feat(governance): 新增服務健康缺口矩陣

docs/LOGBOOK.md

@@ -1,3 +1,38 @@
+## 2026-06-05｜P1-005 服務健康缺口與過期端點本地完成
+
+**背景**：接續 P1-004 AI Provider 路由矩陣正式驗證，依工作清單推進 `P1-005`。本段只建立 committed service health gap matrix、只讀 API 與治理頁顯示；不做 live probe / external probe、不重啟 service / pod / host、不修改 endpoint / ConfigMap、不讀 Secret / Redis / DB payload、不發 Telegram / AwoooP 通知、不觸發 workflow / deploy / reload / runtime execution。
+
+**本輪完成**：
+- 新增 `service_health_gap_matrix_v1` schema 與 `docs/evaluations/service_health_gap_matrix_2026-06-05.json`。
+- 新增 `GET /api/v1/agents/service-health-gap-matrix` 與 service guard，強制驗證 read-only mode、operation / approval boundaries、rollup consistency、health gap / stale endpoint / target evidence 與 operator denial。
+- 治理頁 `/zh-TW/governance?tab=automation-inventory` 新增「服務健康缺口與過期端點」區塊，將健康目標、需處置、過期端點、健康缺口與不可誤讀合約拆成 KPI、摘要磚、目標卡與缺口清單，降低文字牆閱讀負擔。
+- 同步 automation backlog / inventory snapshot：current `P1-005`、next `P1-006`、backlog overall `88%`、P1 `95%`、done `21/24`、inventory tasks `31`。
+
+**目前數字**：
+- Service health targets：`10`。
+- 需處置 targets：`5`。
+- Stale endpoints：`3`。
+- Health gaps：`5`。
+- Service restart / endpoint change / active probe / notification send / runtime execution allowed counts 全部 `0`。
+
+**本地驗證**：
+- JSON parse 通過：`service_health_gap_matrix_2026-06-05.json`、`service_health_gap_matrix_v1.schema.json`、automation backlog / inventory snapshots、zh-TW / en messages。
+- 目標測試通過：service health gap matrix service / API、automation inventory / backlog snapshot service / API 共 `25 passed`。
+- Python py_compile 通過：`apps/api/src/services/service_health_gap_matrix.py`、`apps/api/src/api/v1/agents.py`。
+- zh-TW / en i18n key 差異 `0`；web typecheck 通過；Next production build 通過，治理頁 First Load JS `387 kB`；`source-control-owner-response-guard.py`、`security-mirror-progress-guard.py` 與 `git diff --check` 通過。
+- 本地 API readback：service health gap matrix 回 `service_health_gap_matrix_v1`、current `P1-005`、next `P1-006`、targets `10`、需處置 `5`、stale endpoints `3`、health gaps `5`；backlog 回 overall `88%`、done `21/24`；inventory 回 tasks `31`。
+
+**待補**：
+- 推送 Gitea 後等待 deploy marker，補 production API readback 與 desktop / mobile browser smoke，再更新正式驗證紀錄。
+
+**邊界**：
+- Live probe、external health probe、service / pod / host restart、rollout restart、endpoint / ConfigMap 修改、provider switch、paid API call、Secret payload read、通知發送、workflow / deploy / reload / runtime execution 全部仍未批准。
+- P1-005 UI / API 可見只代表 committed service health gap matrix 完成，不代表任何健康檢查已 live 執行，也不代表 runtime gate、S4.9 owner response gate 或 security acceptance gate 已提高。
+
+**下一步**：
+1. 完成本輪 build、Gitea push、正式 deploy marker 追蹤與 production smoke。
+2. P1-006：在 UI 顯示更細緻的 service health evidence cards。
+
 ## 2026-06-05｜P1-004 AI Provider 路由矩陣正式上線
 
 **背景**：接續 P1-003 監控合約與降噪矩陣正式驗證，依工作清單推進 `P1-004`。本段只建立 AI Router / Ollama / OpenClaw / Nemotron / Gemini provider route 的 committed 只讀矩陣、API 與治理頁顯示；不切換 provider、不呼叫 Gemini / NVIDIA / Claude、不改 `USE_AI_ROUTER`、不修改 fallback order 或 `OLLAMA_*`、不讀 Secret payload、不進 shadow / canary、不觸發 workflow / deploy / reload / runtime execution。

docs/ai/AI_AGENT_AUTOMATION_WORKLIST_2026-06-04.md

@@ -10,10 +10,10 @@
 |---|---:|---|---|
 | Agent 市場治理 | 72% | 進行中 | `agent_market_governance_snapshot_v1`、API、UI 分頁、每週觀察流程 |
 | Nemotron 實際整合應用 | 30% | 完整回放前仍被關卡擋下 | `blocked_needs_evidence`，下一關是 `refresh_source_evidence_then_5_record_smoke_only` |
-| 工具 / 服務 / 套件 AI 自動化 | 87% | P0 已完成，P1 套件 / 供應鏈主線已完成；備份 / DR 主線已完成到異地 / escrow 準備度顯示；任務批准邊界、進度彙總、P1-001 執行面只讀矩陣、P1-002 Gitea 工作流程 / runner 健康合約、P1-003 監控合約 / 降噪矩陣與 P1-004 AI 供應商路由矩陣已完成，下一主線是 P1-005 服務健康缺口與過期端點偵測 | 狀態分類、盤點 schema、權限矩陣、靜態盤點種子、只讀 API、UI 骨架、驗證、自動化待辦 schema / 快照 / API / 分組 UI、Backup / DR 目標盤點、準備度矩陣、備份通知政策、Backup / DR 證據 UI、復原演練批准包模板、異地 / escrow 準備度狀態、任務批准邊界、確定性進度彙總、Python 套件 / 供應鏈只讀基線、JS pnpm/npm 只讀基線、Docker build surface 只讀基線、CVE / license / drift 嚴重度政策、定期依賴漂移與外部資料來源檢查設計、依賴升級批准包模板、runtime_surface_inventory_v1 schema / snapshot / API / UI、gitea_workflow_runner_health_v1 schema / snapshot / API / UI、observability_contract_matrix_v1 schema / snapshot / API / UI、ai_provider_route_matrix_v1 schema / snapshot / API / UI 已完成 |
+| 工具 / 服務 / 套件 AI 自動化 | 88% | P0 已完成，P1 套件 / 供應鏈主線已完成；備份 / DR 主線已完成到異地 / escrow 準備度顯示；任務批准邊界、進度彙總、P1-001 執行面只讀矩陣、P1-002 Gitea 工作流程 / runner 健康合約、P1-003 監控合約 / 降噪矩陣、P1-004 AI 供應商路由矩陣與 P1-005 服務健康缺口矩陣已完成，下一主線是 P1-006 在 UI 顯示 service health 證據卡 | 狀態分類、盤點 schema、權限矩陣、靜態盤點種子、只讀 API、UI 骨架、驗證、自動化待辦 schema / 快照 / API / 分組 UI、Backup / DR 目標盤點、準備度矩陣、備份通知政策、Backup / DR 證據 UI、復原演練批准包模板、異地 / escrow 準備度狀態、任務批准邊界、確定性進度彙總、Python 套件 / 供應鏈只讀基線、JS pnpm/npm 只讀基線、Docker build surface 只讀基線、CVE / license / drift 嚴重度政策、定期依賴漂移與外部資料來源檢查設計、依賴升級批准包模板、runtime_surface_inventory_v1 schema / snapshot / API / UI、gitea_workflow_runner_health_v1 schema / snapshot / API / UI、observability_contract_matrix_v1 schema / snapshot / API / UI、ai_provider_route_matrix_v1 schema / snapshot / API / UI、service_health_gap_matrix_v1 schema / snapshot / API / UI 已完成 |
 | 本工作清單與分析報告 | 100% | 已完成 | 本 MD 文件 |
 
