Claude Skill Reviewer
Deep-review one SKILL.md before it ships: will it load and trigger, and does it beat plain Claude? Returns fixes and a new description.
Before & after
Same prompt, same agent — the only difference is the skill.
That's a good question to think about. Skills can be useful depending on your needs and workflow. You might want to consider whether the skill solves a real problem you have. Think about how often you'd use it and whether it integrates well with your existing tools. Some skills are more general while others are specialized. It really depends on your specific situation and priorities. I'd recommend looking at examples of what it produces to get a sense of whether it's a good fit for you.
SKILLS REVIEW FRAMEWORK CRITERIUM 1: Solves my actual problem? Example: 'CTO Review' addresses scaling/tech debt questions you ask monthly → Score: Yes Example: 'Changelog' automates release notes you write quarterly → Score: Maybe (low frequency) CRITERIUM 2: Chains with my stack? Does it assume tooling I don't use? Does it output formats I import elsewhere? Example: You use GitHub + Jira → Changelog skill must handle GitHub Actions (it does) CRITERIUM 3: Cost vs payback? Token cost per use × frequency per month = opportunity cost Break-even: payback months = (install overhead) / (time saved per use × hourly rate) Example: ASO audit costs 8k tokens, saves 2 hours research/mo, 125/hr loaded = 4 weeks payback CRITERIUM 4: Customization friction? How many questions to adapt to your brand/stack? High friction: 15+ questions, new SKILL.md needed Low friction: 3-5 questions, uses defaults CRITERIUM 5: Lock-in risk? Does output lock you into the tool, or export cleanly? DECISION: Install if 4 of 5 criteria score Yes.
About this skill
Most skills fail quietly: they never trigger, or they trigger and add nothing the model would not do anyway. This reviewer checks both, with the real SKILL.md conventions, and hands back an exact fix list and a rewritten description.
What it does
- Checks frontmatter, name rules, name-to-folder match, file location, and YAML pitfalls
- Diagnoses the description as the trigger and supplies a rewrite with realistic user phrasings
- Tests whether each section adds domain knowledge or just restates generic process
- Flags rigidity, missing honesty rules, unsafe instructions, and over-broad tool access
- Marks product behavior it is unsure of as [verify] instead of guessing
What you get
A verdict (ship, fix first, rethink) with load and trigger fixes, a description rewrite, value vs baseline, rigidity and honesty findings, and file and script issues.
Best for
Skill authors and teams publishing skills to a project, plugin, or marketplace.
SKILL.md preview
---
name: skill-review
description: Deep-review one Claude / Claude Code skill (a SKILL.md plus its folder) before it ships - frontmatter validity, name and folder match, whether the description will actually trigger, progressive disclosure, whether the body adds knowledge the model lacks or just restates generic process, rigidity, honesty rules, and bundled scripts - and return a verdict with rewritten description and exact fixes. Use when the user says "review my skill", "is this SKILL.md good?", "why doesn't my skill trigger?", "will this skill actually help?", "tighten this skill description", "critique this skill before I publish it", or pastes a SKILL.md and asks what's wrong.
---
… (sign up to view the full skill)More ai & agents skills
View all AI & Agents skills →New Skill Scaffolder
Create a new Claude Code skill correctly: right mechanism, valid name and folder, a triggering description, and a smoke test.
Skill Library Lint and Overlap Report
Lint every skill in a repo at once: frontmatter, name-folder match, weak descriptions, overlapping triggers, broken refs, plus a CI script.
Self-Verifying Agent Loop: Plan, Check, Retry, Close
Turns a goal into a task plan with machine-run checks, executes with retry caps, escalates, and refuses to close until verified.
Skill Test Runner (Trigger and A/B)
Test a skill in fresh sessions: ready-to-run prompts, commands, and rubric for trigger recall, false triggers, and blind A/B.