents[3]
+    workflow_dir = repo_root / ".gitea" / "workflows"
+    committed_workflow_file_count = sum(
+        1
+        for path in workflow_dir.iterdir()
+        if path.is_file() and path.suffix.lower() in {".yml", ".yaml"}
+    )
 
     assert (
         payload["schema_version"] == "gitea_actions_secret_variable_name_inventory_v1"
     )
     assert payload["status"] == "gitea_actions_secret_variable_name_inventory_ready"
     assert payload["ready"] is True
-    assert payload["summary"]["workflow_file_count"] == 13
+    assert payload["summary"]["workflow_file_count"] == committed_workflow_file_count
     assert payload["summary"]["referenced_secret_name_count"] == 26
     assert payload["summary"]["referenced_variable_name_count"] == 0
     assert payload["summary"]["secret_values_collected"] is False
diff --git a/apps/api/tests/test_gitea_capability_ci_cd_optimization_readback_api.py b/apps/api/tests/test_gitea_capability_ci_cd_optimization_readback_api.py
index 4448216bf..09fe07d73 100644
--- a/apps/api/tests/test_gitea_capability_ci_cd_optimization_readback_api.py
+++ b/apps/api/tests/test_gitea_capability_ci_cd_optimization_readback_api.py
@@ -139,6 +139,9 @@ def _readback_payload() -> dict:
 
 def test_gitea_capability_readback_rolls_up_complete_feature_surface() -> None:
     payload = _readback_payload()
+    workflow_inventory = payload["workflow_source_inventory"]
+    workflow_rows = workflow_inventory["workflow_rows"]
+    workflow_file_count = len(workflow_rows)
 
     assert (
         payload["schema_version"] == "gitea_capability_ci_cd_optimization_readback_v1"
@@ -148,20 +151,53 @@ def test_gitea_capability_readback_rolls_up_complete_feature_surface() -> None:
     assert payload["rollups"]["expected_product_count"] == 12
     assert payload["rollups"]["ssh_verified_product_repo_count"] == 12
     assert payload["rollups"]["dev_prod_environment_split_ready_count"] == 12
-    assert payload["workflow_source_inventory"]["workflow_file_count"] == 13
-    professional_feature_rollups = payload["workflow_source_inventory"][
+    assert workflow_inventory["workflow_file_count"] == workflow_file_count
+    professional_feature_rollups = workflow_inventory[
         "professional_feature_rollups"
     ]
-    assert professional_feature_rollups["gitea_ignored_syntax_workflow_count"] == 10
-    assert professional_feature_rollups["otel_workflow_count"] == 3
-    assert professional_feature_rollups["argocd_gitops_workflow_count"] == 1
-    assert professional_feature_rollups["notification_or_webhook_workflow_count"] == 9
-    assert (
-        professional_feature_rollups["registry_or_package_channel_workflow_count"] == 4
+    ignored_syntax_keys = {
+        "declares_concurrency",
+        "declares_permissions",
+        "declares_timeout_minutes",
+        "declares_continue_on_error",
+        "declares_environment",
+    }
+    assert professional_feature_rollups[
+        "gitea_ignored_syntax_workflow_count"
+    ] == sum(
+        any(row.get(key) is True for key in ignored_syntax_keys)
+        for row in workflow_rows
+    )
+    assert professional_feature_rollups["otel_workflow_count"] == sum(
+        row.get("uses_otel") is True for row in workflow_rows
+    )
+    assert professional_feature_rollups["argocd_gitops_workflow_count"] == sum(
+        row.get("uses_argocd_gitops") is True for row in workflow_rows
+    )
+    assert professional_feature_rollups[
+        "notification_or_webhook_workflow_count"
+    ] == sum(
+        row.get("uses_notification_or_webhook") is True for row in workflow_rows
+    )
+    assert (
+        professional_feature_rollups["registry_or_package_channel_workflow_count"]
+        == sum(
+            row.get("uses_registry_or_package_channel") is True
+            for row in workflow_rows