-AI Agent 自動化工作包目前完成度：**87%**。本工作清單文件本身完成度：**100%**。
+AI Agent 自動化工作包目前完成度：**88%**。本工作清單文件本身完成度：**100%**。
 
 完成度計算模型：
 
@@ -870,7 +870,7 @@ UI：
 | P1-002 | 完成 | 100 | Hermes | 盤點 Gitea 工作流程與 runner 健康合約 | `gitea_workflow_runner_health_v1` / `GET /api/v1/agents/gitea-workflow-runner-health` / Gitea 健康合約 UI | 只讀；不修改 workflow、不重啟 runner、不停止 container、不讀 Secret、不發通知 |
 | P1-003 | 完成 | 100 | Hermes | 盤點 Prometheus / Alertmanager / SigNoz / Grafana 監控合約 | `observability_contract_matrix_v1` / `GET /api/v1/agents/observability-contract-matrix` / 監控合約與降噪 UI | 只讀；不修改 alert rules、不改 receiver/route、不建立 silence、不寫 Grafana、不發通知 |
 | P1-004 | 完成 | 100 | OpenClaw | 盤點 AI Router / Ollama / Nemotron / Gemini provider 路徑 | `ai_provider_route_matrix_v1` / `GET /api/v1/agents/ai-provider-route-matrix` / AI 供應商路由矩陣 UI | 只讀；不切 provider、不呼叫付費 API、不改 fallback order、不進 shadow/canary |
-| P1-005 | 待辦 | 0 | OpenClaw | 偵測服務健康缺口與過期端點 | 需處置清單 | 不重啟 |
+| P1-005 | 完成 | 100 | OpenClaw | 偵測服務健康缺口與過期端點 | `service_health_gap_matrix_v1` / `GET /api/v1/agents/service-health-gap-matrix` / 服務健康缺口與過期端點 UI | 只讀；不重啟、不改 endpoint、不做 active probe、不發通知 |
 | P1-006 | 待辦 | 0 | Hermes | 在 UI 顯示 service health 證據卡 | 狀態卡 | 瀏覽器驗證 |
 | P1-007 | 待辦 | 0 | OpenClaw | 建立 service health 失敗限定 Telegram / AwoooP 對應 | 通知合約 | 不發成功洗版 |
 
@@ -1102,24 +1102,23 @@ UI：
 本次同步：
 
 ```text
-進度：87%。
+進度：88%。
 目前優先級：P1。
-目前任務：P1-004 盤點 AI Router / Ollama / Nemotron / Gemini provider 路徑。
+目前任務：P1-005 偵測服務健康缺口與過期端點。
 狀態變更：待辦 -> 完成。
-證據：ai_provider_route_matrix_v1 schema / snapshot；GET /api/v1/agents/ai-provider-route-matrix；治理頁 AI 供應商路由矩陣區塊；automation backlog 87%；inventory tasks 30。
-目前數字：provider routes 7；需處置 routes 2；需人工批准 candidate gates 3；source gaps 3；blocked candidates 1；provider switch / paid API call / shadow or canary / runtime route change allowed counts 全部 0；backlog done 20/23；overall 87%；P1 95%；WS3 監控自動化 100%。
-驗證：JSON parse 通過；AI provider route matrix service / API、automation inventory / backlog snapshot service / API 目標測試 `25 passed`；Python py_compile 通過；zh-TW / en i18n key 差異 `0`；web typecheck 通過；Next production build 通過；source-control-owner-response guard、security-mirror-progress guard、git diff --check 通過；本地 API readback 回 `ai_provider_route_matrix_v1`、current `P1-004`、next `P1-005`、routes `7`、approval gates `3`；backlog 回 `87%`、done `20/23`；inventory 回 tasks `30`；本地 desktop `1440x1000` 與 mobile `390x844` browser smoke 通過，`AI 供應商路由矩陣`、`P1-004`、`P1-005`、`87%`、`Ollama 三層`、`Nemotron 候選`、`不可誤讀合約` 可見，`horizontalOverflow=0`、overflowing elements `0`、危險互動入口 `0`。
-正式驗證：code commit `45556f8f` 已推 `gitea main`；deploy marker `c619446b chore(cd): deploy 45556f8 [skip ci]`；Gitea code-review `#2607` 成功；CD `#2606` tests 與 build-and-deploy 成功，post-deploy-checks job 顯示 `Blocked by required conditions`，已由人工只讀 production smoke 補正式驗證。Production health `healthy/prod/mock_mode=false`；AI provider route matrix API 回 `ai_provider_route_matrix_v1`、current `P1-004`、next `P1-005`、routes `7`、approval gates `3`、provider switch allowed `0`；backlog API 回 overall `87%`、done `20/23`；inventory API 回 tasks `30`、`production_change_blocked=1`；production desktop `1440x1000` 與 mobile `390x844` smoke 通過，`AI 供應商路由矩陣`、`P1-004`、`P1-005`、`87%`、`Ollama 三層`、`Nemotron 候選`、`不可誤讀合約`、`允許入口 0` 可見；console error `0`、HTTP failed response `0`、`horizontalOverflow=0`、overflowing elements `0`、危險互動入口 `0`。
-阻擋：provider switch、production routing change、USE_AI_ROUTER 切換、fallback order 修改、OLLAMA_* endpoint / ConfigMap 修改、Gemini / Claude / NVIDIA 付費呼叫或頻率提高、OpenClaw 取代 / 降級、Nemotron shadow / canary、external live probe / benchmark、Secret payload read、通知發送、workflow/deploy/reload/runtime execution 仍全部禁止。
-下一步：P1-005 偵測服務健康缺口與過期端點。
+證據：service_health_gap_matrix_v1 schema / snapshot；GET /api/v1/agents/service-health-gap-matrix；治理頁服務健康缺口與過期端點區塊；automation backlog 88%；inventory tasks 31。
+目前數字：service health targets 10；需處置 targets 5；stale endpoints 3；health gaps 5；service restart / endpoint change / active probe / notification send / runtime execution allowed counts 全部 0；backlog done 21/24；overall 88%；P1 95%；WS3 監控自動化 100%。
+驗證：JSON parse 通過；service health gap matrix service / API、automation inventory / backlog snapshot service / API 目標測試 `25 passed`；Python py_compile 通過；zh-TW / en i18n key 差異 `0`；web typecheck 通過；source-control-owner-response guard、security-mirror-progress guard、git diff --check 通過；本地 API readback 回 `service_health_gap_matrix_v1`、current `P1-005`、next `P1-006`、targets `10`、需處置 `5`、stale endpoints `3`、health gaps `5`；backlog 回 `88%`、done `21/24`；inventory 回 tasks `31`。
+正式驗證：待 code commit 推送 Gitea、取得 deploy marker 後補 production API readback 與 desktop / mobile browser smoke。
+阻擋：live probe、external health probe、service / pod / host restart、rollout restart、endpoint / ConfigMap 修改、provider switch、paid API call、Secret payload read、通知發送、workflow/deploy/reload/runtime execution 仍全部禁止。
+下一步：P1-006 在 UI 顯示 service health 證據卡。
 ```
 
