Lightweight Skills Eliminate AI Agent Guesswork and Unreliable Code Reviews

Claude Code agents often fail not in code generation but upstream (vague task interpretation) and downstream (insufficient review). These two open-source, dependency-free skills solve both with explicit specifications and evidence-based audits, enabling verifiable, production-ready agent-assisted development.