Use when the user wants to: spawn, stop, or monitor agents; send messages or follow-ups; run agents in parallel; collect results; schedule recurring work with triggers; or diagnose daemon issues. Trigger on any mention of 'agentd', and also on multi-agent patterns like 'run these in parallel', 'spawn a child agent', 'set a timed wakeup', or 'route a webhook into an agent'. Do NOT trigger for generic subprocess/asyncio, Celery/task-queue, or systemd/launchd questions unrelated to agentd.
2026-07-05