Create golden test sets

This skill is for Code Review & Testing and helps you evaluate llm model performance, score agent trajectories, and assess rag system quality.