What problem does it solve?
Helps authors and teams reliably create, test, and iterate Claude Code skills by providing a repeatable workflow for drafting SKILL.md files, generating test prompts, running with-skill vs baseline evaluations, and packaging deliverables. It reduces manual validation work, prevents regressions, and surfaces objective metrics so improvements are evidence-driven.
Core Features & Use Cases
- Guided Skill Authoring: Prompts and templates to capture intent, trigger conditions, output formats, and examples for a new or existing skill.
- Eval & Benchmarking Workflow: Create eval sets, spawn parallel runs, capture timing/tokens, grade expectations, and aggregate benchmark summaries.
- Iteration Loop: Draft descriptions, run train/test splits, automatically propose improved descriptions, and re-evaluate until results stabilize.
- Packaging & Presentation: Tools to package a validated skill into a distributable .skill file and generate human-facing review pages for qualitative feedback.
- Use Case: Turn a rough idea ("convert invoices to CSV") into a production-ready skill with test cases, graders, benchmark reports, and a polished SKILL.md.
Quick Start
Ask the assistant to draft a SKILL.md for your intent, propose 2–3 realistic test prompts, and run the evaluation loop to generate initial benchmark results.