 ## 13. 立即執行順序
 
-1. P1-005：偵測服務健康缺口與過期端點。
-2. P1-006：在 UI 顯示 service health 證據卡。
-3. P1-007：建立 service health 失敗限定 Telegram / AwoooP 對應。
-4. P2 / P3 必須等 P1 服務、監控與 provider runtime surface 可見且關卡穩定後再做。
+1. P1-006：在 UI 顯示 service health 證據卡。
+2. P1-007：建立 service health 失敗限定 Telegram / AwoooP 對應。
+3. P2 / P3 必須等 P1 服務、監控、provider route 與 service health evidence cards 可見且關卡穩定後再做。
 
 ## 14. 目前風險
 

docs/evaluations/ai_agent_automation_backlog_2026-06-04.json

@@ -1,34 +1,34 @@
 {
   "schema_version": "ai_agent_automation_backlog_v1",
-  "generated_at": "2026-06-05T13:28:00+08:00",
+  "generated_at": "2026-06-05T14:05:00+08:00",
   "source_inventory_snapshot_ref": "docs/evaluations/ai_agent_automation_inventory_snapshot_2026-06-04_static_seed.json",
   "program_status": {
-    "overall_completion_percent": 87,
+    "overall_completion_percent": 88,
     "current_priority": "P1",
-    "current_task_id": "P1-004",
-    "next_task_id": "P1-005",
+    "current_task_id": "P1-005",
+    "next_task_id": "P1-006",
     "read_only_mode": true
   },
   "rollups": {
-    "total_items": 23,
+    "total_items": 24,
     "by_priority": {
-      "P1": 21,
+      "P1": 22,
       "P2": 1,
       "P3": 1
     },
     "by_status": {
-      "done": 20,
+      "done": 21,
       "planned": 3
     },
     "by_gate_status": {
-      "read_only_allowed": 20,
+      "read_only_allowed": 21,
       "production_change_blocked": 1,
       "cost_approval_required": 1,
       "blocked_by_evidence": 1
     },
     "by_owner_agent": {
       "hermes": 12,
-      "openclaw": 10,
+      "openclaw": 11,
       "nemotron": 1
     }
   },
@@ -421,6 +421,63 @@
         ]
       }
     },
+    {
+      "item_id": "AUTO-P1-005",
+      "priority": "P1",
+      "status": "done",
+      "workstream_id": "WS3",
+      "source_asset_id": "service_health_gap_matrix",
+      "source_signal_kind": "runtime_evidence_gap",
+      "title": "偵測服務健康缺口與過期端點",
+      "owner_agent": "openclaw",
+      "recommended_action": "已建立 service health gap matrix，只整理 committed health / endpoint / stale evidence；不做 live probe、不重啟、不改 endpoint。",
+      "action_class": "observe",
+      "gate_status": "read_only_allowed",
+      "risk_level": "critical",
+      "evidence_refs": [
+        "docs/evaluations/service_health_gap_matrix_2026-06-05.json",
+        "GET /api/v1/agents/service-health-gap-matrix",
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "scripts/health_check_session.sh",
+        "docs/evaluations/ai_provider_route_matrix_2026-06-05.json",
+        "docs/evaluations/observability_contract_matrix_2026-06-05.json",
+        "docs/evaluations/gitea_workflow_runner_health_2026-06-05.json"
+      ],
+      "acceptance_criteria": [
+        "不做 live probe 或 external health probe",
+        "不重啟 service / pod / host，不 rollout restart",
+        "不修改 endpoint / ConfigMap / public URL",
+        "不讀 Secret payload、不發通知、不觸發 workflow / deploy / runtime execution"
+      ],
+      "next_review": "P1-006",
+      "approval_boundary": {
+        "mode": "read_only_allowed",
+        "display_summary": "只允許整理 committed service health gap evidence 與需處置清單；不得 live probe、重啟、改 endpoint 或發通知。",
+        "allowed_actions": [
+          "讀取 committed snapshot",
+          "整理健康缺口",
+          "顯示治理 UI",
+          "準備 operator review 清單"
+        ],
+        "blocked_actions": [
+          "production_write",
+          "runtime_execution",
+          "destructive_operation",
+          "secret_plaintext_collection",
+          "live_probe",
+          "service_restart",
+          "endpoint_change",
+          "notification_send"
+        ],
+        "requires_operator_approval_for": [
+          "live probe",
+          "服務重啟",
+          "endpoint / ConfigMap 修改",
+          "Telegram / AwoooP 通知發送",
+          "runtime execution"
+        ]
+      }
+    },
     {
       "item_id": "AUTO-P1-007",
       "priority": "P1",
@@ -1142,9 +1199,9 @@
     "destructive_operation_allowed": false
   },
   "item_approval_boundary_rollup": {
-    "total_items": 23,
+    "total_items": 24,
     "by_mode": {
-      "read_only_allowed": 20,
+      "read_only_allowed": 21,
       "production_change_blocked": 1,
       "cost_approval_required": 1,
       "blocked_by_evidence": 1
@@ -1155,14 +1212,11 @@
       "AUTO-P3-001"
     ],
     "items_with_blocked_operations": [
-      "AUTO-P1-303",
-      "AUTO-P1-304",
-      "AUTO-P1-305",
-      "AUTO-P1-306",
       "AUTO-P1-001",
       "AUTO-P1-002",
       "AUTO-P1-003",
       "AUTO-P1-004",
+      "AUTO-P1-005",
       "AUTO-P1-007",
       "AUTO-P1-101",
       "AUTO-P1-102",
@@ -1176,22 +1230,26 @@
       "AUTO-P1-204",
       "AUTO-P1-205",
       "AUTO-P1-206",
+      "AUTO-P1-303",
+      "AUTO-P1-304",
+      "AUTO-P1-305",
+      "AUTO-P1-306",
       "AUTO-P2-004",
       "AUTO-P3-001"
     ]
   },
   "progress_summary": {
-    "overall_percent": 87,
-    "done_items": 20,
+    "overall_percent": 88,
+    "done_items": 21,
     "planned_items": 3,
-    "total_items": 23,
+    "total_items": 24,
     "formula": "round(done_items / total_items * 100)，只有 status=done 計入完成；planned/in_progress/blocked/deferred/rejected 不計入。",
     "by_priority": [
       {
         "priority": "P1",
         "completion_percent": 95,
-        "done_items": 20,
-        "total_items": 21
+        "done_items": 21,
+        "total_items": 22
       },
       {
         "priority": "P2",
@@ -1215,14 +1273,30 @@
         "total_items": 3,
         "next_task_id": "P3-001"
       },
+      {
+        "workstream_id": "WS8",
+        "display_name": "產品 UI",
+        "completion_percent": 100,
+        "done_items": 2,
+        "total_items": 2,
+        "next_task_id": "complete"
+      },
       {
         "workstream_id": "WS3",
         "display_name": "監控自動化",
         "completion_percent": 100,
-        "done_items": 4,
-        "total_items": 4,
+        "done_items": 5,
+        "total_items": 5,
         "next_task_id": "complete"
       },
+      {
+        "workstream_id": "WS7",
+        "display_name": "安全執行關卡",
+        "completion_percent": 0,
+        "done_items": 0,
+        "total_items": 1,
+        "next_task_id": "P1-007"
+      },
       {
         "workstream_id": "WS4",
         "display_name": "備份與 DR 自動化",
@@ -1246,22 +1320,6 @@
         "done_items": 0,
         "total_items": 1,
         "next_task_id": "P2-004"
-      },
-      {
-        "workstream_id": "WS7",
-        "display_name": "安全執行關卡",
-        "completion_percent": 0,
-        "done_items": 0,
-        "total_items": 1,
-        "next_task_id": "P1-007"
-      },
-      {
-        "workstream_id": "WS8",
-        "display_name": "產品 UI",
-        "completion_percent": 100,
-        "done_items": 2,
-        "total_items": 2,
-        "next_task_id": "complete"
       }
     ]
   }

docs/evaluations/ai_agent_automation_inventory_snapshot_2026-06-04_static_seed.json

