What problem does it solve?
Reduce friction and iteration time for authoring Claude skills by providing a complete toolkit to initialize skill templates, validate frontmatter, run trigger and output evaluations, and package skills for distribution.
Core Features & Use Cases
- Skill Initialization: Template-based initialization to scaffold SKILL.md, scripts/, references/, and assets/.
- Eval & Benchmarking: Parallel with-skill vs baseline evals, grader/comparator/analyzer agents, aggregation, and interactive review viewer for human feedback.
- Description Optimization & Packaging: Iterative description improvement (improve_description.py, run_loop.py) and packaging/validation tooling to produce distributable skill archives.
- Use Case: Create a new skill, run automated trigger and output tests, optimize the description to improve activation, aggregate benchmark results, and package the skill for sharing.
Quick Start
Ask Claude to initialize a skill folder, run the automated eval loop to optimize the description, and package the skill using the included scripts.