Runbooks Way¶
Source: hooks/ways/itops/runbooks/runbooks.md
Frontmatter
| Field | Value |
|---|---|
description |
Runbook automation, executable playbooks, SOPs as code, operational procedures |
vocabulary |
runbook playbook sop procedure operational automation executable standard checklist step-by-step operations |
pattern |
runbook|runbook.?(automation|executable)|playbook|sop.?(automation|as.?code)|operational.?procedure |
refire |
0.15 |
scope |
agent, subagent |
What Claude Should Do¶
When handling operational tasks, check for existing runbooks before writing ad-hoc scripts.
Creating a Runbook¶
A runbook documents a repeatable operational procedure:
runbooks/
└── procedure-name/
├── runbook.sh (or .py, .ts) # The procedure
└── README.md # When/how to use
The Procedure¶
- Include error handling for every external call (API, SSH, database)
- Ensure idempotent execution — safe to run twice
- Log each step's outcome for post-incident review
- Exit with clear success/failure status and message
The README¶
- When to use: trigger conditions, symptoms
- Prerequisites: required access, credentials, tools
- Expected outcome: what success looks like
- Escalation: when this runbook isn't enough
Runbook Quality Standards¶
- Test in a non-production environment before relying on it
- Include a rollback or "undo" section where applicable
- Keep steps atomic — each step succeeds completely or fails cleanly
- Don't mix diagnosis with remediation — separate "find the problem" from "fix it"
When a Runbook Doesn't Exist¶
If handling a novel operational issue: 1. Resolve the incident first 2. Then offer to create a runbook from the steps taken