| name | incident |
| description | Incident response workflow for production issues |
| when_to_use | Use when production is broken and needs structured response — classify severity (P0-P3), diagnose via Sentry traces + Vercel logs, fix or roll back, verify recovery, write the postmortem on a GitHub issue. Triggers on: incident, production is down, outage, P0, users can't log in. |
| argument-hint | <description> |
Incident response workflow for production issues.
Arguments: $ARGUMENTS (description of the incident, e.g., "hogwarts login page returning 500")
Steps:
- Classify severity:
- P0: Service completely down, customers affected
- P1: Service degraded, some features broken
- P2: Minor issue, workaround available
- Gather context:
- Check Sentry for related errors (sentry MCP)
- Check Vercel deployment logs (vercel MCP)
- Check Neon database status (neon MCP)
- Check recent git commits that might have caused it (github MCP)
- Diagnose:
- Use sse agent for server-side error analysis
- Check if the issue is in the latest deployment
- Determine if rollback is needed
- Respond based on severity:
- P0: Alert Abdout via Slack immediately, recommend rollback
- P1: Create Linear issue, delegate to tech-lead
- P2: Log for next sprint, inform support agent
- Track:
- Create Linear issue with all context
- Start resolution timer
- Update support agent for customer communication
- Post-mortem (after resolution):
- What broke and why
- How was it detected
- How to prevent recurrence
- Update guardian agent's checklist if needed
Reference: .claude/agents/ops.md for incident workflow
Reference: .claude/agents/guardian.md for security implications