Incident flow
Production problems are tracked as incidents attached to the task they relate
to — not as task statuses. Each task carries an incident log (incidents.json side
file), and incidents have their own status machine.
Incident lifecycle
reported → investigating → production-fix → fixed → verified → closed
| Status | Meaning |
|---|---|
reported | Incident created and logged. |
investigating | Root cause being diagnosed. |
production-fix | A fix is being applied to production. |
fixed | Fix applied. |
verified | Fix confirmed to resolve the incident. |
closed | Incident fully resolved and closed. |
Agent
The /task-incident agent handles incidents:
investigate, apply a production fix, and document the root cause. Production
problems are redirected here rather than being filed as new tasks.
CLI commands
insight-flow incident-create --id Nxx --title "..." --severity critical|high|medium|low
insight-flow incident-status --id Nxx --incident <iid> --status investigating|production-fix|verified|closed
insight-flow incident-resolve --id Nxx --incident <iid> --rootCause "..." --fix "..."
insight-flow incident-list --id Nxx
See the CLI: Incidents group for full flags. A task's open
incident count surfaces on its dashboard card via the openIncidentCount summary
field.