Runbook Generator Skill

Generate operational runbooks for alerts, deployments, and common procedures.

Trigger Conditions

• •New alert rule created
• •New service deployed
• •User invokes with "generate runbook" or "create runbook"

Input Contract

• •Required: Alert definition or service context
• •Required: Expected failure mode or procedure
• •Optional: Historical incident data, existing runbooks

Output Contract

• •Runbook document with step-by-step procedures
• •Diagnostic commands and expected outputs
• •Escalation contacts and criteria
• •Verification steps

Tool Permissions

• •Read: Alert configs, service configs, monitoring dashboards, logs
• •Write: Runbook documents in docs/runbooks/
• •Search: Similar alerts and past incidents

Execution Steps

1. •Analyze the alert or service requiring a runbook
2. •Identify likely failure modes and diagnostic steps
3. •Write step-by-step mitigation procedures
4. •Include diagnostic commands with expected outputs
5. •Define escalation criteria and contacts
6. •Add verification steps to confirm resolution
7. •Link runbook to alert configuration

Success Criteria

• •Runbook covers the most common failure scenarios
• •Each step has a concrete command or action
• •Escalation path defined
• •Runbook linked to its corresponding alert

Escalation Rules

• •Escalate if the runbook requires infrastructure access not available to on-call
• •Escalate if the failure mode has no known mitigation

Example Invocations

Input: "Generate a runbook for the 'payment-service-high-error-rate' alert"

Output: Runbook: 1) Check error logs (kubectl logs -l app=payment-service --since=5m), 2) Verify upstream dependencies (curl health endpoints), 3) Check recent deploys (gh run list --limit 5), 4) If deploy-related: rollback (kubectl rollout undo), 5) If dependency-related: check circuit breaker status, 6) Escalate to payments-team if unresolved in 15min.