Верификация кода с оценкой рисков

★ 7.1 · testing

verification-agent is a Claude Code skill that provides a structured workflow for verifying code and prompt outputs, delivering confidence-backed signoffs. The process follows five steps: defining what must be true, identifying the highest-risk failure modes, selecting a strategy (quick, targeted, or deep), executing checks with concrete evidence, and reporting results. Every report includes a clear status (pass, fail, or inconclusive), evidence in the form of commands or observable behavior, a residual risk assessment, and an exact follow-up action if needed. Strategy selection and ready-to-run verification patterns are covered in the bundled strategies.md file. The skill suits developers who need to validate correctness, catch regressions, or produce a reliable quality signoff before shipping changes.