claude-skills
CLAUDE.md62 starsChanged 9 months agoMITView on GitHub
This is an Agent Skills repository following the open standard that makes skills compatible with Claude Code, OpenAI Codex, ChatGPT, and other AI coding assistants. It contains UX/UI evaluation skills for auditing interfaces using industry-standard methodologies. Validate structure (descriptions ≤200 chars, frontmatter, naming, SKILL.md size) with the bundled script: To exercise a real install of your local working copy (including uncommitted branch changes), install into a scratch target outside the repo, passing the repo path as the source: ⚠️ Do…
In this repository14 other files it gives its agents.
- ai-assessment-scaleSkill
- cognitive-walkthroughSkill
- don-norman-principles-auditSkill
- gdpr-auditSkill
- information-architecture-auditSkill
- iso-27090-ai-securitySkill
- iso-42001-ai-governanceSkill
- nielsen-heuristics-auditSkill
- nist-ai-rmfSkill
- owasp-ai-testingSkill
- owasp-llm-top10Skill
- ui-design-reviewSkill
- ux-audit-rethinkSkill
- wcag-accessibility-auditSkill
The file
Show its rules and text
We split this CLAUDE.md into the rules it gives an agent
13 rules. About 0 are common, 8 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 32
After installing, invoke the skill through natural language (the AI agent will automatically detect when to use it based on the YAML …
- Very specific: line 49
Every SKILL.md must follow this exact structure:
- Very specific: line 75
Use kebab-case for skill names (e.g., wcag-accessibility-audit)
- Distinctive: line 86
Update README.md: Add skill to the skills table
- Very specific: line 88
Update CHANGELOG.md: Document the addition under [Unreleased]
- Very specific: line 122
Scope: Use skill name or affected area (e.g., wcag-audit, contributing)
- Distinctive: line 126
No capitalization on first letter
- Distinctive: line 127
No period at end
- Distinctive: line 165
Before committing a new skill, verify:
- Distinctive: line 178
When modifying skills or documentation:
- Distinctive: line 180
README.md: Update skills table, installation examples, or feature descriptions
- Very specific: line 181
CHANGELOG.md: Add entry under [Unreleased] following Keep a Changelog format
- Distinctive: line 192
Write skills to be agent-agnostic.
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Repository Purpose
This is an Agent Skills repository following the open standard that makes skills compatible with Claude Code, OpenAI Codex, ChatGPT, and other AI coding assistants. It contains UX/UI evaluation skills for auditing interfaces using industry-standard methodologies.
Testing Skills Locally
Validate structure (descriptions ≤200 chars, frontmatter, naming, SKILL.md size) with the bundled script:
npm run validate # or: node scripts/validate.jsTo exercise a real install of your local working copy (including uncommitted branch changes), install into a scratch target outside the repo, passing the repo path as the source:
mkdir -p /tmp/skills-test && cd /tmp/skills-test
npx skills add /absolute/path/to/this/repo --skill skill-name
# installed under /tmp/skills-test/.agents/skills/<skill-name>/ (SKILL.md + references/)⚠️ Do not run
npx skills add .from the repository root. The CLI treats the cwd as the install target, so it creates.agents/inside the repo, replacesskills/<name>/with a symlink into it, and writesskills-lock.json— clobbering the source. Keeping the target in/tmp(cwd ≠ repo) leaves the repo untouched; the repo is only read as the source.
After installing, invoke the skill through natural language (the AI agent will automatically detect when to use it based on the YAML description).
Repository Architecture
Core Structure
skills/ # All skills live here
├── skill-name/ # Each skill is a directory
│ ├── SKILL.md # REQUIRED: YAML frontmatter + markdown instructions
│ ├── scripts/ # OPTIONAL: Executable helpers (.sh, .py, .js)
│ ├── references/ # OPTIONAL: Reference documentation
│ └── assets/ # OPTIONAL: Templates, examplesSKILL.md Anatomy
Every SKILL.md must follow this exact structure:
---
name: skill-name # kebab-case, matches directory name
description: Brief description... # MAX 200 chars - AI uses this to determine invocation
---
# Skill Title
## When to Use This Skill
[Scenarios for invocation]
## Inputs Required
[Required and optional parameters]
## Procedure
[Step-by-step instructions the AI agent executes]
## Output Format
[Expected output structure]Critical constraints:
descriptionfield MUST be ≤200 characters (AI uses this for skill matching)- YAML frontmatter must parse correctly
- Use kebab-case for skill names (e.g.,
wcag-accessibility-audit)
Adding New Skills
- Create directory structure:
mkdir -p skills/your-skill-name- Create SKILL.md with YAML frontmatter + instructions
- Update README.md: Add skill to the skills table
- Update CHANGELOG.md: Document the addition under
[Unreleased] - Validate locally:
npm run validate(For a real install test, run npx skills add from outside the repo — see "Testing Skills Locally".)
