verification-before-completion

Verifying code changes before completion requires running the exact test or build command, reading its complete output and exit code, and presenting that evidence instead of assuming the fix worked.

Code review can still fail because completion claims often lack verifiable evidence. Reviewers need fresh output from tests, linters, or builds to validate agent-produced changes rather than trusting premature success reports.

Enforce evidence-first testing by applying a verification gate that identifies, runs, and reads the exact command proving a claim, requiring full output and exit code before committing or creating a pull request.

A verification gate is a process that identifies, runs, and verifies the exact command proving a work claim. You need it for commit checks to prevent premature or dishonest completion claims by insisting on fresh, verifiable evidence.

Yes, you can validate agent-produced code changes by running the full verification command, reading its complete output and exit code, and requiring that evidence before preparing commits or creating a pull request.

Common false-claim scenarios include assuming tests, linters, or builds passed without reading output, or agents reporting success without exit codes. Required evidence means capturing full command output to prove readiness.