watchdog

To automatically resume a stalled pipeline, configure `.pipeline/config.yaml` to enable the watchdog and ensure cron is running; it reads state, monitors heartbeat and lockfile, then triggers resume to continue progress without manual intervention.

A cron watchdog is a periodic monitor that reads `.pipeline/state.yaml` and heartbeat signals within a cron-driven loop to detect stalled executions, managing a lockfile to prevent concurrent resumes before triggering a pipeline resume.

To prevent race conditions when auto-resuming cron pipelines, the watchdog applies safe lock handling by managing a lockfile, ensuring isolation from user data and preventing concurrent resume triggers from overlapping.

Yes, you need cron running to use this pipeline heartbeat monitor, because the watchdog registers a cron entry to drive periodic checks of `.pipeline/state.yaml` and handle retries and notifications automatically.

Pipelines stall when heartbeat signals fail or execution hangs; the watchdog logs retries and notifications by writing to `.pipeline/watchdog.log` while monitoring `.pipeline/config.yaml` and state files.