@@ -1,11 +1,11 @@
 {
   "schema_version": "ai_agent_automation_inventory_snapshot_v1",
-  "generated_at": "2026-06-05T13:28:00+08:00",
+  "generated_at": "2026-06-05T14:05:00+08:00",
   "program_status": {
     "overall_completion_percent": 100,
     "current_priority": "P1",
-    "current_task_id": "P1-004",
-    "next_task_id": "P1-005",
+    "current_task_id": "P1-005",
+    "next_task_id": "P1-006",
     "read_only_mode": true
   },
   "status_taxonomy": {
@@ -451,6 +451,21 @@
         "docs/evaluations/docker_build_surface_inventory_2026-06-04.json"
       ],
       "next_action": "P1-206 產生 Docker base image digest、binary source、CVE 與 rebuild approval package。"
+    },
+    {
+      "asset_id": "service_health_gap_matrix",
+      "domain_id": "services",
+      "display_name": "服務健康缺口與過期端點矩陣",
+      "asset_type": "read_only_health_gap_matrix",
+      "status": "ready_for_review",
+      "gate_status": "read_only_allowed",
+      "owner_agent": "openclaw",
+      "risk_level": "high",
+      "evidence_refs": [
+        "docs/evaluations/service_health_gap_matrix_2026-06-05.json",
+        "GET /api/v1/agents/service-health-gap-matrix"
+      ],
+      "next_action": "P1-006 在治理頁顯示 service health 證據卡；不得做 live probe、重啟或改 endpoint。"
     }
   ],
   "workstreams": [
@@ -1484,6 +1499,44 @@
           "付費 API 呼叫"
         ]
       }
+    },
+    {
+      "task_id": "P1-005",
+      "priority": "P1",
+      "status": "done",
+      "completion_percent": 100,
+      "owner_agent": "openclaw",
+      "title": "偵測服務健康缺口與過期端點",
+      "output": "docs/evaluations/service_health_gap_matrix_2026-06-05.json + GET /api/v1/agents/service-health-gap-matrix",
+      "gate_status": "read_only_allowed",
+      "next_action": "完成 committed service health gap matrix；下一步 P1-006 在 UI 顯示 service health 證據卡。",
+      "approval_boundary": {
+        "mode": "read_only_allowed",
+        "display_summary": "只允許整理 committed service health gap evidence 與需處置清單；不得 live probe、重啟、改 endpoint 或發通知。",
+        "allowed_actions": [
+          "讀取 committed snapshot",
+          "整理健康缺口",
+          "顯示治理 UI",
+          "準備 operator review 清單"
+        ],
+        "blocked_actions": [
+          "production_write",
+          "runtime_execution",
+          "destructive_operation",
+          "secret_plaintext_collection",
+          "live_probe",
+          "service_restart",
+          "endpoint_change",
+          "notification_send"
+        ],
+        "requires_operator_approval_for": [
+          "live probe",
+          "服務重啟",
+          "endpoint / ConfigMap 修改",
+          "Telegram / AwoooP 通知發送",
+          "runtime execution"
+        ]
+      }
     }
   ],
   "evidence": [
@@ -1818,6 +1871,18 @@
       "source_ref": "GET /api/v1/agents/ai-provider-route-matrix",
       "status": "done",
       "summary": "只讀呈現 AI Router / Ollama / OpenClaw / Nemotron / Gemini provider route、候選 Gate、來源缺口與不可誤讀合約；不切 provider、不呼叫付費 API、不讀 Secret。"
+    },
+    {
+      "evidence_id": "service_health_gap_matrix_snapshot",
+      "kind": "committed_snapshot",
+      "ref": "docs/evaluations/service_health_gap_matrix_2026-06-05.json",
+      "result": "P1-005 committed service health gap matrix 已建立，涵蓋 10 個 service health target、3 個 stale endpoint、5 個 health gap；重啟、endpoint 修改、active probe、通知與 runtime execution 入口皆為 0。"
+    },
+    {
+      "evidence_id": "service_health_gap_matrix_api",
+      "kind": "read_only_api",
+      "ref": "GET /api/v1/agents/service-health-gap-matrix",
+      "result": "只讀 API 回傳 service_health_gap_matrix_v1；不做 live probe、不重啟服務、不改 endpoint、不讀 Secret/Redis/DB payload、不發通知。"
     }
   ],
   "approval_boundaries": {
@@ -1828,10 +1893,10 @@
     "destructive_operation_allowed": false
   },
   "task_approval_boundary_rollup": {
-    "total_tasks": 30,
+    "total_tasks": 31,
     "by_mode": {
       "ready_for_operator_review": 1,
-      "read_only_allowed": 27,
+      "read_only_allowed": 28,
       "approval_required": 1,
       "production_change_blocked": 1
     },
@@ -1852,12 +1917,8 @@
       "P1-001",
       "P1-002",
       "P1-003",
-      "P1-301",
-      "P1-302",
-      "P1-303",
-      "P1-304",
-      "P1-305",
-      "P1-306",
+      "P1-004",
+      "P1-005",
       "P1-101",
       "P1-102",
       "P1-103",
@@ -1870,7 +1931,12 @@
       "P1-204",
       "P1-205",
       "P1-206",
-      "P1-004"
+      "P1-301",
+      "P1-302",
+      "P1-303",
+      "P1-304",
+      "P1-305",
+      "P1-306"
     ]
   }
 }

