evidence-proof

You can verify code changes with evidence by automating test execution, capturing pass/fail outputs, manually checking critical workflows, and documenting any residual risks to ensure correctness before deployment.

A confidence score in code verification is a numerical rating assigned based on a defined rubric, evaluating test results and manual checks to objectively quantify the reliability of recent code changes.

To gather proof that a bug fix is correct, execute automated tests like npm test, verify the fix manually on a local server, and generate a confidence report detailing test outputs, remaining risks, and a rollback plan.

Automated evidence gathering captures detailed test outputs including pass/fail counts, coverage percentages, and duration, providing concrete proof that all code changes are backed by systematic testing.

You should document unknowns and residual risks during the confidence scoring phase, ensuring that any incomplete verifications or potential edge cases are explicitly recorded before considering a code change complete.

Generating a rollback plan is a core use case after verifying code changes, where the skill details test results, remaining risks, and rollback procedures to prevent the deployment of untested features.