+        )
+    )
+    assert professional_feature_rollups[
+        "package_release_promotion_workflow_count"
+    ] == sum(
+        row.get("uses_package_release_promotion_contract") is True
+        for row in workflow_rows
+    )
+    assert professional_feature_rollups["sbom_or_provenance_workflow_count"] == sum(
+        row.get("uses_sbom_or_provenance") is True for row in workflow_rows
+    )
+    assert professional_feature_rollups["artifact_transfer_workflow_count"] == sum(
+        row.get("uses_artifact_transfer") is True for row in workflow_rows
     )
-    assert professional_feature_rollups["package_release_promotion_workflow_count"] == 1
-    assert professional_feature_rollups["sbom_or_provenance_workflow_count"] == 1
-    assert professional_feature_rollups["artifact_transfer_workflow_count"] == 1
     assert (
         payload["workflow_source_inventory"]["secret_variable_rollups"][
             "secret_values_collected"
@@ -203,8 +239,8 @@ def test_gitea_capability_readback_rolls_up_complete_feature_surface() -> None:
     assert capability_by_id["wiki_knowledge_base"]["status"] == "source_ready"
     assert capability_by_id["gitea_actions_workflows"]["status"] == "verified_ready"
     assert capability_by_id["gitea_actions_workflows"]["metrics"] == {
-        "workflow_source_file_count": 13,
-        "workflow_contract_count": 13,
+        "workflow_source_file_count": workflow_file_count,
+        "workflow_contract_count": workflow_file_count,
     }
     assert capability_by_id["gitea_actions_runners"]["status"] == "verified_ready"
     assert (
diff --git a/apps/api/tests/test_gitea_workflow_runner_health_api.py b/apps/api/tests/test_gitea_workflow_runner_health_api.py
index ee9734a94..a65eb96fe 100644
--- a/apps/api/tests/test_gitea_workflow_runner_health_api.py
+++ b/apps/api/tests/test_gitea_workflow_runner_health_api.py
@@ -20,9 +20,14 @@ def test_gitea_workflow_runner_health_endpoint_returns_committed_snapshot():
     assert data["program_status"]["current_task_id"] == "P1-002"
     assert data["program_status"]["next_task_id"] == "P1-003"
     assert data["program_status"]["read_only_mode"] is True
-    assert data["rollups"]["total_workflows"] == len(data["workflow_records"]) == 13
-    assert data["rollups"]["workflows_with_schedule"] == 5
-    assert data["rollups"]["workflows_with_workflow_dispatch"] == 13
+    workflow_records = data["workflow_records"]
+    assert data["rollups"]["total_workflows"] == len(workflow_records)
+    assert data["rollups"]["workflows_with_schedule"] == sum(
+        "schedule" in row["triggers"] for row in workflow_records
+    )
+    assert data["rollups"]["workflows_with_workflow_dispatch"] == sum(
+        "workflow_dispatch" in row["triggers"] for row in workflow_records
+    )
     assert data["rollups"]["workflow_ids_requiring_runner_attestation"] == []
     assert data["rollups"]["runner_contracts_requiring_action"] == []
     assert data["rollups"]["notification_contracts_quiet_success_count"] == 3
@@ -64,5 +69,16 @@ def test_gitea_workflow_runner_health_endpoint_returns_committed_snapshot():
         if row["workflow_id"] == "harbor_110_local_repair"
     )
     assert harbor_repair["runner_evidence_status"] == "non110_host_runner_mapped"
+    mcp_artifact_mirror = next(
+        row
+        for row in workflow_records
+        if row["workflow_id"] == "mcp_external_artifact_mirror"
+    )
+    assert mcp_artifact_mirror["runner_labels"] == ["awoooi-non110-host"]
+    assert mcp_artifact_mirror["runner_evidence_status"] == (
+        "non110_host_runner_mapped"
+    )
+    assert mcp_artifact_mirror["notification_policy"] == "read_only_no_notify"
+    assert mcp_artifact_mirror["notify_bridge_calls"] == 0
