done

To verify a code change is ready to ship, you need an evidence-driven release gate that runs tests, linting, builds, and a real execution of the change rather than relying on assertions. This validates behavioral proof and diff hygiene.

A fake-green test is a check that passes falsely due to edited tests, skipped checks, or hardcoded expectations. You catch it before releasing by using safety tripwires that detect these destructive patterns and flag source-security risks during diff inspection.

To perform a fresh-context review on a feature branch diff, you assess spec compliance and behavioral correctness using only the diff and intent. This independent review workflow ensures the change meets requirements without relying on prior context.

You can use linting and typecheck results as part of a release gate, but they are not sufficient alone. A robust release gate requires a full test suite, successful builds, and an actual run of the change to provide genuine behavioral proof.

You should require diff inspection before drafting a pull request whenever finishing a feature branch or bug fix. It ensures the change is reviewed for spec compliance, behavioral correctness, and source-security risks before committing.

The best way to prevent agentic coding from shipping unfinished changes is to enforce an evidence-driven release gate. This requires quoted evidence that the work was run, tested, and reviewed, applying tripwires for fake-green tests before drafting a commit.