# Operations runbook

## Service and ownership

- System/version, owner, on-call, and escalation contacts:
- Dashboards, traces, logs, and dependency status:

## Symptoms and first checks

| Symptom | Confirming signals | Likely causes | Safe first action |
|---|---|---|---|

## Containment

- Kill switch and traffic isolation:
- Read-only/degraded mode:
- Candidate/model/prompt/tool rollback:
- Secret or permission revocation:

## Recovery and verification

- Restore/reconcile steps:
- Data/effect integrity checks:
- Regression/eval suite:
- Criteria to resume traffic:

## Communication

- Incident channel, update cadence, and stakeholder templates: