Skip to content

Ship better skills.
Catch issues before your users do.

SkillCheck validates your Agent Skills, Claude Code plugin manifests, and MCP server tool lists against the open standard. Pro adds accessibility, security, anti-slop, and marketplace-governance checks. Built for Claude Code.

Check My Skill View on GitHub

Built for Claude Code · Follows the Agent Skills open standard · v3.20.0

Validates skills, plugins, and MCP servers anywhere they're used. Pro connects as an MCP tool server.

v3.20: skills, plugins, and MCP servers

v3.20 widens scope from SKILL.md only to three artifact types. Pro binaries are available for macOS (arm64, amd64), Linux (amd64, arm64), and Windows (amd64).

Cat 24: Plugin Manifests

  • + check-plugin validates .claude-plugin/plugin.json
  • + Free: schema validation against the Anthropic reference
  • + Pro: governance recommendations (maintainers, evals, deprecation)

Cat 23: MCP Servers

  • + check-mcp validates a server's tools/list response
  • + Free: tool-count thresholds and 1:1 CRUD-API smell
  • + Pro: identity-override, annotations, CLI literacy, code-orchestration

Benchmark proof

  • + Cat 24 Pro rerun against all 21 plugins in anthropics/knowledge-work-plugins
  • + Schema-floor lock holds: zero criticals across the full reference
  • + Empirical finding: 0 of 21 declare maintainers or evals

Also new

  • + WASM validator handles all three input types in the browser
  • + Cat 25 (Memory Governance) design spec published; implementation tracking Anthropic memory exemplars

Full changelog · Free tier on GitHub

You built a skill. But is it good?

Agent Skills are now an open standard adopted by Microsoft, Cursor, and dozens of coding agents. Your skill needs to work everywhere.

You won't know until someone complains. Or you run SkillCheck.

Tested at scale

We scan public skill repositories regularly. As SkillCheck adds checks, it catches more; scores reflect a stricter bar, not worse skills.

1,950

skills scanned

58

average score out of 100

72%

missing proper descriptions

3.6%

scored Excellent

How the numbers moved

v3.12 (82 checks)

1,613 skills, avg 65

v3.19 (115 checks)

1,950 skills, avg 58

New categories

+6 (patterns, density, evals)

Repos scanned

8 → 12

Last scanned: April 2026 with SkillCheck v3.18.0. 12 repos including:

microsoft/skills anthropics/skills ComposioHQ/awesome-claude-skills Jeffallan/claude-skills alirezarezvani/claude-skills davepoon/buildwithclaude coleam00/second-brain-skills

What It Checks

Category Tier What We Catch
Structure Free Missing fields, invalid names, broken YAML, XML injection, effort/maxTurns/disallowedTools validation
Body Free Content requirements, length, formatting, anti-pattern format lint, MCP tool qualification
Naming Free Conventions, specificity, reserved words, gerund naming
Semantics Free Contradictions, ambiguous instructions, wisdom/platitude detection, workflow-steps-in-description, description trigger style, railroading detection
Quality Patterns Free Examples, error handling, triggers, output format, structured instructions, prerequisites, gotchas sections
Anti-Slop Pro "Let's dive in", hedge words, filler phrases
Visual & WCAG Pro Color contrast, accessibility, visual design checks for skills that produce UI
Security Pro PII detection, credential safety, path traversal, Bash guardrail validation
Token Budget Pro Context efficiency, budget analysis, frontmatter overhead
Enterprise Pro Hardcoded paths, env config, audit support, metadata validation, setup/config handling, data storage path safety
Workflow Pro Exit criteria, unbounded iteration, deploy-without-verify, feedback loops. Based on the Trail of Bits anti-pattern catalog.
Agent Readiness Pro 28 checks across 6 pillars: reference integrity, eval readiness, orchestration safety, autonomy design, composability, observability. Maturity scoring from L0 (manual) to L3 (autonomous-ready).
Design Pattern Free + Pro Classifies skills into Reviewer, Generator, Inversion, Pipeline, or Tool Wrapper patterns (Google ADK taxonomy). Pro validates pattern-specific requirements.
Trigger Collision Pro Cross-skill directory scan detecting overlapping trigger phrases (Jaccard similarity) and artifact pipeline gaps (orphan producers, missing consumers).
Eval Kit Pro Auto-generates test prompts (should-trigger and should-NOT-trigger) for manual skill verification. Negative controls sourced from trigger collision data.
Knowledge Density Free + Pro Flags hollow gotcha/troubleshooting sections (Free). Pro rewards specific thresholds, consequence explanations, experience markers, debugging sequences, decision logic, and concrete code references.
Skill Discovery Free + Pro Scans your config files (CLAUDE.md, .cursorrules, .windsurfrules, AGENTS.md, hooks) for procedural knowledge that should be a skill. Free finds candidates. Pro generates the SKILL.md files.
Plugin Manifests (Cat 24, v3.20) Free + Pro Validates .claude-plugin/plugin.json for Claude Code marketplaces. Free: schema (kebab-case names, semver, required fields, layout). Pro: marketplace-governance recommendations (maintainers, change-gate evals, deprecation format). Calibrated against anthropics/knowledge-work-plugins as the schema floor.
MCP Tool Servers (Cat 23, v3.20) Free + Pro Validates an MCP server's tools/list response for Agent Integration Readiness. Free: tool-count thresholds, 1:1 CRUD-API mirror smell. Pro: identity-override resistance, annotation completeness, CLI literacy in descriptions, code-orchestration pattern detection. Floor-lock against modelcontextprotocol reference servers.

