メニュー
Use when detecting or classifying agent failures. Trigger with health check failures, heartbeat timeouts, or error events. Loaded by ai-maestro-chief-of-staff-main-agent
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when applying or checking GitHub labels for agent assignment and status tracking. Trigger with label management or taxonomy lookup requests. Loaded by ai-maestro-chief-of-staff-main-agent
CHIEF-OF-STAFF's role in the PRRD / TRDD / Kanban workflow. COS does NOT own kanban columns directly — it ROUTES messages and proposals between the team layer (ORCH/ARCH/INT/MEMBER) and the governance layer (MANAGER). Use when COS forwards a PRRD proposal, relays approval decisions, broadcasts TRDD status updates, or aggregates team status for AMAMA.
Use when sending failure notifications after operation errors. Trigger with failure events, error reporting, or incident notification. Loaded by ai-maestro-chief-of-staff-main-agent
Use when managing team registry, inter-agent messaging, role boundaries, or sub-agent delegation. Trigger with team coordination tasks. Loaded by ai-maestro-chief-of-staff-main-agent
Use when requesting or processing agent acknowledgments before proceeding with operations. Trigger with acknowledgment requests or ACK timeout handling. Loaded by ai-maestro-chief-of-staff-main-agent
Use when replacing a failed agent with a new instance. Trigger with agent replacement or failover events. Loaded by ai-maestro-chief-of-staff-main-agent
| name | amcos-failure-detection |
| description | Use when detecting or classifying agent failures. Trigger with health check failures, heartbeat timeouts, or error events. Loaded by ai-maestro-chief-of-staff-main-agent |
| user-invocable | false |
| license | Apache-2.0 |
| compatibility | Requires AI Maestro installed. |
| metadata | {"author":"Emasoft","version":"1.0.0"} |
| context | fork |
| agent | ai-maestro-chief-of-staff-main-agent |
Detect and classify agent failures in a multi-agent system coordinated via AI Maestro. Covers heartbeat monitoring, message delivery failure detection, and severity classification.
$CLAUDE_PROJECT_DIR/.amcos/agent-health/heartbeat-config.json$CLAUDE_PROJECT_DIR/.amcos/agent-health/task-tracking.json| Mechanism | Signal | Response Time |
|---|---|---|
| Heartbeat timeout | Missed pings | 30-60 seconds |
| Message delivery failure | API error | Immediate |
| Acknowledgment timeout | No ACK | 5-15 minutes |
| Task completion timeout | Stalled progress | Variable |
amcos-recovery-execution; Terminal -> amcos-agent-replacement| Category | Recovery | Example |
|---|---|---|
| Transient | Auto (< 5 min) | Network hiccup, rate limit |
| Recoverable | Intervention | Session hibernated, OOM |
| Terminal | Replacement | Host crash, disk corruption |
DETECT --> CLASSIFY --> RESPOND
Transient --> Wait & Retry
Recoverable --> amcos-recovery-execution
Terminal --> amcos-agent-replacement
| Result | Next Action |
|---|---|
| Transient detected | Wait 5 min, verify auto-recovery |
| Recoverable detected | Route to amcos-recovery-execution |
| Terminal detected | Route to amcos-agent-replacement |
| Issue | Resolution |
|---|---|
| Agent unresponsive to ping | Wait 30s, retry, then classify |
| Cannot determine failure type | Default to recoverable |
| AI Maestro API unreachable | Use fallback file-based communication |
# Send health check ping to suspect agent
amp-send.sh ampa-svgbbox-impl "Health check" high '{"type":"ping"}'
Expected: agent responds within 30s; if no response, classify as failure.
Copy this checklist and track your progress: