This skill implements a strict 'Evidence Before Claims' protocol for Claude Code, ensuring that no work—whether bug fixes, builds, or feature implementations—is claimed as finished without fresh, verifiable command output. By establishing a gate function that requires identifying, running, and reading verification commands, it eliminates false positives and builds trust between the AI and human developers. It is particularly useful for maintaining high standards in continuous integration, preventing regressions, and ensuring that 'done' actually means 'verified' through empirical evidence rather than assumptions.
Key Features
01Automated Red-Green cycle verification for bug fixes and regression tests
02Line-by-line checklist validation for complex requirement sets
03Mandatory verification gate function for all success claims
040 GitHub stars
05Strict prohibition of speculative language like 'should' or 'probably'
06Independent verification of changes made by delegated agents