How It Works

Free

Claude reads a skill file and runs checks inline. No install, no binary.

1 Copy skill-check folder to ~/.claude/skills/
2 Say "skillcheck my skill" in Claude Code
3 Get instant feedback

Pro

A standalone binary that runs locally and connects to Claude Code as a tool server.

1 Download binary from purchase email
2 Add to ~/.claude.json with license key
3 All Pro checks unlocked automatically

CLI commands (v3.20)

SKILL.md

skillcheck check \
  ~/.claude/skills/my-skill

Plugin manifest

skillcheck check-plugin \
  --pro \
  .claude-plugin/plugin.json

MCP server

skillcheck check-mcp \
  --pro \
  tools-list.json

--pro flags require SKILLCHECK_LICENSE. Or paste any of the three input types into the browser validator; runs entirely client-side via WebAssembly. 

## my-awesome-skill Check Results [PRO]

### Critical Issues (1)
- Description missing WHEN clause. Add trigger context.

### Warnings (2)
- Line 47: Vague term "several"; specify a number
- No error handling documented

### Passed Checks: 77 / 82 applicable

### Pro Scores
✓ Anti-slop: 92/100
✓ WCAG AA: Pass
✓ Enterprise ready: Yes
✓ Agent Readiness: L2 Orchestratable (78/100)

Status: Needs Attention

Validated Skills

Browse all

Skills validated by SkillCheck get a public report page with scores, badges, and shareable links. Here are some top-scoring examples.

Reports coming soon.

Run skillcheck report your-skill/SKILL.md to generate your first report.

Pricing

Free

$0 forever

  • ✓ Structure validation
  • ✓ Body & naming checks
  • ✓ Semantic consistency
  • ✓ Quality patterns (strengths)
  • ✓ No install needed
  • ✓ Open source (MIT)
Get Started

Pro

$79 lifetime

  • ✓ Everything in Free
  • ✓ Anti-slop detection
  • ✓ Security scanning
  • ✓ Token budget analysis
  • ✓ WCAG accessibility
  • ✓ Enterprise readiness
  • ✓ Agent Readiness (L0-L3)
  • ✓ Plugin manifests (Cat 24)
  • ✓ MCP server tool lists (Cat 23)
  • ✓ Go binary & MCP server
Get Pro
COMING SOON

Team

$49 /month

up to 10 users

  • ✓ Everything in Pro
  • ✓ Custom rulesets
  • ✓ Team dashboard
  • ✓ CI/CD integration

