review-agent-skills

Reviews Claude Code skill files against authoring best practices and schema rules.

Updated May 19, 2026
One-click install
npx skills add https://github.com/kmosher/claude-plugins --skill review-agent-skills-kmosher
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: review-agent-skills
Source: https://github.com/kmosher/claude-plugins/tree/main/plugins/kmosher-review/skills/review-agent-skills
Command: npx skills add https://github.com/kmosher/claude-plugins --skill review-agent-skills-kmosher

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Claude Code skills often fail silently: descriptions that never trigger, frontmatter that drifts from the schema, broken references to bundled scripts, or missing side-effect guards. This Skill reviews PR diffs touching skill files and catches those authoring defects before merge. ## Core Features & Use Cases - Frontmatter and description validation: Enforces schema rules for name, description length, trigger phrasing, voice, and allowed-tools syntax against canonical Anthropic references. - Structural and safety checks: Verifies referenced files exist, scripts use ${CLAUDE_SKILL_DIR}, side-effecting skills set disable-model-invocation: true, and renames update all cross-references. - Severity-calibrated findings: Emits JSONL findings tagged with rule codes (e.g. R-DESCRIPTION-TRIGGERS), source docs, and P0–P3 severity so authors can verify each claim. - Use Case: A PR adds a new skill under plugins/foo/skills/bar/SKILL.md. Run this lens to catch that its description starts with "Use this skill when…", its body is second-person, and its scripts/run.sh is never referenced. ## Quick Start Review the skill file changes in this PR for authoring issues and report findings with rule codes and severity.

Frequently Asked Questions about review-agent-skills

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review Claude Code SKILL.md files in a pull request?▼

Run this review lens on any PR diff touching SKILL.md, command, agent, or plugin manifest files. It checks frontmatter schema, description triggers, body voice, and file references, returning JSONL findings with rule codes and P0–P3 severity.

What makes a good Claude Code skill description?▼

A good description states when to invoke the skill in third person, stays well under 1,024 characters, and carries a few high-value quoted trigger phrases. Descriptions starting with "Use this skill when…" or lacking concrete triggers are flagged as anti-patterns.

When does this skill review lens not apply?▼

It declines when the diff only touches skill test fixtures, generated eval artifacts, or pure documentation outside the skill body. Pure prose polish belongs to the review-legibility lens, and script correctness belongs to review-code.

Why does my Claude Code skill never trigger?▼

Common causes are a description with no concrete trigger phrases, second-person phrasing, or trigger overlap with another installed skill. This lens flags empty trigger surfaces and duplicate routing as findings.

Does a skill with side effects need special frontmatter?▼

Yes. Skills that deploy, commit, push, send, post, or delete must set disable-model-invocation: true so they only run on explicit invocation. The lens flags unguarded side effects and overly broad allowed-tools grants.