     assert "workflow 修改批准" in data["operator_contract"]["must_not_interpret_as"]
     assert "Secret 已讀取或可輸出" in data["operator_contract"]["must_not_interpret_as"]
diff --git a/apps/web/src/components/awooop/status-chain.tsx b/apps/web/src/components/awooop/status-chain.tsx
index 448ee8cba..4f358eae1 100644
--- a/apps/web/src/components/awooop/status-chain.tsx
+++ b/apps/web/src/components/awooop/status-chain.tsx
@@ -18,6 +18,22 @@ export interface AwoooPStatusChain {
   verification?: string | null;
   needs_human?: boolean | null;
   next_step?: string | null;
+  run_terminal_precedence?: {
+    active?: boolean | null;
+    reason?: string | null;
+    run_id?: string | null;
+    run_state?: string | null;
+    error_code?: string | null;
+    outcome_timeout_no_write_terminal?: boolean | null;
+    timeout_at?: string | null;
+    terminal_at?: string | null;
+    terminalizer_runtime_write_performed?: boolean | null;
+    outcome_runtime_write_status?: string | null;
+    transport_replayed?: boolean | null;
+    incident_resolution_authorized?: boolean | null;
+    runtime_closure_verified?: boolean | null;
+    safe_next_action?: string | null;
+  } | null;
   operator_outcome?: {
     schema_version?: string;
     state?: string | null;
@@ -466,6 +482,13 @@ export interface AwoooPStatusChain {
 
 function toneClass(chain?: AwoooPStatusChain | null) {
   if (!chain) return "border-[#d8d3c7] bg-[#faf9f3] text-[#5f5b52]";
+  if (
+    chain.run_terminal_precedence?.active === true
+    || chain.stage_status === "failed"
+    || String(chain.repair_state ?? "").startsWith("failed_")
+  ) {
+    return "border-[#e2a29b] bg-[#fff0ef] text-[#9f2f25]";
+  }
   if (chain.repair_state === "auto_repaired_verified") {
     return "border-[#9bc7a4] bg-[#f0faf2] text-[#17602a]";
   }
@@ -477,6 +500,11 @@ function toneClass(chain?: AwoooPStatusChain | null) {
 
 function chainIcon(chain?: AwoooPStatusChain | null) {
   if (!chain) return Route;
+  if (
+    chain.run_terminal_precedence?.active === true
+    || chain.stage_status === "failed"
+    || String(chain.repair_state ?? "").startsWith("failed_")
+  ) return TriangleAlert;
   if (chain.repair_state === "auto_repaired_verified") return CheckCircle2;
   if (chain.needs_human || chain.repair_state === "blocked_manual_required") return ShieldCheck;
   return Activity;
@@ -557,6 +585,7 @@ export function AwoooPStatusChainPanel({
       approve_post_apply_verifier_plan: t("nextActions.reviewPostApplyVerifier"),
       assign_km_writeback_owner_after_verified_execution: t("nextActions.prepareKmPlaybookWriteback"),
       assign_playbook_trust_owner_after_verified_execution: t("nextActions.prepareKmPlaybookWriteback"),
+      await_new_source_recurrence_after_agent99_relay_recovery: t("nextActions.collectEvidenceOrWait"),
     };
     return labels[key] ?? valueOrEmpty(value, emptyLabel);
   };
diff --git a/config/security/runtime-image-mirror-policy.json b/config/security/runtime-image-mirror-policy.json
index 4a75d9c28..d1ba157dd 100644
--- a/config/security/runtime-image-mirror-policy.json
+++ b/config/security/runtime-image-mirror-policy.json
@@ -278,6 +278,29 @@
           "max_surge": 1
         }
       }
+    },
+    {
+      "asset_id": "runtime-image:sealed-secrets-controller",
+      "source_index_digest": "sha256:74cd190f424b591dd82a234685c8c81d50b33af807e03470bc12b23d202e0106",
+      "platform_digest": "sha256:1827b36853e124ac0dd2554f7cd55c04952ecb7c38ff19a4a8d93d633811ed68",
+      "target_registry_digest": "sha256:2692460ed95f213760550cb508c881e1490a2f0ce4affbed31a0bf900ebba363",
+      "push_ref": "registry.wooo.work/awoooi/runtime-mirror/sealed-secrets-controller:0.26.0-linux-amd64",
+      "runtime_ref": "192.168.0.110:5000/awoooi/runtime-mirror/sealed-secrets-controller@sha256:2692460ed95f213760550cb508c881e1490a2f0ce4affbed31a0bf900ebba363",
