What problem does it solve? Writing effective SKILL.md files for AI agents is error-prone: vague triggers, bloated context, missing output contracts, and platform-specific quirks cause skills to under-trigger, over-trigger, or fail validation. This Skill provides a structured, research-backed workflow for engineering production-grade skills. ## Core Features & Use Cases - Eight operating modes: Create, Refactor, Review, Audit, Compress, Split, Merge, and Adapt skills across Claude Code, Codex, Copilot, Gemini, and Antigravity. - Deterministic validation gate: Bundled Python scripts check frontmatter rules, kebab-case naming, description length, banned docs, and broken local links, plus a 10-dimension scoring rubric. - Progressive disclosure architecture: Templates, references, and anti-pattern guides keep SKILL.md bodies lean while deep guidance loads on demand. - Use Case: Ask to "create a skill for processing CSV files" and receive a complete skill folder with validated frontmatter, numbered workflow, output contract, and passing rubric scores. ## Quick Start Ask the agent to create a skill for your target workflow, for example by saying "create a skill for generating release notes from merged pull requests".