Runbooks
Operational runbooks for running, debugging, and recovering the Curate-Me platform. Each runbook is a focused, step-by-step procedure for a specific situation.
Incidents & recovery
- Incident response — triage, escalation, and communication
- Incident update template — post-incident review
- MongoDB incident — recovery and backup restoration
- Redis incident — recovery and session restoration
- Runner stuck — recover or force-terminate a stuck runner
- Runner provision failure — diagnose launch failures
- Webhook disaster recovery — recover webhook delivery
- Launch flake handling — flaky runner launches & retries
Gateway & governance
- Gateway high latency — diagnose performance degradation
- Rate limit hit — handle rate-limit errors & backoff
- PII blocked — unblock PII detection false positives
- Governance cascading denials — policy chain failures
- Provider failover loop — resolve failover chain issues
- Cost accumulation lag — reconcile billing report delays
- Budget exceeded — handle cost overruns & spending controls
Operations & deployment
- Deployment procedure — deploy to the production VPS
- Runner operations — scaling, provisioning, deprovisioning
- Rotate Claude Code OAT — rotate the Claude Code OAuth token (current)
- OAuth token rotation — legacy token-push flow (see rotate-claude-code-oat)
- Agent identity provisioning — Microsoft 365 agent identities
- Slack integration — bot, token, and webhook setup
- Onboarding — first-customer onboarding sequence
See also: Template runbooks — starting points for common autopilot template workflows.