- Commit using Conventional Commits:
git commit -m "feat(your-skill-name): add skill for X evaluation"Commit Convention (REQUIRED)
This repository uses Conventional Commits specification:
<type>(<scope>): <subject>
<body>
<footer>Types:
feat: New skill or feature (feat(wcag-audit): add contrast checker)fix: Bug fix (fix(nielsen): correct severity rating)docs: Documentation (docs(readme): update examples)style: Formatting only (style(skill): fix markdown)refactor: Code restructuring (refactor(wcag): simplify structure)test: Tests (test(skills): add validation)chore: Maintenance (chore: update dependencies)
Scope: Use skill name or affected area (e.g., wcag-audit, contributing)
Subject rules:
- Imperative mood: "add" not "added" or "adds"
- No capitalization on first letter
- No period at end
- ≤50 characters
Examples:
# Good
feat(cognitive-walkthrough): add task analysis skill
fix(wcag-audit): correct WCAG 2.2 criteria references
docs(contributing): clarify testing process
# Bad
feat: new skill # Too vague, no scope
fix: fixed bugs # Not imperative, vague
feat(wcag): Added comprehensive audit # Capitalized, not imperativeCurrent Skills
UX/UI Evaluation Suite:
ux-audit-rethink- Holistic UX audit (IxDF): 7 factors + 5 usability characteristics + 5 interaction dimensions + redesign proposalsdon-norman-principles-audit- 7 principles: discoverability, affordances, signifiers, feedback, mapping, constraints, conceptual modelsnielsen-heuristics-audit- 10 usability heuristics with severity ratings (0-4)wcag-accessibility-audit- WCAG 2.1/2.2 compliance (A/AA/AAA levels, 4 POUR principles)cognitive-walkthrough- Task-specific deep-dive using 4 cognitive questions to evaluate learnabilityui-design-review- Visual design evaluation: typography, color, spacing, hierarchy, consistency, branding (10 dimensions)
AI Governance & Security:
7. iso-42001-ai-governance - AI governance (ISO 42001:2023): Risk management, ethics, security, transparency, regulatory compliance (EU AI Act, GDPR)
8. nist-ai-rmf - AI risk assessment (NIST AI RMF 1.0): Govern, Map, Measure, Manage functions for trustworthy AI
9. owasp-llm-top10 - Security audit for LLM/GenAI apps (OWASP Top 10 for LLM Apps 2025)
10. owasp-ai-testing - AI trustworthiness testing (OWASP AI Testing Guide v1): 32 test cases across 4 layers
11. gdpr-audit - Technical GDPR audit of code, plans, schemas, or IaC: processing map + article-cited findings (not legal advice)
12. ai-assessment-scale - Measure AI contribution to a project using the AI Assessment Scale (AIAS) 5-level framework
13. information-architecture-audit - Information architecture (Rosenfeld/Morville four systems, 8 principles): organization, labeling, navigation, search, card sorting, tree testing
14. iso-27090-ai-security - AI security assessment (ISO/IEC 27090:2026): Threat-based gap assessment across 7 domains (governance, supply chain/AIBOM, data poisoning, model protection, inputs/outputs, monitoring) with maturity scale and client-ready report
Skill Quality Standards
Before committing a new skill, verify:
- YAML description ≤200 characters
- Clear "When to Use This Skill" section
- Step-by-step procedure with actionable instructions
- Defined output format
- No spelling/grammar errors
- Validated locally with
npm run validate(do not runnpx skills add .from the repo root) - Added to README.md skills table
- Added to CHANGELOG.md under
[Unreleased] - Uses kebab-case naming
Documentation Updates
When modifying skills or documentation:
README.md: Update skills table, installation examples, or feature descriptions
CHANGELOG.md: Add entry under [Unreleased] following Keep a Changelog format
CONTRIBUTING.md: Reference for contribution guidelines, testing, and PR process
Multi-Agent Compatibility
Skills in this repository follow the Agent Skills open standard (December 2025) adopted by:
- Anthropic Claude Code
- OpenAI Codex CLI
- ChatGPT
- Any agent supporting the specification
Write skills to be agent-agnostic. Avoid references to specific AI agents in skill instructions.
Installation Distribution
Users install skills via:
# From GitHub (primary)
npx skills add mastepanoski/claude-skills --skill skill-name
# From npm (if published)
npx skills add @mastepanoski/claude-skills --skill skill-nameThe files field in package.json ensures only skills/ directory is distributed.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

