impl-validator

To validate agent output, you pass a structured check block to the validator, which identifies discrepancies between intended goals and produced results, returning a structured pass, warn, or fail verdict with specific issues to fix.

A structured check block is the readable input format required by the validator to evaluate produced results against stated goals, enabling the system to surface mismatches and return a structured pass, warn, or fail verdict.

Yes, you can use this validator in subagent mode to automatically review outputs generated by other skills or agents across iterative development cycles, returning actionable issues and a clear pass, warn, or fail verdict.

You enforce quality gates by using the validator's structured pass, warn, and fail verdicts to automatically approve, flag, or block agent outputs, enabling rapid improvement cycles by returning actionable issues in a consistent format.

The best way to surface mismatches is to use an automated validator that systematically checks implementation alignment, identifying and reporting discrepancies before they reach end users with a structured verdict.

You should not use an automated output validator when you lack a readable check block input, as the validator requires this structured input to evaluate results and return a structured pass, warn, or fail verdict.