The Completion Quality Gate acts as an automated senior reviewer for Claude Code, enforcing a strict separation of concerns between the agent that writes code and the agent that verifies it. By evaluating tasks against ten specific criteria—including specification compliance, linting, multi-level testing (unit, integration, E2E), security audits, and visual verification—it ensures that no subtask is marked as finished until it meets production-grade requirements. This skill is essential for maintaining code integrity in complex AI-driven workflows by providing structured feedback and automated retries when standards are not met, preventing the common pitfall of an AI 'grading its own homework.'
주요 기능
01Automated feedback loop for error correction and retries
02Comprehensive 10-point quality verification system
03Structured result logging in YAML for audit trails
040 GitHub stars
05Independent reviewer enforcement to prevent self-verification bias
06Integrated security audits and visual output checks
사용 사례
01Automating quality assurance for AI-generated pull requests
02Ensuring consistent linting, typing, and testing standards in autonomous workflows
03Verifying multi-step coding tasks against original project specifications
