Skip to content

Repository files navigation


Master Skills

Master Skills

The unified skill library for Claude, Codex, Cursor, Antigravity & every AI agent.

2,658 skills · 15 domains · one machine-readable catalog · zero dependencies


CI Release npm License Skills Domains PRs welcome


⚡ Install  ·  🗂 Domains  ·  📦 Manifest  ·  ▶️ Usage  ·  ❓ FAQ


npx skills add sinhoneyy/master-skills


✨ What is Master Skills

One source of truth for agent skills. Instead of juggling scattered, inconsistent collections, you get a single clean catalog — deduplicated, domain-organized, and indexed.

Master Skills is a standardized catalog of 2,658 AI-agent skills, sorted into 15 normalized domains. Every skill is runtime-agnostic — it works with Claude, Codex, Cursor, Antigravity, and anything else that loads SKILL.md skills. Each one has a stable id, version, domain, and slash trigger, all described in one authoritative manifest: skills_index.json.

2,658
unique skills
15
domains
0
dependencies
1
manifest file

⚡ Install — choose your tool

One command installs the skills into the path your host expects:

Tool Install First use
Claude Code npx master-skills --claude /<skill> … (or the Claude plugin marketplace)
Cursor npx master-skills --cursor @<skill> …
Gemini CLI npx master-skills --gemini Use <skill> to …
Codex CLI npx master-skills --codex Use <skill> to …
Antigravity IDE npx master-skills --antigravity @<skill> …
Antigravity CLI (agy) npx master-skills --agy /<skill> …
Kiro npx master-skills --kiro @<skill> …
OpenCode npx master-skills --opencode opencode run @<skill> …
AdaL CLI npx master-skills --adal Use <skill> to …
Custom path npx master-skills --path ./my-skills depends on your tool

Filter what gets installed:

npx master-skills --claude --category security,backend   # only those domains
npx master-skills --cursor --risk safe,none              # only low-risk skills
npx master-skills --list                                 # show all categories
npx master-skills --path ./skills --dry-run              # preview, write nothing

Default tool paths are best-effort — if your host stores skills elsewhere, use --path.

Or clone and point your agent's loader at skills/ directly. No build step, no dependencies. ✅


🧩 Install as a plugin (Claude · Codex · Antigravity)

Master Skills ships as one plugin with a manifest for each platform, so the same catalog installs everywhere:

Claude Code

/plugin marketplace add sinhoneyy/master-skills
/plugin install master-skills@master-skills

Codex — point Codex at the repo; it reads .codex-plugin/plugin.json (skills → ./skills/).

Antigravity — add the repo as a plugin source; it reads .agents/plugins/marketplace.json.

Platform Manifest Bundles?
Claude Code .claude-plugin/marketplace.json + plugin.json ✅ all-skills + 10 themed
Antigravity .agents/plugins/marketplace.json ✅ all-skills + 10 themed
Codex .codex-plugin/plugin.json single full plugin (Codex uses one plugin; filter by category)

Themed bundles

Prefer a focused set over all 2,658? Install just the role bundle you need (Claude & Antigravity):

/plugin install master-skills-security-engineer@master-skills
/plugin install master-skills-frontend-designer@master-skills
Bundle Skills Domains
master-skills-integrations 812 integrations
master-skills-ai-agent-builder 388 agents · ai-ml · prompt-engineering
master-skills-backend-engineer 310 backend · data-db
master-skills-frontend-designer 227 web-frontend · design-ux
master-skills-devops-cloud 160 devops-cloud
master-skills-marketing-growth 123 content-marketing
master-skills-security-engineer 98 security
master-skills-qa-testing 53 testing
master-skills-productivity 52 productivity
master-skills-mobile-developer 14 mobile

The full master-skills plugin (all 2,658) remains installable on its own.


🗂 The 15 domains

Domain Skills What's inside
🔌 integrations 812 App connectors, automation, webhooks
🧰 general 421 Cross-cutting, no single domain
⚙️ backend 233 APIs, servers, frameworks, GraphQL
🧠 prompt-engineering 193 Prompting, RAG, context, evaluation
☁️ devops-cloud 160 Docker, K8s, Terraform, AWS/Azure/GCP, CI/CD
🤖 agents 154 Multi-agent, orchestration, autonomous
🎨 web-frontend 143 React/Vue/Svelte/Angular, CSS, components
📣 content-marketing 123 SEO, copywriting, social, email
🔐 security 98 Pentest, vuln analysis, auth, crypto
🖌️ design-ux 84 Figma, canvas, brand, UI/UX
🗄️ data-db 77 SQL, Postgres/Mongo/Redis, pipelines
🧪 testing 53 Unit/E2E/TDD, Playwright, pytest
📄 productivity 52 PDF/XLSX/PPTX/DOCX, office, slides
🧬 ai-ml 41 Training, embeddings, vectors, NLP
📱 mobile 14 iOS, Android, Flutter, Swift, RN

Live counts always reflect skills_index.json → categories.


