cyberneticlibrary

Audit evaluation coverage for phases

gsd:eval-reviewskillsetup L13,035
davepoon/buildwithclaude
What it does

Retroactively audit phase evaluation coverage

Best for

Finishing a phase when you want to verify evaluation coverage aligned with the plan, or identifying what post-launch monitoring is needed.

Inputs
  • · phase number (optional, defaults to last completed)
Outputs
  • · EVAL-REVIEW.md
  • · COVERED/PARTIAL/MISSING scores
  • · remediation plan
Requires
  • · Read
  • · Write
  • · Bash
  • · Glob
  • · Grep
  • · Task
  • · AskUserQuestion
Preconditions

Phase directory exists; AI-SPEC.md documents evaluation strategy; phase has been executed

Failure modes

AI-SPEC.md missing or incomplete; phase not yet executed; evaluation dimension unclear

Trust signals
  • · Scores each eval dimension as COVERED/PARTIAL/MISSING
  • · Produces actionable remediation plan
  • · References AI-SPEC.md strategy
  • · Audits post-execution (catches missed tests)