verification-before-completion

Verification before completion requires running fresh verification commands and evaluating their outputs before asserting success. This process captures evidence like test results and build outputs to prevent premature completion claims and hidden defects in code changes.

A premature completion claim in code review occurs when success is asserted without fresh verification evidence. It is detected by enforcing a process that demands running tests, builds, and evaluations, requiring their outputs as proof before allowing the completion status to be declared.

To attach verification evidence before merging a pull request, run the full verification workflow for your task and capture the command outputs. If failures occur, report the status and root cause alongside the evidence instead of declaring the work complete.

Required verification steps include running tests, builds, code reviews, and deployment checks. The expected outputs must demonstrate successful execution for each claim, providing fresh evidence that satisfies the predefined completion criteria before success is declared.

You should not rely solely on automated verification when a task requires subjective human judgment or when verification commands cannot produce deterministic expected outputs. Automated guardrails are designed for objective evidence like test suites and build statuses, not nuanced evaluations.