Audit advanced Works with: Any AI

Runbook and On-Call Documentation Audit

Table output Has example

Check that an operator at 3 a.m. can follow it.

The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.

Audit an operational runbook. Paste the runbook for one alert or service: [RUNBOOK_TEXT]. Last incident where it was used: [LAST_INCIDENT]. Skills of the people on call: [ONCALL_SKILLS].

Read it as a tired newcomer. Check that the first step states how to confirm the problem, commands are exact and safe, expected outputs are shown, escalation contacts are current, rollback steps exist, and dangerous actions carry warnings. Compare with what happened in the last incident.

Return a table with Step, Problem and Fix, a rewritten first page, and a drill plan to test the runbook with someone who did not write it.

Also suggest a standard layout for every runbook, with symptoms, checks, safe actions, rollback and escalation, and a review date. Add a rule that each incident ends with an update to the relevant runbook, and a quarterly test with a volunteer who has not used it before.

Finish with a one-line summary for the engineering lead.

Then design the on-call experience around the runbook. Create a template for alert descriptions that states impact, likely causes, the first diagnostic command, a decision tree for the next steps and the escalation path, with a link from the alert to the runbook. Define a drill programme: monthly game days in which an engineer who did not write the runbook follows it during a simulated failure in a non-production environment, with an observer noting every hesitation. Write a feedback form for after real incidents asking which runbook steps helped and which misled. Provide a review rotation so that every runbook is validated at least twice a year, and a dashboard of runbook age. Add guidance on writing for a tired reader, with short steps, safe defaults and warnings in capitals only when action could cause harm. Include a safe rollback for each risky action.

Fill in before use

Optional: type your details and the prompt above updates. Nothing is saved or sent.

Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.

Read-through (invented "High API latency" runbook)
Step 1: "Check the dashboard." Which dashboard? Link missing. Red.
Step 3: restart command given without saying which environment. Red; add warning.
Step 5: "Escalate to Dana" (Dana left in March). Red.
No rollback step after restart. Amber.
Last real incident: the on-call engineer needed 14 minutes to find the right log query.
Rewritten first page: symptom, impact, check A (exact command and expected output), check B, safe action, who to call, how to confirm recovery.
Drill: a colleague who did not write it follows it in staging; observer notes each hesitation.

• audit