aeo— plugin

<a name="top"></a>

<div align="center">

<br />

<img src="https://img.shields.io/badge/%F0%9F%93%9A_MASTER_SKILLS-f97316?style=for-the-badge&labelColor=0a0a0f" alt="Master Skills" height="52" />

<h1>Master&nbsp;Skills</h1>

<h3>The unified skill library for <b>Claude</b>, <b>Codex</b>, <b>Cursor</b>, <b>Antigravity</b> &amp; every AI agent.</h3>

<p><i>2,658 skills · 15 domains · one machine-readable catalog · zero dependencies</i></p>

<br />

<br />

[⚡ Install](#-install) &nbsp;·&nbsp; [🗂 Domains](#-the-15-domains) &nbsp;·&nbsp; [📦 Manifest](#-the-manifest) &nbsp;·&nbsp; [▶️ Usage](#️-using-a-skill) &nbsp;·&nbsp; [❓ FAQ](#-faq)

</div>

<br />

npx skills add sinhoneyy/master-skills

<br />

✨ 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.

<table> <tr> <td align="center"><b>2,658</b><br/>_{unique skills}</td> <td align="center"><b>15</b><br/>_domains</td> <td align="center"><b>0</b><br/>_dependencies</td> <td align="center"><b>1</b><br/>_{manifest file}</td> </tr> </table>

⚡ Install — choose your tool

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

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.

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

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

🗂 The 15 domains

<div align="center">

</div>

_{Live counts always reflect skills_index.json → categories.}

📦 The manifest

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

<details> <summary><b>📋 Schema — top level</b></summary>

<br/>

</details>

<details> <summary><b>🧩 Schema — per-skill entry</b></summary>

<br/>

</details>

{
  "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

<details> <summary><b>How are skills organized?</b></summary> <br/> 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 <code>general</code>. </details>

<details> <summary><b>How do I find a specific skill?</b></summary> <br/> Search <a href="skills_index.json"><code>skills_index.json</code></a> by <code>id</code>, <code>category</code>, or <code>trigger</code> — it's a flat list designed to be grepped or loaded programmatically. </details>

<details> <summary><b>Is the catalog self-contained?</b></summary> <br/> Yes. After cloning, everything you need is in <code>skills/</code> and the manifest — no network access or build step required to use it. </details>

🤝 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.

<div align="center"> <br/> _{Built as one unified catalog. No dependencies. No lock-in.} <br/><br/>

[⬆ back to top](#top)

</div>