Overview

The AI Skills Library is a curated catalog of building blocks that extend the capabilities of AI agents. Browse, search, and install them to enhance your workflows — from writing and reviewing code to running operational and business processes.

The catalog now spans four kinds of entries. Skills are the core unit; plugins bundle several units together; MCP servers connect agents to live tools and data; and marketplaces group entries into installable sources. Every entry carries a trust level and can be installed from its detail page.

The library can be viewed in two modes, switchable at the foot of the page: a Human view (the browsable catalog described here) and an Agent view intended to be read by an AI agent directly.

What's in the catalog

Use the type filter at the top of the library to browse each kind of entry on its own, or view everything together.

Understanding plugins

AI coding agents are extensible through skills, hooks, and tool servers — but each is normally configured separately, with no standard way to package them together. A plugin solves that: it is a single installable unit that bundles related extensions so they can be shared, versioned, and updated as one.

In this library, a plugin most commonly bundles one or more skills together with the MCP servers those skills rely on. A plugin can hold several of each — or just a single skill. There is no minimum; the bundle is whatever the author needs to ship together.

On an MCP server's own page you can see this relationship in reverse: it lists which plugin declares it and which skills reference it — so you can trace a single tool server back to everything that uses it.

The Open Plugins standard

Plugins in this library follow Open Plugins — an open standard for packaging agent extensions into distributable plugins that any conformant tool can install and run. Under the standard, a plugin is simply a directory with a manifest and components in known locations:

my-plugin/
├── .plugin/
│   └── plugin.json       # Manifest: name, version, metadata
├── skills/               # Agent Skills (SKILL.md)
├── .mcp.json             # MCP tool servers
├── agents/               # Specialized sub-agents
├── hooks/                # Event-driven automation
└── rules/                # Coding standards

Our catalog currently surfaces the skills and MCP servers a plugin contains. The standard defines further component types — sub-agents, hooks, rules, and LSP servers — which a plugin may also carry. For the full component reference and specification, see open-plugins.com.

How a plugin installs

When you install a plugin, the host tool scans its directory, discovers the components in their default locations, and registers them. Each component is namespaced under the plugin name to prevent conflicts, then activated — skills become available, and any MCP servers start running.

Finding entries

Press / or click the search bar to find entries by name, description, tag, keyword, or author. Three filter groups then narrow the catalog:

Each entry also carries a more specific category tag on its card — for example #Platform #DevOps #Code Review — which rolls up under the Platform / Business classification.

Switch between grid and list view using the toggle in the toolbar; the result count updates live as you filter. To publish your own entry, use Submit Skills in the header — see Contributing.

Detail pages

Click any entry to open its detail page. Beyond a description and installation options, each type surfaces the metadata and relationships relevant to it, so you can judge and trace it before installing.

Installing

Every detail page shows the installation options available for that entry. Which options appear depends on the entry type and your environment.

Before you install

Each method has its own prerequisites. Check you have what an option needs before using it:

npx skills add SAP/leanix-ai-plugins --skill automations-toolkit

/plugin marketplace add SAP/leanix-ai-plugins
/plugin install sap-leanix@leanix-ai-plugins

{
  "mcpServers": {
    "leanix": {
      "type": "http",
      "url": "https://mcp.leanix.net/services/mcp-server/v1/mcp?toolsets=…"
    }
  }
}

{
  "mcpServers": {
    "ui5": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@ui5/mcp-server"]
    }
  }
}

Trust & checks

Every entry carries a trust level so you can evaluate its origin at a glance:

Contributing

The library uses a bring-your-own-repo model — your code stays in your own public GitHub repository. You don't submit code here; instead you register your repository, and a maintainer onboards it so it appears on the site.

1. Structure your repository

   Place each skill under a skills/<skill-slug>/ directory with a SKILL.md file:

   Repository layout

   your-repo/
   └── skills/
       └── <skill-slug>/
           └── SKILL.md

   The SKILL.md must include name and description frontmatter. Author and license information should be discoverable somewhere in the repo — a README, LICENSE file, or package.json. Plugins follow the Open Plugins directory layout described above.

2. Register your entry

   Once the repository is public on github.com, open a Register a New Skill issue. A maintainer reviews the submission and onboards the repository to the library.

For improvements to existing entries or documentation fixes, see the Contribution Guidelines.