+      "workload": {
+        "kind": "deployment",
+        "namespace": "kube-system",
+        "name": "sealed-secrets-controller",
+        "container": "sealed-secrets-controller"
+      },
+      "availability_guard": {
+        "minimum_ready_replicas": 1,
+        "maximum_effective_unavailable": 0,
+        "minimum_effective_surge": 1,
+        "enforced_strategy": {
+          "max_unavailable": 0,
+          "max_surge": 1
+        }
+      }
     }
   ]
 }
diff --git a/docs/LOGBOOK.md b/docs/LOGBOOK.md
index 8f4eb97c4..bd7166202 100644
--- a/docs/LOGBOOK.md
+++ b/docs/LOGBOOK.md
@@ -1,3 +1,25 @@
+## 2026-07-15 — P0 MCP Playwright immutable mirror 與獨立供應鏈 verifier
+
+**完成內容**：
+- mcp.so 僅保留 discovery URL；可執行 bytes 只從 policy allowlist 的 npm registry exact URLs 取得。`@playwright/mcp=0.0.78`、`playwright` / `playwright-core=1.62.0-alpha-1783623505000` 全部固定 version、SRI SHA-512、SHA-1、npm ECDSA signature、SLSA subject、dependency closure 與 Apache-2.0 license。
+- 新增 fail-closed artifact controller：拒絕 redirect、非 allowlist host、floating version、dependency/metadata drift、unsafe/duplicate tar entry、已知 OSV vulnerability 與未通過簽章的 artifact；產生 deterministic CycloneDX 1.5 SBOM 與 scratch OCI data bundle，不執行 package install、MCP、browser 或 container。
+- 新增完全不 import controller 的獨立 verifier：由 Harbor immutable digest 重新 pull/save image、只把 layer 當資料讀取，限制 file/byte/path/symlink/whiteout，要求 exact 8-file bundle，再次重算 manifest/SBOM/tarball hash、以 OpenSSL 重驗 npm signature 並重解 SLSA DSSE subject。
+- Gitea `mcp-external-artifact-mirror.yaml` 不使用遠端 action；push、manual 與每週三 pinned revalidation 都先跑 controller/verifier tests、host-pressure gate、Harbor password-stdin、獨立 readback與 logout。receipt 以 ephemeral askpass normal push 到專用 `mcp-artifact-receipts` audit branch，不再推 main、取消有效 CD 或保存 credential。
+- Gitea run `#5174` 成功；controlled run `dce032df-0a9a-476f-940b-e61a94865e1b` 產生 internal digest `sha256:de002c418bbb94fb4609539c4259c39ccb7f8b353f8bac8fbcba9e5742ad79ed`，獨立 verifier 為 `completed_internal_mirror_verified`，audit branch commit `9c08ada5c0f15bb6924304af29cd13c8376182ff`。
+
+**source/test evidence**：
+- artifact controller/verifier focused tests `24 passed`；MCP control-plane/federation/version-lifecycle `20 passed`。
+- 新增第 14 條 Gitea workflow 後，CD `#5173` 以 `2627 passed / 1 failed` 找出 workflow health contract 仍固定 13；已同步 workflow health snapshot、secret-name inventory、Gitea capability readback 與 priority work-order count，聚焦回歸 `40 passed`，secret names 維持 `26`、secret values collected=`false`。
+- Ruff、py_compile、JSON/YAML parse、freeze/floating-reference guard 與 `git diff --check` 通過。
+
+**尚未完成 / 不誤報**：
+- 目前只完成 immutable internal mirror 與供應鏈 verifier；`deployment_allowed=false`、`replay_allowed=false`、`shadow_allowed=false`、`canary_allowed=false`。尚未實作 registered public-origin replay adapter、compatibility replay、shadow/canary、runtime upgrade、post-verifier 或 rollback execution。
+- 外部 catalog / tarball / attestation 未寫 RAG；只保留 normalized digest、SBOM 與 supply-chain receipt。未讀 secret value、`.env`、raw session、SQLite/auth；未連線或下載 GitHub / Actions / hosted artifact，也未使用 `npx -y`、`@latest`、`:latest`、force push 或 production route mutation。
+- 本筆 inventory drift 修正與 catalog mirror state 尚待最新 main CD/deploy marker/production UI readback；不得由 `#5174` mirror success 倒推 AWOOOI production 已部署。
+
+**下一步**：
+- 先取得包含本修正的 Gitea CD terminal與 production MCP page readback；再以 public-only origin allowlist 建立 repo-owned replay adapter，要求 compatibility、prompt-injection、latency、accessibility、session/file/download no-write verifier與 rollback receipt 全數通過，才可進第一個 noncritical shadow/canary。
+
 ## 2026-07-15 — P0 MCP EwoooC/MOMO durable production federation source closure
 
 **完成內容**：
diff --git a/docs/evaluations/gitea_workflow_runner_health_2026-06-05.json b/docs/evaluations/gitea_workflow_runner_health_2026-06-05.json
