What problem does it solve?
Capture, evaluate, and iteratively improve Claude Skill definitions and workflows so authors can reliably produce repeatable, testable, and well-triggering skills without manual trial-and-error.
Core Features & Use Cases
- Skill authoring workflow: Guided process for drafting SKILL.md, defining triggers, and organizing scripts/references/assets for a new skill.
- Evaluation & benchmarking: Tools and scripts to run trigger/eval suites, spawn with-skill vs baseline comparisons, aggregate metrics, and surface pass rates, timing, and token usage.
- Iteration & optimization: Automates drafting assertions, running with/without-skill baselines, generating an HTML review viewer, and optimizing SKILL.md descriptions to improve trigger accuracy.
- Use case: A skill maintainer wanting to add a new capability can draft a skill, generate test prompts, run parallel evaluations, inspect outputs in the viewer, collect feedback, and iterate until results meet acceptance criteria.
Quick Start
Tell the assistant to help you create a new Claude skill by drafting SKILL.md, authoring test prompts, running the eval loop, reviewing results in the viewer, and iterating on the skill and description until benchmarks and human feedback are satisfactory.
