What problem does it solve? Skill authors need an independent, clean-context check that a finished or modified skill conforms to plugin conventions before it goes live. This Skill acts as that reviewer: it validates naming, mandatory frontmatter fields, model/effort pairing, and instruction quality, then returns a verdict with findings ranked by severity — without ever editing the skill itself. ## Core Features & Use Cases - Convention compliance check: Verifies kebab-case naming, the mandatory frontmatter five (name, description, when_to_use, model, effort), the 1536-character description limit, and justification comments for omitted fields. - Instruction quality review: Checks imperative style, triggering boundaries against neighboring skills and agents, allowed-tools fit, and path anchoring rules depending on where the skill lives (plugin vs. project .claude/skills/). - Structured verdict output: Returns Approve or Return-for-rework with a findings table classified as Blocking, Recommendation, or Note, plus verification of unresolved findings from previous review rounds. - Use Case: After write-skill finishes a draft, it spawns this reviewer as a subagent each iteration; the reviewer reads the skill and the write-skill conventions, reports blocking findings, and the loop repeats until approval. ## Quick Start Ask the assistant to review the skill at a given path against the plugin conventions and return the verdict with findings.