Changelog

v3.20.0 ·April 2026

  • New (Free + Pro): check-plugin for .claude-plugin/plugin.json validation (Cat 24)
  • New (Free + Pro): check-mcp for MCP server tools/list validation (Cat 23)
  • Pro Cat 23: identity-override critical, code-orchestration strength, annotation completeness, CLI-literacy detection
  • Pro Cat 24: governance recommendations (maintainers, change-gate evals, deprecation format)
  • Schema-floor lock empirically confirmed: 21/21 plugins in anthropics/knowledge-work-plugins pass clean
  • Browser validator: three-tab input (SKILL.md, plugin.json, tools/list)
  • Cat 25 Memory Governance design spec published; implementation tracking exemplar samples

v3.19.0 ·April 2026

  • New (Pro): 3 anti-slop patterns; verb inflation, promotional adjectives, vague attributions
  • 115 total checks across 22 categories

v3.18.0 ·April 2026

  • New (Free + Pro): Knowledge Density checks (22.1-22.7); hollow content detection (Free), density strengths (Pro)
  • Based on anti-distill classifier inversion: six categories of high-value tacit knowledge
  • 112 total checks across 22 categories

v3.17.0 ·April 2026

  • New (Pro): Eval Kit generation; auto-generated test prompts (explicit, implicit, contextual, negative controls) per validated skill
  • Negative controls sourced from trigger collision data; names which skill each negative belongs to
  • 109 total checks across 21 categories

v3.16.0 ·April 2026

  • New (Pro): GitHub Actions integration; validate skills in CI pipelines
  • Configurable fail thresholds (critical, warning, none) and structured outputs

v3.15.0 ·March 2026

  • New (Free): Design Pattern Classification; detects Reviewer, Generator, Inversion, Pipeline, Tool Wrapper patterns
  • New (Pro): Pattern-specific deep checks; validates criteria, output specs, question frameworks, stage docs per pattern
  • New (Pro): Artifact Contract Validation; checks produces/consumes frontmatter with metadata and discovery docs
  • New (Pro): Trigger Collision Detection; cross-skill directory scan for overlapping triggers and artifact pipeline gaps
  • 105 total checks across 20 categories

v3.14.0 ·March 2026

  • New (Free): description trigger style check (summary vs trigger condition)
  • New (Free): railroading detection (over-prescriptive language)
  • New (Free): gotchas section recognized as quality strength
  • New (Pro): setup/config handling check (missing-config fallback)
  • New (Pro): data storage path safety (skill-dir persistence warning)
  • 87 total checks across 18 categories

v3.13.0 ·March 2026

  • New: effort field validation (low/medium/high)
  • New: maxTurns field validation with runaway detection
  • New: disallowedTools validation with overlap detection
  • Updated: model field now accepts aliases (opus, sonnet, haiku)
  • Spec version updated to 2026-03-20

v3.12.0 ·March 2026

  • New: skillcheck discover command for multi-platform config scanning
  • Scans Claude Code, Cursor, Windsurf, Codex CLI configurations
  • Pro: auto-generates SKILL.md files from discovered candidates
  • New: skillcheck_discover MCP tool

v3.11.0 ·March 2026

  • 18 check categories across free and pro tiers
  • 82 total checks (now 115 in v3.19.0)

v3.10.0 ·March 2026

  • New: Agent Readiness validation (L0-L3 maturity scoring)
  • Orchestration safety, composability, observability, autonomy design checks

Full history on GitHub

FAQ

Is this official Anthropic tooling?

No. SkillCheck is an independent project. The Free tier validates against the Agent Skills open standard (created by Anthropic). Pro adds extra quality checks like anti-slop detection, WCAG accessibility, and security scanning.

What platforms does it work with?

Built and tested for Claude Code. Since it follows the Agent Skills open standard and MCP protocol, it should work with other compatible tools like Cursor, VS Code, and Windsurf.

What about MCP servers and plugin manifests?

As of v3.20, SkillCheck validates three artifact types. check for SKILL.md (the original product). check-plugin for .claude-plugin/plugin.json (Cat 24, Skill Marketplace Governance). check-mcp for an MCP server's tools/list response (Cat 23, Agent Integration Readiness). Free covers schema and surface-level signals; Pro adds the governance and deeper-quality rubric.

Can I use it in CI/CD?

Pro includes a GitHub Actions integration. Add uses: olgasafonova/skillcheck@v3 to your workflow to block PRs that introduce low-quality skills.

What if I disagree with a check?

SkillCheck reports issues; you decide what to fix. Some checks are suggestions, not requirements.

Ready to ship better skills?

Check My Skill View on GitHub