What problem does it solve? Long-running AI agent systems accumulate silent failures: stuck tasks, retry loops, exposed secrets, and missing logs. This Skill automates periodic background health checks so operators catch degradation early without manually inspecting every subsystem. ## Core Features & Use Cases - Five-Step Health Workflow: Checks gateway/database/process liveness, audits tasks stuck in RUNNING or BLOCKED states, reviews artifacts and file permissions, scans for unauthenticated bindings or secret exposure, and generates a daily digest. - Token-Efficient Operation: Returns HEARTBEAT_OK and skips reprocessing when no state has changed, batching email, calendar, health, and alert checks into a single turn. - Escalation Rules: Flags security risks and failures persisting longer than 48 hours for developer attention. - Use Case: An operations engineer schedules a daily run that audits an agent orchestrator deployment, writes a digest to _audit/digests/YYYY-MM-DD_digest.md, and only escalates when a task has been retry-looping for two days. ## Quick Start Invoke the heartbeat-monitoring workflow with the daily argument to run a full system health check and generate today's digest.
