darwin-skill - Evaluate and optimize SKILL.md files
Evaluates and iteratively improves SKILL.md files through rubric scoring, test prompts, git versioning, user checkpoints, and result cards.
Tags
Updated: 2026-10-04Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Score skills across eight dimensions
- Design representative test prompts
- Compare skill and baseline outputs
- Optimize one dimension per round
- Keep only scored improvements
- Revert unsuccessful changes
- Record optimization results
- Generate visual result cards
Inputs
- Target SKILL.md files
- Optimization scope
- User confirmations
- Existing results.tsv
- Existing test-prompts.json
- Result card template
Outputs
- Scored evaluation reports
- Modified SKILL.md files
- test-prompts.json files
- results.tsv entries
- Git branches and commits
- Reverted changes
- PNG result cards
Requirements
- Filesystem access
- Git for version control
- Subagent support for full testing
- Model capable of editing files