This skill establishes a rigorous 'Iron Law' for Claude Code, preventing the AI from making premature or hallucinated claims of completion. By mandating a specific five-step gate (Identify, Run, Read, Verify, Claim), it ensures that every code change, bug fix, or feature implementation is backed by fresh, empirical evidence from test suites, build logs, or linters. It is designed to eliminate 'it should work' syndrome and build high-trust development workflows where every assertion is supported by an exit code and actual output.
주요 기능
01Systematic cross-referencing of requirements against actual command output
02Evidence-based reporting for tests, builds, and linting status
03Mandatory execution of verification commands before task completion
040 GitHub stars
05Detection and prevention of 'red flag' language like 'should' or 'probably'
06Rationalization prevention to stop premature commits and pushes
