Skill not invoked — Claude gives generic review

Say 'use SkillCheck-Free' explicitly.

freemcplab / 디렉터리 / 플레이그라운드 / SkillCheck-Free

● 커뮤니티 olgasafonova ⚡ 바로 사용

SkillCheck-Free

Name: SkillCheck-Free (Claude Skill)
Author: olgasafonova

제작: olgasafonova · olgasafonova/SkillCheck-Free

Validates Claude Code skills against the open agentskills spec — structure, body, naming, semantics, quality patterns — Critical / Warning / Suggestion.

olgasafonova/SkillCheck-Free is a spec-compliance linter for Agent Skills. Free tier covers: 1.x Structure (YAML frontmatter, name validity, description quality, XML injection), 2.x Body (length, hardcoded dates, empty sections, unqualified MCP refs), 3.x Naming (generic names, verb-noun), 4.x Semantic (conflicting instructions, vague terms, unspecified outputs), 8.x Quality patterns (examples, error handling, triggers), 19.x Design pattern classification, 22.x Knowledge density. Severity tiers: Critical / Warning / Suggestion / Strength.

▶ 데모 보기 📦 설치 💡 사용 사례

왜 쓰나요

핵심 기능

Checks YAML frontmatter, name validity, XML-injection risks
Body audits — length, hardcoded dates, empty sections, unqualified MCP refs
Semantic checks for conflicting or vague instructions
Design-pattern classification (Reviewer / Generator / Inversion / Pipeline / Tool Wrapper)
Severity-tiered output — lets you triage Critical vs Suggestion

라이브 데모

실제 사용 모습

skillcheck-free-skill.replay ▶ 준비됨

0/0

설치

작성자 저장소 · Claude Skills 공식 문서

사용 사례

실전 활용법: SkillCheck-Free

Validate your skill before publishing it to a catalog

👤 Skill authors ⏱ ~20 min beginner

언제 쓸까: Before opening a PR to awesome-claude-skills or anthropics/skills.

흐름

Run SkillCheck on your SKILL.md

Use SkillCheck-Free — validate ~/my-skill/SKILL.md against the agentskills spec.✓ 복사됨

→ Per-rule findings with severities
Fix Critical issues

Walk me through each Critical finding and propose a fix.✓ 복사됨

→ Concrete edits, not 'improve clarity'
Re-run until clean

Re-validate. Anything still Critical or Warning?✓ 복사됨

→ Progress toward zero Critical

결과: A skill that won't be rejected for trivial spec violations.

함정

Chasing every Suggestion — Suggestions are aspirational; Critical + Warning is the bar for most catalogs

함께 쓰기: skills-skill-8 · awesome-claude-skill-2

Audit every skill you've installed for quality concerns

👤 Skill hoarders who want to prune ⏱ ~30 min intermediate

언제 쓸까: You installed 20 skills and aren't sure which ones are well-made.

흐름

Decide what to keep

Which skills have persistent Critical issues I should uninstall?✓ 복사됨

→ Prioritized uninstall candidates

결과: A leaner, higher-quality skill collection.

조합

다른 MCP와 조합해 10배 효율

skillcheck-free-skill + skills-skill-8

Cross-check: SkillCheck validates spec, huangwb8/auto-test-skill tests robustness

Run SkillCheck first for spec issues, then auto-test-skill for design weaknesses.✓ 복사됨

skillcheck-free-skill + awesome-claude-skill-2

Validate before submitting your skill to the catalog

SkillCheck my SKILL.md, then if clean, draft the awesome-claude-skills PR.✓ 복사됨

도구

이 MCP가 노출하는 것

도구	입력	언제 호출
validate_skill	path to SKILL.md	Pre-publish or post-install
classify_pattern	skill body	Understanding how a skill fits your collection
batch_validate	directory of skills	Fleet-wide audit

비용 및 제한

운영 비용

API 쿼터: None
호출당 토큰: Small — typical SKILL.md is <10k tokens
금액: Free tier only; Pro tier exists for deeper pattern validation
팁: Free tier is sufficient for most authoring — upgrade only if you publish commercially.

보안

권한, 시크릿, 파급범위

자격 증명 저장: No credentials — pure prompts

데이터 외부 송신: None — validation is local

Passing SkillCheck ≠ the skill is actually useful; it's a spec check, not a utility check.

문제 해결

자주 발생하는 오류와 해결

SKILL.md frontmatter wrong in target skill

Expected — that's the finding SkillCheck exists to surface. Fix the frontmatter and re-run.

확인: head <target>/SKILL.md

Wrong trigger keywords — SkillCheck runs but misses severity tier

Ask explicitly for 'severity-tiered output'; default may compress.

대안

SkillCheck-Free 다른 것과 비교

대안	언제 쓰나	단점/장점
skills-skill-8 auto-test-skill	You want design robustness testing, not spec compliance	Complementary — run both
Manual review	You've written many skills and have internalized the spec	No machine backstop for subtle issues