Critique the Incident Runbook Everyone Signed Off but Nobody Tests

The runbook got signed off in a meeting and nobody has opened it since, which is exactly the problem. This checks every step is specific enough to follow under pressure, that escalation is clear, and that nothing in it assumes knowledge only the person who wrote it actually has.

<context>
You are a meticulous site reliability engineer who reviews incident runbooks specifically for whether they would actually work during a real incident, not just whether they read well in a document review. {SYSTEM_NAME} is the system the runbook covers.
</context>

<paste>
[paste the incident runbook here]
</paste>

<task>
**Critique the runbook:**
1. Check every step is specific enough to follow under pressure: exact commands, exact dashboards, exact thresholds, not vague instructions like "check if the service is healthy"
2. Identify any step that assumes access, tooling, or knowledge that an on-call engineer who did not write the runbook might not have
3. Check the runbook includes a clear escalation path and a clear point at which to stop trying and escalate
4. Flag anything that looks like it was true once but may now be outdated, based on internal inconsistencies in the document
</task>

<output_format>
- A short verdict: usable under pressure, needs specific fixes, or unsafe to rely on
- Step by step notes, only for steps that need changing, each with the reason
- A rewritten version of the two weakest steps
- Tone: direct and unsparing, written as if this runbook will be used at 3am by someone who has never seen it before
</output_format>