feat(web): add homepage blueprint drilldown

apps/web/messages/en.json

@@ -474,6 +474,57 @@
             "km": "KM / PlayBook"
           }
         },
+        "inspector": {
+          "title": "Stage Inspector",
+          "openTarget": "Open Work Surface",
+          "fields": {
+            "owner": "Owner",
+            "evidence": "Evidence Source",
+            "nextAction": "Next Step"
+          },
+          "stages": {
+            "signal": {
+              "owner": "OpenClaw + AlertChain",
+              "evidence": "Alertmanager / Sentry / SigNoz / Telegram callback trace",
+              "nextAction": "Attach the signal to an AwoooP run dossier and produce a traceable fingerprint"
+            },
+            "intake": {
+              "owner": "AwoooP Run Monitor",
+              "evidence": "Runs list / timeline / alert_operation_log / callback evidence",
+              "nextAction": "Link incident_id, trace_ref, and run_id so the alert does not stop at Telegram"
+            },
+            "ai": {
+              "owner": "OpenClaw leads decisions; Hermes drafts KM",
+              "evidence": "AI route status / selected provider / skipped lanes",
+              "nextAction": "Keep GCP-A -> GCP-B -> 111 -> Gemini fallback order and record the lane"
+            },
+            "mcp": {
+              "owner": "MCP Gateway",
+              "evidence": "K8s / Prometheus / Sentry / SigNoz / Gitea / self-hosted MCP results",
+              "nextAction": "Write MCP evidence back to the dossier so the LLM does not decide by guessing"
+            },
+            "playbook": {
+              "owner": "OpenClaw + PlayBook trust gate",
+              "evidence": "Quality gate / work items / playbook match / execution history",
+              "nextAction": "Fill execution, repair, approval, and learning evidence before promotion"
+            },
+            "ansible": {
+              "owner": "AwoooP Executor + Ansible lane",
+              "evidence": "ansible_runtime / check-mode count / pending check-mode / blockers",
+              "nextAction": "Clear ansible_playbook_binary_missing first, then run check-mode without direct apply"
+            },
+            "approval": {
+              "owner": "Approval Coordinator + SRE owner",
+              "evidence": "Approvals / risk gate / run timeline / manual_required reason",
+              "nextAction": "Allow low-risk automation, keep high-risk work under human approval and audit trail"
+            },
+            "verify": {
+              "owner": "Hermes + KM owner",
+              "evidence": "KM stale candidates / post-execution verification / playbook learning",
+              "nextAction": "Hermes drafts updates, owners review before KM write, then stale ratio is monitored"
+            }
+          }
+        },
         "values": {
           "verified": "verified {verified}/{evaluated}",
           "topGate": "{gate} missing {count}",

apps/web/messages/zh-TW.json

@@ -475,6 +475,57 @@
             "km": "KM / PlayBook"
           }
         },
+        "inspector": {
+          "title": "Stage Inspector",
+          "openTarget": "打開工作面",
+          "fields": {
+            "owner": "主責",
+            "evidence": "證據來源",
+            "nextAction": "下一步"
+          },
+          "stages": {
+            "signal": {
+              "owner": "OpenClaw + AlertChain",
+              "evidence": "Alertmanager / Sentry / SigNoz / Telegram callback trace",
+              "nextAction": "把 signal 併入 AwoooP run dossier，產生可追蹤 fingerprint"
+            },
+            "intake": {
+              "owner": "AwoooP Run Monitor",
+              "evidence": "runs list / timeline / alert_operation_log / callback evidence",
+              "nextAction": "關聯 incident_id、trace_ref、run_id，避免告警只停在 Telegram"
+            },
+            "ai": {
+              "owner": "OpenClaw 主判斷；Hermes 產 KM 草稿",
+              "evidence": "AI route status / selected provider / skipped lanes",
+              "nextAction": "維持 GCP-A → GCP-B → 111 → Gemini fallback 順序並記錄 lane"
+            },
+            "mcp": {
+              "owner": "MCP Gateway",
+              "evidence": "K8s / Prometheus / Sentry / SigNoz / Gitea / 自建 MCP 查證結果",
+              "nextAction": "把 MCP 查證結果寫回 dossier，讓 LLM 不靠猜測判斷"
+            },
+            "playbook": {
+              "owner": "OpenClaw + PlayBook trust gate",
+              "evidence": "quality gate / work items / playbook match / execution history",
+              "nextAction": "補齊 execution、repair、approval、learning evidence 後才允許升級"
+            },
+            "ansible": {
+              "owner": "AwoooP Executor + Ansible lane",
+              "evidence": "ansible_runtime / check-mode count / pending check-mode / blockers",
+              "nextAction": "先解除 ansible_playbook_binary_missing，再跑 check-mode，不直接 apply"
+            },
+            "approval": {
+              "owner": "Approval Coordinator + SRE owner",
+              "evidence": "approvals / risk gate / run timeline / manual_required reason",
+              "nextAction": "低風險才進自動化，高風險保留人工審批與 audit trail"
+            },
+            "verify": {
+              "owner": "Hermes + KM owner",
+              "evidence": "KM stale candidates / post-execution verification / playbook learning",
+              "nextAction": "Hermes 產草稿，owner 審核後寫入 KM，並觀察 stale ratio 下降"
+            }
+          }
+        },
         "values": {
           "verified": "verified {verified}/{evaluated}",
           "topGate": "{gate} 缺 {count}",