ferrox-eval-review

To audit evaluation coverage gaps in a completed AI phase, review the implemented work against the project's AI spec. This process identifies missing tests and weak verification, producing a structured EVAL-REVIEW report with scores and remediation guidance.

An EVAL-REVIEW report is a structured document generated by auditing a completed AI phase. It captures evaluation coverage findings, identifies missing verification strategies, provides a final score, a verdict, and a concrete remediation plan to fix discovered gaps.

Yes, a project AI spec is required to check for missing evaluation cases. The audit process reads project workflow references to compare planned evaluation strategies against actual implementation, ensuring the phase meets documented AI-SPEC expectations.

The best way to generate remediation guidance for incomplete review gates is running a retroactive evaluation audit. This produces a concrete fix plan with a score and verdict, allowing teams to methodically improve weak coverage areas before issues ship.

You should run an evaluation coverage audit immediately after an AI feature phase is done. This retroactive review checks evaluation rigor, catches missing verification steps, and ensures the planned evaluation strategy was actually carried through before shipping.