verification-before-completion

To stop premature code completion claims, enforce a strict verification workflow that requires executing test suites, linters, and builds, then analyzing their fresh output before making any success assertions. This evidence-based approach prevents false positives and costly rework.

The best way to verify bug fixes is to run the specific verification commands fresh, read the full output to confirm zero failures, and only then claim completion with the attached output evidence. Avoid relying on outdated test results or external reports.

Linter success does not guarantee build correctness because linting only checks code style and syntax, not compilation or runtime behavior. You must execute build and test commands separately and analyze their empirical output to prove overall code quality.

To enforce an evidence-based testing workflow before committing code, identify the command that proves your claim, execute it fresh, read the full output, verify it confirms the claim, and only then state completion with the output evidence.

You do not need external tools to verify agent task completion claims, but you must enforce a gate function that prioritizes empirical evidence over assumptions. Execute verification commands directly and analyze the output instead of trusting external reports.

You should never trust outdated test results for completion claims. Always execute verification commands fresh and read the full output to confirm zero failures before declaring any task complete, fixed, or passing.