docs/evaluations/service_health_gap_matrix_2026-06-05.json

@@ -0,0 +1,439 @@
+{
+  "schema_version": "service_health_gap_matrix_v1",
+  "generated_at": "2026-06-05T14:05:00+08:00",
+  "program_status": {
+    "overall_completion_percent": 100,
+    "current_priority": "P1",
+    "current_task_id": "P1-005",
+    "next_task_id": "P1-006",
+    "read_only_mode": true
+  },
+  "source_refs": [
+    "docs/schemas/service_health_gap_matrix_v1.schema.json",
+    "docs/reference/SERVICE-ENDPOINTS.md",
+    "scripts/health_check_session.sh",
+    "docs/reference/ALERT-TAXONOMY-CATALOG.md",
+    "apps/api/src/core/config.py",
+    "apps/api/src/api/v1/health.py",
+    "apps/api/src/services/ai_provider_route_matrix.py",
+    "docs/evaluations/ai_provider_route_matrix_2026-06-05.json",
+    "docs/evaluations/observability_contract_matrix_2026-06-05.json",
+    "docs/evaluations/gitea_workflow_runner_health_2026-06-05.json",
+    "k8s/awoooi-prod/05-deployment-web.yaml",
+    "apps/api/src/services/metrics_service.py",
+    "apps/api/src/services/ai_providers/ollama.py",
+    "apps/api/src/services/ai_providers/openclaw_nemo.py",
+    "scripts/ops/backup-health-textfile-exporter.py",
+    "docs/ai/AI_AGENT_AUTOMATION_WORKLIST_2026-06-04.md"
+  ],
+  "rollups": {
+    "total_targets": 10,
+    "by_kind": {
+      "api_health": 1,
+      "web_health": 1,
+      "ai_provider_health": 2,
+      "observability_health": 3,
+      "devops_health": 1,
+      "backup_health": 1,
+      "security_edge_health": 1
+    },
+    "by_status": {
+      "verified": 5,
+      "action_required": 5
+    },
+    "by_freshness_status": {
+      "fresh_readback": 1,
+      "manifest_mapped": 1,
+      "source_mismatch": 3,
+      "committed_health_hook": 1,
+      "stale_evidence": 1,
+      "committed_exporter": 1,
+      "source_heartbeat_contract": 1,
+      "manifest_only": 1
+    },
+    "target_ids_requiring_action": [
+      "gitea_workflow_runner_health_contract",
+      "kali_scanner_health_reference",
+      "ollama_three_layer_health_contract",
+      "openclaw_health_endpoint_contract",
+      "prometheus_alertmanager_endpoint_reference"
+    ],
+    "health_gap_ids": [
+      "endpoint_reference_stale_hosts",
+      "gitea_runner_attestation_health_gap",
+      "health_check_script_not_authoritative",
+      "openclaw_nemo_rca_health_review",
+      "security_scanner_health_evidence_gap"
+    ],
+    "stale_endpoint_ids": [
+      "legacy_188_ollama_provider_endpoint",
+      "openclaw_8088_comment_vs_8089_contract",
+      "prometheus_alertmanager_110_188_split"
+    ],
+    "critical_target_ids": [
+      "ollama_three_layer_health_contract",
+      "openclaw_health_endpoint_contract",
+      "production_api_health_public",
+      "gitea_workflow_runner_health_contract"
+    ],
+    "read_only_denials_total": 10,
+    "service_restart_allowed_count": 0,
+    "endpoint_change_allowed_count": 0,
+    "active_probe_allowed_count": 0,
+    "notification_send_allowed_count": 0,
+    "runtime_execution_allowed_count": 0
+  },
+  "service_health_targets": [
+    {
+      "target_id": "production_api_health_public",
+      "display_name": "Production API Health",
+      "kind": "api_health",
+      "status": "verified",
+      "risk_level": "critical",
+      "freshness_status": "fresh_readback",
+      "health_contract": "`/api/v1/health` 是正式環境健康讀回入口，回傳 PostgreSQL / Redis / Ollama / OpenClaw / SigNoz 等 component 狀態與 `mock_mode=false`。",
+      "endpoint_contract": "正式站以 `https://awoooi.wooo.work/api/v1/health` 讀回；P1-005 只記錄現有讀回形狀，不新增 probe。",
+      "evidence_refs": [
+        "apps/api/src/api/v1/health.py",
+        "docs/LOGBOOK.md",
+        "docs/runbooks/AWOOOP-RLS-CANARY-WAVE1.md"
+      ],
+      "next_action": "P1-006 只把健康證據卡顯示到 UI；不得把讀回成功解讀成可重啟或改 endpoint。"
+    },
+    {
+      "target_id": "awoooi_web_health_manifest",
+      "display_name": "AWOOOI Web /api/health",
+      "kind": "web_health",
+      "status": "verified",
+      "risk_level": "high",
+      "freshness_status": "manifest_mapped",
+      "health_contract": "Web deployment liveness / readiness / startup probe 皆指向 `/api/health`。",
+      "endpoint_contract": "只讀 K8s manifest；不 patch deployment、不 rollout restart、不改 public URL。",
+      "evidence_refs": [
+        "k8s/awoooi-prod/05-deployment-web.yaml"
+      ],
+      "next_action": "P1-006 顯示 manifest health contract；任何 probe path 變更需另開部署批准。"
+    },
+    {
+      "target_id": "ollama_three_layer_health_contract",
+      "display_name": "Ollama 三層健康合約",
+      "kind": "ai_provider_health",
+      "status": "action_required",
+      "risk_level": "critical",
+      "freshness_status": "source_mismatch",
+      "health_contract": "目前 AI provider route truth 是 GCP-A → GCP-B → 111；Gemini 只作 final fallback。",
+      "endpoint_contract": "`SERVICE-ENDPOINTS.md` 仍有 188 Ollama 參考，`scripts/health_check_session.sh` 又檢查 111 / 110；需把 endpoint truth 收斂為 owner-attested source，不可直接改路由。",
+      "evidence_refs": [
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "scripts/health_check_session.sh",
+        "docs/evaluations/ai_provider_route_matrix_2026-06-05.json",
+        "apps/api/src/services/ai_providers/ollama.py"
+      ],
+      "next_action": "只開 health evidence card 與 endpoint truth review；不得改 `OLLAMA_*`、不重啟 Ollama、不做 live benchmark。"
+    },
+    {
+      "target_id": "openclaw_health_endpoint_contract",
+      "display_name": "OpenClaw 健康端點合約",
+      "kind": "ai_provider_health",
+      "status": "action_required",
+      "risk_level": "critical",
+      "freshness_status": "source_mismatch",
+      "health_contract": "OpenClaw 仍是生產仲裁核心；health check 使用 `OPENCLAW_URL/health`。",
+      "endpoint_contract": "`SERVICE-ENDPOINTS.md` 已標 8089，但 `openclaw_nemo.py` 註解仍殘留 8088；需收斂註解與 runbook truth。",
+      "evidence_refs": [
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "apps/api/src/services/ai_providers/openclaw_nemo.py",
+        "docs/evaluations/ai_provider_route_matrix_2026-06-05.json"
+      ],
+      "next_action": "只提出 source cleanup；不得重啟 OpenClaw、不得把 health gap 當成替換 OpenClaw 批准。"
+    },
+    {
+      "target_id": "prometheus_alertmanager_endpoint_reference",
+      "display_name": "Prometheus / Alertmanager 端點參考",
+      "kind": "observability_health",
+      "status": "action_required",
+      "risk_level": "high",
+      "freshness_status": "source_mismatch",
+      "health_contract": "Prometheus / Alertmanager 只允許 read-only readiness evidence；Alertmanager 仍必須走 AWOOOI API，不得指向 OpenClaw receiver。",
+      "endpoint_contract": "`SERVICE-ENDPOINTS.md` 記錄 188:9090 / 9093，`health_check_session.sh` 檢查 110:9090 / 9093；需要 owner attestation。",
+      "evidence_refs": [
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "scripts/health_check_session.sh",
+        "docs/evaluations/observability_contract_matrix_2026-06-05.json"
+      ],
+      "next_action": "P1-006 顯示 source mismatch；不得改 alert rule、receiver、route 或 silence。"
+    },
+    {
+      "target_id": "signoz_clickhouse_health_contract",
+      "display_name": "SigNoz / ClickHouse 健康合約",
+      "kind": "observability_health",
+      "status": "verified",
+      "risk_level": "high",
+      "freshness_status": "committed_health_hook",
+      "health_contract": "metrics service 有 ClickHouse health check；observability matrix 已把 SigNoz / ClickHouse 視為 committed evidence。",
+      "endpoint_contract": "只讀 committed health hook 與 matrix；不改 SigNoz query、不改 ClickHouse datasource。",
+      "evidence_refs": [
+        "apps/api/src/services/metrics_service.py",
+        "docs/evaluations/observability_contract_matrix_2026-06-05.json"
+      ],
+      "next_action": "後續若要 live query，需另開批准；P1-005 只呈現 committed evidence。"
+    },
+    {
+      "target_id": "sentry_source_health_contract",
+      "display_name": "Sentry Source Health",
+      "kind": "observability_health",
+      "status": "verified",
+      "risk_level": "medium",
+      "freshness_status": "source_heartbeat_contract",
+      "health_contract": "Source provider heartbeat 與 upstream canary 已被 observability matrix 納入來源證據；P1-005 不呼叫 Sentry API。",
+      "endpoint_contract": "只讀 source-link / heartbeat contract；不改 webhook、不讀 Sentry token。",
+      "evidence_refs": [
+        "docs/evaluations/observability_contract_matrix_2026-06-05.json",
+        "docs/reference/SERVICE-ENDPOINTS.md"
+      ],
+      "next_action": "P1-006 顯示 source heartbeat contract；token / webhook 變更需另行批准。"
+    },
+    {
+      "target_id": "gitea_workflow_runner_health_contract",
+      "display_name": "Gitea Workflow / Runner Health",
+      "kind": "devops_health",
+      "status": "action_required",
+      "risk_level": "critical",
+      "freshness_status": "stale_evidence",
+      "health_contract": "Gitea workflow runner health matrix 已完成，但 runner owner attestation gap 仍需處置。",
+      "endpoint_contract": "只讀 Gitea workflow / runner snapshot；不重啟 runner、不停止 container、不觸發 workflow。",
+      "evidence_refs": [
+        "docs/evaluations/gitea_workflow_runner_health_2026-06-05.json",
+        "docs/LOGBOOK.md"
+      ],
+      "next_action": "保留 owner attestation 待補；不得用 P1-005 觸發 Gitea Actions 或 runner mutation。"
+    },
+    {
+      "target_id": "backup_health_exporter_freshness",
