What problem does it solve? When a persistent Claude Code agent misbehaves — memory search returns nothing, heartbeats stop firing, or the browser bridge won't connect — there is no built-in way to find the root cause. This Skill runs structured diagnostic checks across the entire agent workspace and optionally applies safe auto-repairs. ## Core Features & Use Cases - Comprehensive health checks: Inspects config, identity, memory, SQLite index, QMD, crons, hooks, HTTP bridge, messaging plugins, and dreaming via the agent_doctor MCP tool. - Safe auto-repair mode: With --fix, creates missing memory directories, syncs the index, and deletes stale BOOTSTRAP.md files before re-running checks. - Cron audit integration: Refreshes the cron registry audit from live CronList output so heartbeat and dreaming schedules are reported accurately. - Use Case: After running /agent:create, invoke the doctor to verify the setup is clean, or run it with --fix when memory search suddenly returns empty results. ## Quick Start Ask the agent to run a health check on the workspace, or say "run doctor with --fix" to also apply safe automatic repairs.