gsd-eval-review

Audit AI development phases by retroactively comparing execution logs against defined specifications. This process identifies unverified code and technical debt by verifying if the evaluation strategy stated in AI-SPEC.md was actually executed.

Generate a remediation plan by running a retroactive evaluation audit to identify test coverage gaps. This produces a structured EVAL-REVIEW.md file containing scores, verdicts, and actionable steps to fix missing edge case validations.

Evaluation coverage gap analysis is a quality assurance process that compares completed execution logs against project-specific evaluation standards. It documents missing test coverage and unverified performance benchmarks to ensure compliance with the initial specification.

Yes, an AI-SPEC.md file is required. The audit process checks execution logs against the defined evaluation strategy stated in the specification, meaning the audit cannot verify compliance without this foundational project-specific document.

Run an evaluation coverage audit after completing a complex feature implementation. This ensures all edge cases and performance benchmarks defined in the initial specification were properly tested and validated before moving to the next phase.

Automated evaluation audits are limited to comparing execution logs against existing defined specifications. They cannot invent new evaluation criteria or test unlogged behaviors, meaning any gaps in the initial AI-SPEC.md will remain unaddressed by the audit.