+      "display_name": "Backup Health Exporter Freshness",
+      "kind": "backup_health",
+      "status": "verified",
+      "risk_level": "high",
+      "freshness_status": "committed_exporter",
+      "health_contract": "backup health textfile exporter 已定義 backup freshness、restore drill freshness、offsite freshness 與 escrow freshness metrics。",
+      "endpoint_contract": "只讀 exporter contract；不執行 backup / restore / offsite sync，不寫 credential marker。",
+      "evidence_refs": [
+        "scripts/ops/backup-health-textfile-exporter.py",
+        "docs/evaluations/backup_dr_readiness_matrix_2026-06-04.json",
+        "docs/evaluations/offsite_escrow_readiness_status_2026-06-05.json"
+      ],
+      "next_action": "維持成功不洗版；失敗或 stale 才進 P1-007 通知合約。"
+    },
+    {
+      "target_id": "kali_scanner_health_reference",
+      "display_name": "Kali Scanner Health Reference",
+      "kind": "security_edge_health",
+      "status": "action_required",
+      "risk_level": "medium",
+      "freshness_status": "manifest_only",
+      "health_contract": "`SERVICE-ENDPOINTS.md` 記錄 Kali Scanner `192.168.0.112:8080`，但 P1-005 不做 live probe。",
+      "endpoint_contract": "只讀服務端點參考；不啟動掃描、不呼叫 scanner、不改 host 服務。",
+      "evidence_refs": [
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "docs/LOGBOOK.md"
+      ],
+      "next_action": "P1-006 僅顯示需 owner attestation；真正掃描健康檢查需另行批准。"
+    }
+  ],
+  "health_gaps": [
+    {
+      "gap_id": "endpoint_reference_stale_hosts",
+      "display_name": "端點參考存在主機 truth drift",
+      "status": "action_required",
+      "severity": "high",
+      "summary": "服務端點文件、health script、AI provider route matrix 對 Ollama / Prometheus / Alertmanager / OpenClaw 的主機來源不完全一致。",
+      "target_ids": [
+        "ollama_three_layer_health_contract",
+        "openclaw_health_endpoint_contract",
+        "prometheus_alertmanager_endpoint_reference"
+      ],
+      "evidence_refs": [
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "scripts/health_check_session.sh",
+        "docs/evaluations/ai_provider_route_matrix_2026-06-05.json"
+      ],
+      "next_action": "建立 owner attestation / source cleanup，不直接改 endpoint。"
+    },
+    {
+      "gap_id": "gitea_runner_attestation_health_gap",
+      "display_name": "Runner 健康仍缺 owner attestation",
+      "status": "action_required",
+      "severity": "high",
+      "summary": "P1-002 已列 runner attestation gap；P1-005 只把它納入 service health 缺口，不觸發 runner 操作。",
+      "target_ids": [
+        "gitea_workflow_runner_health_contract"
+      ],
+      "evidence_refs": [
+        "docs/evaluations/gitea_workflow_runner_health_2026-06-05.json"
+      ],
+      "next_action": "由 P1-006 顯示，P1-007 才定義失敗限定通知。"
+    },
+    {
+      "gap_id": "health_check_script_not_authoritative",
+      "display_name": "健康檢查腳本不是唯一真相來源",
+      "status": "proposal_required",
+      "severity": "medium",
+      "summary": "`scripts/health_check_session.sh` 可供人工會前檢查，但它與正式 K8s / public health / provider matrix 的 truth chain 需對齊。",
+      "target_ids": [
+        "production_api_health_public",
+        "prometheus_alertmanager_endpoint_reference",
+        "kali_scanner_health_reference"
+      ],
+      "evidence_refs": [
+        "scripts/health_check_session.sh",
+        "docs/reference/SERVICE-ENDPOINTS.md"
+      ],
+      "next_action": "只產生 P1-006 顯示與後續 runbook cleanup 提案。"
+    },
+    {
+      "gap_id": "openclaw_nemo_rca_health_review",
+      "display_name": "OpenClaw / Nemo RCA 健康仍需複核",
+      "status": "action_required",
+      "severity": "high",
+      "summary": "AI provider route matrix 將 OpenClaw Nemo RCA lane 標為需處置；P1-005 不改 provider，只列健康缺口。",
+      "target_ids": [
+        "openclaw_health_endpoint_contract"
+      ],
+      "evidence_refs": [
+        "docs/evaluations/ai_provider_route_matrix_2026-06-05.json",
+        "apps/api/src/services/ai_providers/openclaw_nemo.py"
+      ],
+      "next_action": "用 replay / smoke 證據補強，不把 health review 當作 OpenClaw 替換批准。"
+    },
+    {
+      "gap_id": "security_scanner_health_evidence_gap",
+      "display_name": "Kali scanner 只有 endpoint reference",
+      "status": "proposal_required",
+      "severity": "medium",
+      "summary": "Security edge 目前只有服務端點參考，缺少 committed health contract 或 owner attestation。",
+      "target_ids": [
+        "kali_scanner_health_reference"
+      ],
+      "evidence_refs": [
+        "docs/reference/SERVICE-ENDPOINTS.md"
+      ],
+      "next_action": "後續建立 read-only scanner health evidence card；不得啟動掃描。"
+    }
+  ],
+  "stale_endpoints": [
+    {
+      "endpoint_id": "legacy_188_ollama_provider_endpoint",
+      "display_name": "legacy 188 Ollama endpoint reference",
+      "status": "action_required",
+      "severity": "high",
+      "stale_ref": "`docs/reference/SERVICE-ENDPOINTS.md` 仍列 `192.168.0.188:11434` 為 Ollama。",
+      "current_truth": "P1-004 provider route truth 是 GCP-A → GCP-B → 111；188 不再作為 AWOOOI Ollama provider。",
+      "evidence_refs": [
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "docs/evaluations/ai_provider_route_matrix_2026-06-05.json",
+        "apps/api/src/services/ai_providers/ollama.py"
+      ],
+      "next_action": "準備 source cleanup 批准包；不得改 `OLLAMA_URL` 或重啟 Ollama。"
+    },
+    {
+      "endpoint_id": "prometheus_alertmanager_110_188_split",
+      "display_name": "Prometheus / Alertmanager 110 與 188 參考分裂",
+      "status": "action_required",
+      "severity": "medium",
+      "stale_ref": "`SERVICE-ENDPOINTS.md` 使用 188，`health_check_session.sh` 使用 110。",
+      "current_truth": "P1-005 只判定為 source mismatch，需要 owner attestation；不判定 live service down。",
+      "evidence_refs": [
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "scripts/health_check_session.sh",
+        "docs/evaluations/observability_contract_matrix_2026-06-05.json"
+      ],
+      "next_action": "只補 truth owner；不得改 Alertmanager route 或 receiver。"
+    },
+    {
+      "endpoint_id": "openclaw_8088_comment_vs_8089_contract",
+      "display_name": "OpenClaw 8088 註解與 8089 合約不一致",
+      "status": "action_required",
+      "severity": "medium",
+      "stale_ref": "`openclaw_nemo.py` 註解仍提到 `192.168.0.188:8088`。",
+      "current_truth": "`SERVICE-ENDPOINTS.md` 與設定使用 OpenClaw 8089；P1-005 只標記註解 / runbook drift。",
+      "evidence_refs": [
+        "apps/api/src/services/ai_providers/openclaw_nemo.py",
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "apps/api/src/core/config.py"
+      ],
+      "next_action": "後續 source cleanup 不得改 provider route 或降級 OpenClaw。"
+    }
+  ],
+  "latest_observations": [
+    {
+      "observation_id": "service_health_gap_matrix_seed",
+      "status": "verified",
+      "summary": "P1-005 只讀收斂 committed health / endpoint / stale evidence；不做 live probe。",
+      "evidence_refs": [
+        "docs/reference/SERVICE-ENDPOINTS.md",
+        "docs/evaluations/ai_provider_route_matrix_2026-06-05.json",
+        "docs/evaluations/observability_contract_matrix_2026-06-05.json"
+      ]
+    },
+    {
+      "observation_id": "active_probe_denied",
+      "status": "preserved",
+      "summary": "本矩陣沒有呼叫任何內網、外部 provider、Sentry、SigNoz、Gitea 或 scanner health endpoint。",
+      "evidence_refs": [
+        "docs/HARD_RULES.md",
+        "docs/ai/AI_AGENT_AUTOMATION_WORKLIST_2026-06-04.md"
+      ]
+    }
+  ],
+  "operator_contract": {
+    "display_mode": "read_only_service_health_gap_matrix",
+    "must_not_interpret_as": [
+      "服務重啟批准",
+      "endpoint 修改批准",
+      "active probe 批准",
+      "live health check 已執行",
+      "Secret payload 已讀取或可輸出",
+      "Telegram 成功通知批准",
+      "workflow / deploy / reload 觸發批准",
+      "provider 切換批准",
+      "runtime execution 授權",
+      "OpenClaw 取代或降級批准"
+    ],
+    "secret_display_policy": "只顯示來源檔案與 env var 名稱，不輸出 secret payload。",
+    "restart_policy": "P1-005 只能列缺口；任何 service / pod / host restart 都需另行批准。",
+    "endpoint_policy": "P1-005 不修改端點、不改 ConfigMap、不改 public URL，只提出 source truth cleanup。",
+    "notification_policy": "成功 smoke 不通知；只有失敗、stale、action-required 才進 P1-007 通知合約。"
+  },
+  "operation_boundaries": {
+    "read_only_api_allowed": true,
+    "service_restart_allowed": false,
+    "pod_restart_allowed": false,
+    "host_restart_allowed": false,
+    "rollout_restart_allowed": false,
+    "endpoint_change_allowed": false,
+    "configmap_patch_allowed": false,
+    "active_probe_allowed": false,
+    "external_health_probe_allowed": false,
+    "live_benchmark_allowed": false,
+    "provider_switch_allowed": false,
+    "paid_api_call_allowed": false,
+    "secret_read_allowed": false,
+    "secret_plaintext_allowed": false,
+    "notification_send_allowed": false,
+    "workflow_trigger_allowed": false,
+    "deploy_trigger_allowed": false,
+    "reload_trigger_allowed": false,
+    "runtime_execution_allowed": false
+  },
+  "approval_boundaries": {
+    "service_restart_approved": false,
+    "endpoint_change_approved": false,
+    "configmap_patch_approved": false,
+    "active_probe_approved": false,
+    "secret_read_approved": false,
+    "notification_send_approved": false,
+    "workflow_trigger_approved": false,
+    "provider_switch_approved": false,
+    "cost_change_approved": false
+  }
+}

