spboyer/sensei

# Summary for Developer Directory Implements an iterative compliance loop using Claude API calls to enhance Agent Skills' frontmatter (triggers, anti-triggers, compatibility metadata) while enforcing test coverage and token budgets. The system runs rule-based scoring, LLM-driven improvements, Jest/pytest validation, and token analysis in cycles until Medium-High compliance is achieved, with optional GEPA evolutionary optimization mode for deep refinement. Integrates with Anthropic's Copilot CLI framework and supports both single-skill and batch operations across skill directories.