Ensures code quality by requiring fresh command-line evidence before claiming any task is finished.
This Claude Code skill enforces a rigorous 'Iron Law' of verification, preventing the AI from prematurely claiming tasks are complete without providing verifiable terminal output. It mandates a structured workflow—Identify, Run, Read, Verify, and Claim—to ensure that every bug fix, feature, and build is backed by current results rather than assumptions. By eliminating speculative language and requiring fresh evidence for every claim, it drastically reduces rework, builds user trust, and prevents the deployment of broken or incomplete code.
Key Features
01Mandates fresh command-line evidence for all completion claims
02Provides a 'Red Flag' system to interrupt premature task handoffs
03Structured verification gate system (Identify, Run, Read, Verify, Claim)
040 GitHub stars
05Bans speculative language like 'should work' or 'probably fixed'
06Validates test passing, build success, and linter cleanliness
Use Cases
01Verifying bug fixes by running reproduction scripts before reporting them solved
02Confirming linter compliance and test suite coverage across newly modified files
03Ensuring local builds pass successfully before initiating a git commit
