Verification Before Completion

소개

This skill implements a rigorous 'Iron Law' for Claude Code, preventing premature or false claims of success by mandating the execution of specific verification commands before any completion assertions. It acts as a gate function that identifies necessary proof—such as passing tests, clean lints, or successful builds—and requires the agent to analyze actual output logs rather than relying on assumptions or previous states. By eliminating 'should-work' rationalizations and requiring hard evidence before commits or PRs, it significantly reduces regressions and maintains a high level of trust in automated workflows.

주요 기능

• Mandatory 'Iron Law' verification protocol for all success claims
• Independent verification of delegated agent tasks and VCS diffs
• Red-green cycle validation for robust regression testing
• Gate function identifying proof-of-work commands before status updates
• Anti-rationalization framework to prevent false confidence
• 1 GitHub stars

사용 사례

• Ensuring zero-failure test results before creating a pull request
• Systematic requirement auditing against a checklist for complex feature implementation
• Validating full build compilation after refactoring to prevent broken environments