docs/schemas/service_health_gap_matrix_v1.schema.json

@@ -0,0 +1,161 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "urn:awoooi:service-health-gap-matrix-v1",
+  "title": "AWOOOI 服務健康缺口只讀矩陣 v1",
+  "description": "以 repo 內 committed service endpoint、health check、K8s manifest、AI provider 與 observability evidence 建立服務健康缺口矩陣；不重啟服務、不改 endpoint、不做 active probe、不讀 secret payload、不送通知。",
+  "type": "object",
+  "required": [
+    "schema_version",
+    "generated_at",
+    "program_status",
+    "source_refs",
+    "rollups",
+    "service_health_targets",
+    "health_gaps",
+    "stale_endpoints",
+    "latest_observations",
+    "operator_contract",
+    "operation_boundaries",
+    "approval_boundaries"
+  ],
+  "properties": {
+    "schema_version": {
+      "type": "string",
+      "const": "service_health_gap_matrix_v1"
+    },
+    "generated_at": {
+      "type": "string",
+      "minLength": 1
+    },
+    "program_status": {
+      "type": "object",
+      "required": [
+        "overall_completion_percent",
+        "current_priority",
+        "current_task_id",
+        "next_task_id",
+        "read_only_mode"
+      ],
+      "properties": {
+        "overall_completion_percent": {
+          "type": "integer",
+          "minimum": 0,
+          "maximum": 100
+        },
+        "current_priority": {
+          "type": "string",
+          "enum": [
+            "P0",
+            "P1",
+            "P2",
+            "P3"
+          ]
+        },
+        "current_task_id": {
+          "type": "string",
+          "minLength": 1
+        },
+        "next_task_id": {
+          "type": "string",
+          "minLength": 1
+        },
+        "read_only_mode": {
+          "type": "boolean",
+          "const": true
+        }
+      },
+      "additionalProperties": false
+    },
+    "source_refs": {
+      "type": "array",
+      "minItems": 1,
+      "items": {
+        "type": "string",
+        "minLength": 1
+      }
+    },
+    "rollups": {
+      "type": "object",
+      "additionalProperties": true
+    },
+    "service_health_targets": {
+      "type": "array",
+      "minItems": 1,
+      "items": {
+        "type": "object",
+        "required": [
+          "target_id",
+          "display_name",
+          "kind",
+          "status",
+          "risk_level",
+          "freshness_status",
+          "health_contract",
+          "endpoint_contract",
+          "evidence_refs",
+          "next_action"
+        ],
+        "additionalProperties": true
+      }
+    },
+    "health_gaps": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "required": [
+          "gap_id",
+          "display_name",
+          "status",
+          "severity",
+          "summary",
+          "target_ids",
+          "evidence_refs",
+          "next_action"
+        ],
+        "additionalProperties": true
+      }
+    },
+    "stale_endpoints": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "required": [
+          "endpoint_id",
+          "display_name",
+          "status",
+          "severity",
+          "stale_ref",
+          "current_truth",
+          "evidence_refs",
+          "next_action"
+        ],
+        "additionalProperties": true
+      }
+    },
+    "latest_observations": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "additionalProperties": true
+      }
+    },
+    "operator_contract": {
+      "type": "object",
+      "additionalProperties": true
+    },
+    "operation_boundaries": {
+      "type": "object",
+      "additionalProperties": {
+        "type": "boolean"
+      }
+    },
+    "approval_boundaries": {
+      "type": "object",
+      "additionalProperties": {
+        "type": "boolean",
+        "const": false
+      }
+    }
+  },
+  "additionalProperties": false
+}