📦 The manifest

A single JSON document, skills_index.json, describes the entire library.

📋 Schema — top level
Field Type Description
name string Always "master-skills"
schemaVersion string Manifest schema version
version string Library version
generated string ISO-8601 UTC build timestamp
skill_count number Total unique skills
categories object { domain: count }
skills object[] One entry per skill
🧩 Schema — per-skill entry
Field Type Description
id string Stable, slugified skill id
name string Human-readable name
version string Skill version (default 1.0.0)
description string One-line summary
category string Assigned domain
tags string[] Tags, if any
trigger string Slash trigger, /+id
path string skills/<id>
{
  "name": "master-skills",
  "schemaVersion": "1.0",
  "skill_count": 2658,
  "categories": { "integrations": 812, "general": 421, "backend": 233, "…": 0 },
  "skills": [
    {
      "id": "api-design-principles",
      "name": "API Design Principles",
      "version": "1.0.0",
      "description": "Guidelines for designing clean, consistent APIs…",
      "category": "backend",
      "trigger": "/api-design-principles",
      "path": "skills/api-design-principles"
    }
  ]
}

📁 Repository layout

master-skills/
├─ 📂 skills/                    # 2,658 skills, flat (domain lives in the manifest)
│   ├─ api-design-principles/SKILL.md
│   ├─ adversarial-reviewer/SKILL.md
│   └─ …
├─ 🧩 .claude-plugin/            # Claude Code plugin + marketplace manifest
├─ 🧩 .codex-plugin/             # Codex plugin manifest
├─ 🧩 .agents/plugins/           # Antigravity marketplace manifest
├─ 🗂️ skills_index.json          # the manifest (id, version, category, trigger, path)
├─ 📜 CREDITS.md                  # license/source notes + full upstream notices
└─ 📖 README.md

▶️ Using a skill

1.  Find it      →  search skills_index.json by id / trigger / category
2.  Open it      →  go to its "path", e.g. skills/api-design-principles/
3.  Invoke it    →  use its "trigger" (/api-design-principles) in your agent

Each skill keeps its full payload — SKILL.md, references, scripts, examples — exactly as published. Or just point your runtime's skill loader at the whole skills/ directory.


❓ FAQ

How are skills organized?
Into 15 normalized domains. A skill's domain comes from its frontmatter when it maps to a known domain, otherwise a keyword heuristic, falling back to general.
How do I find a specific skill?
Search skills_index.json by id, category, or trigger — it's a flat list designed to be grepped or loaded programmatically.
Is the catalog self-contained?
Yes. After cloning, everything you need is in skills/ and the manifest — no network access or build step required to use it.

🤝 Contributing

PRs are welcome! Good first contributions:

  • ➕ Add new skills as skills/<id>/SKILL.md (set the domain via frontmatter)
  • 🏷️ Improve domain classification so fewer skills land in general
  • 🧪 Add manifest validation or search tooling

Once GitHub Actions is approved (when required for first-time fork contributors), CI validates the manifest and lints changed SKILL.md files. Lint results appear in the Actions job summary. Comments are not posted by CI.


📜 License

The Master Skills catalog tooling is MIT-licensed. The root package.json license field covers the catalog tooling; it does not relicense third-party skills. Bundled skills retain their own upstream terms. Consult each skill's nested LICENSE/notice where present and the source attribution in CREDITS.md.

Source imports and compatibility

A new skill in the source index is not automatically an endorsement, published npm release, or proof that every agent host supports it. CI checks repository structure and, for supported imports, bundled examples plus an isolated local npm package/install smoke test. Host-specific behavior and external services require separate verification and any necessary user authorization. Fork PR lint results are always available in the GitHub Actions job summary. Lint reporting uses no PR-comment API or write token.


Built as one unified catalog. No dependencies. No lock-in.

⬆ back to top

Releases

Packages

Contributors

Languages