index 694f595bd..fe7b0fd49 100644
--- a/docs/evaluations/gitea_workflow_runner_health_2026-06-05.json
+++ b/docs/evaluations/gitea_workflow_runner_health_2026-06-05.json
@@ -1,6 +1,6 @@
 {
   "schema_version": "gitea_workflow_runner_health_v1",
-  "generated_at": "2026-07-09T23:20:00+08:00",
+  "generated_at": "2026-07-15T09:57:22+08:00",
   "program_status": {
     "overall_completion_percent": 100,
     "current_priority": "P1",
@@ -21,8 +21,10 @@
     ".gitea/workflows/deploy-alerts.yaml",
     ".gitea/workflows/e2e-health.yaml",
     ".gitea/workflows/harbor-110-local-repair.yaml",
+    ".gitea/workflows/mcp-external-artifact-mirror.yaml",
     ".gitea/workflows/run-migration.yml",
     ".gitea/workflows/type-sync-check.yaml",
+    "config/mcp/playwright-mcp-artifact-policy.json",
     "docs/evaluations/ai_agent_version_lifecycle_update_proposal_2026-07-10.json",
     "scripts/ci/check-gitea-step-env-secrets.js",
     "scripts/ci/cleanup-host-runner-workspace.sh",
@@ -32,22 +34,22 @@
     "scripts/ops/stop-stale-gitea-actions-jobs.sh"
   ],
   "rollups": {
-    "total_workflows": 13,
+    "total_workflows": 14,
     "by_workflow_status": {
-      "manifest_mapped": 13
+      "manifest_mapped": 14
     },
     "by_runner_evidence_status": {
       "non110_runner_mapped": 10,
-      "non110_host_runner_mapped": 3
+      "non110_host_runner_mapped": 4
     },
-    "workflows_with_schedule": 5,
-    "workflows_with_workflow_dispatch": 13,
+    "workflows_with_schedule": 6,
+    "workflows_with_workflow_dispatch": 14,
     "workflows_with_notify_bridge": 6,
     "workflows_with_actionable_or_failure_quiet_policy": 3,
     "workflow_ids_requiring_runner_attestation": [],
     "total_runner_contracts": 5,
     "runner_contracts_requiring_action": [],
-    "notification_contracts_total": 7,
+    "notification_contracts_total": 8,
     "notification_contracts_quiet_success_count": 3,
     "notification_contracts_quiet_success_ids": [
       "agent_market_watch_actionable_only",
@@ -337,6 +339,36 @@
       ],
       "next_action": "維持 no-auto-apply repair guard；不得由 API 直接觸發 repair。"
     },
+    {
+      "workflow_id": "mcp_external_artifact_mirror",
+      "file_ref": ".gitea/workflows/mcp-external-artifact-mirror.yaml",
+      "display_name": "MCP External Artifact Mirror",
+      "scope": "以 exact npm registry URL、committed SHA-512 與 Harbor digest 執行受控外部 MCP artifact mirror；只寫獨立 audit branch，不啟動 MCP、不改 production route。",
+      "status": "manifest_mapped",
+      "risk_level": "high",
+      "triggers": [
+        "push:main",
+        "workflow_dispatch",
+        "schedule"
+      ],
+      "schedule_cadence": "每週三 10:13 Asia/Taipei；cron=13 2 * * 3 UTC",
+      "runner_labels": [
+        "awoooi-non110-host"
+      ],
+      "runner_evidence_status": "non110_host_runner_mapped",
+      "job_count": 1,
+      "notification_policy": "read_only_no_notify",
+      "notify_bridge_calls": 0,
+      "secrets_policy_status": "只使用 workflow secret references 進行 Harbor 登入與獨立 audit branch normal push；本 snapshot 與 readback 不讀取、保存或顯示任何值。",
+      "evidence_refs": [
+        ".gitea/workflows/mcp-external-artifact-mirror.yaml",
+        "config/mcp/playwright-mcp-artifact-policy.json",
+        "scripts/security/external_mcp_artifact_controller.py",
+        "scripts/security/verify_external_mcp_artifact_receipt.py",
+        "scripts/ci/wait-host-web-build-pressure.sh"
+      ],
+      "next_action": "維持 bounded capacity wait、獨立 verifier 與 mcp-artifact-receipts audit branch；不得啟動 artifact、寫 RAG、切 production route 或推回 deploy main。"
+    },
     {
       "workflow_id": "run_migration",
       "file_ref": ".gitea/workflows/run-migration.yml",
@@ -399,7 +431,8 @@
       "used_by_workflows": [
