What problem does it solve? Writing a skill that triggers reliably and produces good outputs is hard to do blind. This Skill guides the full lifecycle of building GitHub Copilot CLI skills: drafting SKILL.md files, running test prompts with and without the skill, grading outputs against assertions, and iterating based on quantitative benchmarks and user feedback. ## Core Features & Use Cases - Guided Skill Authoring: Interviews the user to capture intent, then writes a SKILL.md with a triggering-optimized description, progressive disclosure structure, and bundled scripts/references/assets. - Evaluation Harness: Spawns paired with-skill and baseline runs via sub-agents, grades outputs against assertions, and aggregates results into benchmark.json with pass-rate, timing, and token statistics. - Interactive Review Viewer: Serves a local HTML viewer (eval-viewer/generate_review.py) where users inspect outputs side-by-side, leave feedback, and compare iterations. - Description Optimization: Generates should-trigger and should-not-trigger eval queries, reviews them in an editable HTML page, and iteratively rewrites the skill description to improve triggering accuracy. - Use Case: You want a skill that formats weekly sales spreadsheets. This Skill helps you draft it, runs three test prompts with and without the skill, shows you a benchmark comparing pass rates, and loops until the outputs satisfy you. ## Quick Start Help me create a new Copilot CLI skill that converts meeting notes into action items, then test it with a few prompts.
