What problem does it solve? Writing or refactoring agent skills by hand often produces bloated SKILL.md files, inconsistent frontmatter, and lint failures. This Skill turns a short brief into a convention-compliant skill, or conservatively splits an oversized SKILL.md into layered references without changing behavior, and refuses to hand back output that fails the lint and completeness gate. ## Core Features & Use Cases - Author mode: Interactively interviews you for goal, trigger phrases, tracker interaction, and chain membership, then scaffolds a new skills/<name>/ directory with a router-style SKILL.md, references, and optional repo-local stub. - Split mode: Refactors an existing oversized SKILL.md by moving sections word-for-word into references/ files, preserving the frontmatter description byte-for-byte so routing never breaks. - Hard quality gate: Runs scripts/lint.sh plus split-completeness and readability checks before returning, and supports --dry-run to preview planned files. - Use Case: You say "create a skill for reviewing database migrations" — the Skill asks the few questions that matter, drafts the router body from shared templates, pushes detail into references/, and delivers a lint-clean skill ready to commit. ## Quick Start Ask the assistant to create a new om-skill for your task, or to split an existing oversized SKILL.md into references, and answer the short interview questions it poses.