         "awoooi_onboarding_warning_step",
         "cd_pipeline",
-        "harbor_110_local_repair"
+        "harbor_110_local_repair",
+        "mcp_external_artifact_mirror"
       ],
       "health_contract": "host-level controlled workflows use the explicit awoooi-non110-host label; this snapshot does not restart or mutate the runner.",
       "guardrail_refs": [
@@ -409,7 +442,8 @@
       "evidence_refs": [
         ".gitea/workflows/cd.yaml",
         ".gitea/workflows/harbor-110-local-repair.yaml",
-        ".gitea/workflows/awoooi-onboarding-warning-step.yaml"
+        ".gitea/workflows/awoooi-onboarding-warning-step.yaml",
+        ".gitea/workflows/mcp-external-artifact-mirror.yaml"
       ],
       "next_action": "維持 explicit non110 host label；任何 runner/service 變更另走維護窗口。"
     },
@@ -613,6 +647,22 @@
       ],
       "next_action": "後續補 alert/repair receipt；本 snapshot 不發通知、不觸發 migration 或 repair。"
     },
+    {
+      "contract_id": "mcp_artifact_mirror_no_notify",
+      "display_name": "MCP artifact mirror no-notify",
+      "status": "preserved",
+      "policy_kind": "read_only_no_notify",
+      "success_noise_policy": "artifact mirror 與 verifier 成功只寫獨立 audit branch，不向 Telegram 或產品告警群組發送成功訊息。",
+      "failure_policy": "失敗留在 Gitea workflow 與脫敏 receipt；不得因通知失敗而重送 artifact、切 production route 或啟動 MCP。",
+      "workflow_refs": [
+        "mcp_external_artifact_mirror"
+      ],
+      "evidence_refs": [
+        ".gitea/workflows/mcp-external-artifact-mirror.yaml",
+        "config/mcp/playwright-mcp-artifact-policy.json"
+      ],
+      "next_action": "維持 no-notify 與獨立 audit branch；只在受控營運 readback 顯示 terminal receipt。"
+    },
     {
       "contract_id": "lint_and_typecheck_no_notify",
       "display_name": "Lint / typecheck no-notify",
diff --git a/docs/operations/mcp-control-plane-catalog.snapshot.json b/docs/operations/mcp-control-plane-catalog.snapshot.json
index 312253008..7b5d6576b 100644
--- a/docs/operations/mcp-control-plane-catalog.snapshot.json
+++ b/docs/operations/mcp-control-plane-catalog.snapshot.json
@@ -301,18 +301,21 @@
       "data_boundary": "public_and_test_surfaces",
       "deployment_allowed": false,
       "version_state": "exact_0.0.78_check_verified",
-      "digest_state": "upstream_sha512_pinned_internal_mirror_pending",
-      "sbom_state": "cyclonedx_1.5_check_verified_internal_mirror_pending",
-      "signature_state": "npm_registry_ecdsa_and_slsa_subject_verified",
+      "digest_state": "internal_mirror_digest_verified_sha256_de002c418bbb94fb4609539c4259c39ccb7f8b353f8bac8fbcba9e5742ad79ed",
+      "sbom_state": "cyclonedx_1.5_mirror_readback_verified",
+      "signature_state": "npm_registry_ecdsa_and_slsa_subject_independently_reverified",
       "rag_policy": "screenshots_and_accessibility_receipts_only",
       "blockers": [
-        "internal_mirror_and_independent_digest_readback_pending",
         "registered_public_origin_replay_adapter_missing",
+        "compatibility_replay_not_executed",
         "shadow_canary_verifier_and_rollback_execution_pending"
       ],
       "evidence_refs": [
         "config/mcp/playwright-mcp-artifact-policy.json",
-        "docs/operations/external-mcp-artifacts/playwright-mcp-0.0.78-check.snapshot.json"
+        "docs/operations/external-mcp-artifacts/playwright-mcp-0.0.78-check.snapshot.json",
+        "gitea:mcp-artifact-receipts@9c08ada5c0f15bb6924304af29cd13c8376182ff:docs/operations/external-mcp-artifacts/playwright-mcp-0.0.78-controller.snapshot.json",
+        "gitea:mcp-artifact-receipts@9c08ada5c0f15bb6924304af29cd13c8376182ff:docs/operations/external-mcp-artifacts/playwright-mcp-0.0.78-verifier.snapshot.json",
+        "registry.wooo.work/awoooi/mcp-artifacts/playwright-mcp@sha256:de002c418bbb94fb4609539c4259c39ccb7f8b353f8bac8fbcba9e5742ad79ed"
