Write a Runbook

A runbook answers the question: "It's 3am, this service is paging me, what do I do?" Without one, every incident takes 3x longer because the on-call engineer is figuring out what the service even does. This skill produces that document.

When to use

• Before a service goes to production.
• After an incident — the post-mortem reveals gaps in the runbook.
• Handing off ownership of a service to another team.
• Onboarding a new on-call rotation.

When NOT to use

• Stateless one-off tools nobody pages on — runbooks are for things with availability requirements.
• Internal scripts, dev tooling — overkill.

The runbook structure

A runbook is operational reference, not architecture documentation. It optimizes for the 3am case.