system-diagnostics

To detect stalled agent tasks, you run a system diagnostics health check that identifies stale in-progress work, pending items, and overdue human tasks. This routine heartbeat execution surfaces stuck operations early so you can recover stalled work before it causes invisible stalls.

Fleet heartbeat monitoring aggregates agent heartbeat status to highlight agents that haven't progressed recently. It checks operational metrics and fleet heartbeat status during morning reviews to provide a health summary showing which agents are active and which are quiet.

You check goal staleness by reading and evaluating system state via bus commands that detect agents whose GOALS.md hasn't been updated within a configurable threshold. This troubleshooting step reveals outdated goals causing agents to stop moving or go quiet.

Yes, you can surface overdue human-assigned work by executing the system diagnostics sequence for human tasks. It monitors pending work waiting on humans without updates and prompts reminders so overdue items appear in your critical findings and alerts.

The best way to collect operational metrics for task triage is to run a full diagnostics sequence that captures a snapshot of task and agent activity. This metrics collection provides a snapshot for later review alongside stale task detection and fleet heartbeat status.

You should run system diagnostics for operations monitoring during routine heartbeat execution, morning reviews, and troubleshooting moments when tasks stop moving or agents go quiet. It evaluates stale tasks, goal staleness, human tasks, and fleet heartbeats to report critical findings.