docs/superpowers/specs/2026-04-15-MASTER-ai-autonomous-flywheel-v2.md

@@ -3535,6 +3535,24 @@ Phase 6 完成後
 
 **裁決：** P1-003 只完成 read-only observability contract matrix 與降噪候選顯示。不得把 Prometheus rule count、Alertmanager grouping、SigNoz / Sentry heartbeat、Grafana dashboard JSON、OTEL/Event Exporter evidence 或 classification gap 解讀成 alert rule 變更、receiver/route 變更、OpenClaw receiver 授權、silence、dashboard import、webhook 修改、secret 讀取、通知發送、deploy/reload/workflow 或 runtime execution 批准。成功 smoke 不即時通知洗版；失敗與需處置才進批准流程。
 
+### 2026-06-05 下午 (台北) — P1-005 服務健康缺口與過期端點本地完成
+
+**觸發**：統帥批准繼續，接續 P1-004 正式驗證後推進 `docs/ai/AI_AGENT_AUTOMATION_WORKLIST_2026-06-04.md` 的 P1-005，只建立 committed service health gap evidence、只讀 API 與治理頁顯示。
+
+**已推進：**
+- P1-005：新增 `service_health_gap_matrix_v1` schema 與 `docs/evaluations/service_health_gap_matrix_2026-06-05.json`，涵蓋 API / Web / AI provider / Observability / Gitea / Backup / Security edge 的服務健康目標、健康缺口與過期端點 reference。
+- P1-005：新增 `GET /api/v1/agents/service-health-gap-matrix` 與 service guard，強制驗證 read-only mode、operation / approval boundaries、rollup consistency、target / health gap / stale endpoint evidence、operator denial 與禁止 secret payload key。
+- P1-005：治理頁 `/zh-TW/governance?tab=automation-inventory` 新增「服務健康缺口與過期端點」區塊；顯示健康目標、需處置、過期端點、健康缺口、不可誤讀合約與允許入口 0，不新增任何執行按鈕。
+- 目前數字：service health targets `10`；需處置 targets `5`；stale endpoints `3`；health gaps `5`；service restart / endpoint change / active probe / notification send / runtime execution allowed counts 全部 `0`；automation backlog done `21/24`、overall `88%`、P1 `95%`；inventory tasks `31`。
+- 本地驗證：JSON parse 通過；service health gap matrix service / API、automation inventory / backlog snapshot service / API 目標測試 `25 passed`；Python py_compile 通過；zh-TW / en i18n key 差異 `0`；web typecheck 通過。
+
+**待補驗證：**
+1. Next production build。
+2. `source-control-owner-response-guard.py`、`security-mirror-progress-guard.py` 與 `git diff --check`。
+3. 推送 Gitea 後取得 deploy marker，補 production API readback 與 desktop / mobile browser smoke。
+
+**裁決：** P1-005 只完成 service health gap matrix 的 committed evidence、API 與治理頁顯示。不得把 `service_health_gap_matrix_v1`、健康目標、過期端點、health gaps、source mismatch、owner attestation 或 UI 可見解讀成 live probe、服務重啟、endpoint / ConfigMap 修改、通知發送、provider switch、paid API call、Secret payload read、workflow / deploy / reload 或 runtime execution 批准。
+
 ### 2026-06-05 下午 (台北) — P1-004 AI Provider 路由矩陣正式上線
 
 **觸發**：統帥批准繼續，要求依 `docs/ai/AI_AGENT_AUTOMATION_WORKLIST_2026-06-04.md` 的優先順序推進，並在每階段同步完成度、工作狀態、驗證節點與正式環境頁面檢查。
@@ -3552,3 +3570,20 @@ Phase 6 完成後
 2. 持續保持 S4.9 owner response gate `0%` 與 active runtime gate `0`，不得把 P1-004 UI / API 可見誤讀成 runtime 授權。
 
 **裁決：** P1-004 只完成 read-only AI provider route matrix 與治理頁顯示。不得把 `ai_provider_route_matrix_v1`、Ollama 三層、Gemini final fallback、OpenClaw / Nemo RCA lane、Nemotron blocked candidate、legacy models registry 或 provider route source gaps 解讀成 provider 切換、付費呼叫、fallback 調整、OpenClaw 取代、Nemotron 升級、shadow / canary、Secret payload 讀取、workflow / deploy / reload 或 runtime execution 批准。
+
+### 2026-06-05 下午 (台北) — P1-005 服務健康缺口與過期端點本地完成
+
+**觸發**：統帥批准繼續，並要求治理頁避免文字牆、以圖表 / 卡片 / 專業資訊架構呈現完整資訊，同時強化決策支援覆蓋率的參考性。
+
+**已推進：**
+- P1-005：新增 `service_health_gap_matrix_v1` schema 與 `docs/evaluations/service_health_gap_matrix_2026-06-05.json`，以 committed health / endpoint / stale evidence 建立只讀 service health gap matrix。
+- P1-005：新增 `GET /api/v1/agents/service-health-gap-matrix` 只讀 API 與 service guard，強制拒絕把 snapshot 誤讀成 live probe、service restart、endpoint / ConfigMap 修改、Secret / Redis / DB payload 讀取、通知發送、workflow / deploy / reload 或 runtime execution 授權。
+- P1-005：治理頁 `/zh-TW/governance?tab=automation-inventory` 新增「服務健康缺口與過期端點」區塊；以 KPI、摘要磚、健康目標卡、過期端點清單、健康缺口清單與不可誤讀合約呈現，降低文字牆閱讀負擔，不新增任何執行按鈕。
+- 目前數字：service health targets `10`；需處置 targets `5`；stale endpoints `3`；health gaps `5`；service restart / endpoint change / active probe / notification send / runtime execution allowed counts 全部 `0`；automation backlog done `21/24`、overall `88%`、P1 `95%`、WS3 `100%`；inventory tasks `31`。
+- 本地驗證：JSON parse 通過；service health gap matrix service / API、automation inventory / backlog snapshot API、AI provider route matrix service / API 目標測試 `22 passed`；Python py_compile 通過；zh-TW / en i18n key 差異 `0`；web typecheck 通過；Next production build 通過，治理頁 First Load JS `387 kB`；source-control-owner-response guard、security-mirror-progress guard、`git diff --check` 通過；本地 API readback 回 `service_health_gap_matrix_v1`、current `P1-005`、next `P1-006`、targets `10`、需處置 `5`、stale endpoints `3`、health gaps `5`；backlog 回 `88%`、done `21/24`；inventory 回 tasks `31`。
+
+**待補：**
+1. Gitea push 後取得 deploy marker，補 production API readback 與 desktop / mobile browser smoke。
+2. P1-006：在 UI 顯示更細緻的 service health evidence cards。
+
+**裁決：** P1-005 只完成 read-only service health gap matrix、API 與治理頁顯示。不得把 `service_health_gap_matrix_v1`、健康目標、過期端點、健康缺口、需處置清單或不可誤讀合約解讀成 live health check 已執行、服務重啟批准、endpoint 修改批准、通知發送批准、runtime gate 提高、S4.9 owner response gate 提高或 security acceptance 已成立。