sre-runbook

name	sre-runbook
description	Write SRE runbooks for incident response, on-call procedures, and service recovery. Outputs structured runbooks with SLOs, alert response playbooks, escalation paths, and post-incident templates.
argument-hint	["service name","SLO targets","on-call rotation","alert types"]
allowed-tools	Read, Write

SRE Runbook

A runbook is executable documentation. When an alert fires at 3am, the on-call engineer should be able to open the runbook and follow steps to diagnose and recover — without needing to wake up the service owner or search Slack history.

Process

1. Define SLOs first. Availability %, latency p99, error rate. Without SLOs, you don't know what "broken" means.
2. Enumerate alert types. For each alert, write a dedicated playbook: what it means, immediate actions, escalation.
3. Write diagnosis steps. Structured decision tree: check X, if Y do Z, else do W.
4. Document recovery procedures. Exact commands, not vague instructions. Tested and verified.
5. Define escalation paths. Who to call and when. Include contact info and rotation.
6. Add runbook links to alerts. Every PagerDuty/alertmanager alert links directly to its runbook section.
7. Review after every incident. If the runbook didn't help, update it. Runbooks rot.

Service Overview Template