       ]
     },
     {
diff --git a/scripts/security/tests/test_external_mcp_artifact_controller.py b/scripts/security/tests/test_external_mcp_artifact_controller.py
index cfce6178c..273b68ef6 100644
--- a/scripts/security/tests/test_external_mcp_artifact_controller.py
+++ b/scripts/security/tests/test_external_mcp_artifact_controller.py
@@ -123,13 +123,16 @@ class ExternalMcpArtifactControllerTest(unittest.TestCase):
         self.assertNotIn("@latest", lowered)
         self.assertNotIn(":latest", lowered)
         self.assertNotIn("git push --force", lowered)
+        self.assertIn("cancel-in-progress: true", raw)
         self.assertIn("runs-on: awoooi-non110-host", raw)
         self.assertIn("docker login", raw)
         self.assertIn("--password-stdin", raw)
         self.assertIn("docker logout", raw)
         self.assertIn("verify_external_mcp_artifact_receipt.py", raw)
-        self.assertIn("git merge --no-edit gitea/main", raw)
-        self.assertIn("git push gitea HEAD:main", raw)
+        self.assertIn("RECEIPT_BRANCH: mcp-artifact-receipts", raw)
+        self.assertIn('git merge --no-edit FETCH_HEAD', raw)
+        self.assertIn('HEAD:refs/heads/${RECEIPT_BRANCH}', raw)
+        self.assertNotIn("git push gitea HEAD:main", raw)
         self.assertIn('cron: "13 2 * * 3"', raw)
 
     def test_dependency_closure_rejects_drift(self) -> None:
diff --git a/scripts/security/tests/test_runtime_image_mirror_controller.py b/scripts/security/tests/test_runtime_image_mirror_controller.py
index 8107e2496..e5ec0145c 100644
--- a/scripts/security/tests/test_runtime_image_mirror_controller.py
+++ b/scripts/security/tests/test_runtime_image_mirror_controller.py
@@ -81,7 +81,7 @@ class RuntimeImageMirrorControllerTest(unittest.TestCase):
 
         self.assertEqual(policy.work_item_id, "AIA-P0-006-02A")
         self.assertEqual(policy.risk_level, "high")
-        self.assertEqual(len(policy.images), 17)
+        self.assertEqual(len(policy.images), 18)
         self.assertNotIn("github.com", raw)
         self.assertNotIn("ghcr.io", raw)
         self.assertIn('"external_pull_allowed": false', raw)
@@ -106,7 +106,7 @@ class RuntimeImageMirrorControllerTest(unittest.TestCase):
             )
         )
         container_kinds = [image.workload.container_kind for image in policy.images]
-        self.assertEqual(container_kinds.count("container"), 14)
+        self.assertEqual(container_kinds.count("container"), 15)
         self.assertEqual(container_kinds.count("init_container"), 3)
 
         vpa = next(
@@ -176,6 +176,31 @@ class RuntimeImageMirrorControllerTest(unittest.TestCase):
             ),
         )
 
+        sealed_secrets = next(
+            image
+            for image in policy.images
+            if image.asset_id == "runtime-image:sealed-secrets-controller"
+        )
+        self.assertEqual(
+            sealed_secrets.target_registry_digest,
+            "sha256:2692460ed95f213760550cb508c881e1490a2f0ce4affbed31a0bf900ebba363",
+        )
+        self.assertEqual(sealed_secrets.workload.namespace, "kube-system")
+        self.assertEqual(sealed_secrets.workload.name, "sealed-secrets-controller")
+        self.assertEqual(
+            sealed_secrets.workload.container, "sealed-secrets-controller"
+        )
+        self.assertEqual(
+            sealed_secrets.availability_guard,
+            controller.AvailabilityGuard(
+                minimum_ready_replicas=1,
+                maximum_effective_unavailable=0,
+                minimum_effective_surge=1,
+                enforced_max_unavailable=0,
+                enforced_max_surge=1,
+            ),
+        )
+
     def test_policy_loads_and_validates_init_container_kind(self) -> None:
         image = self._image_with_container_kind("init_container")