Skillibary
Submit
Open Registry

The skill layer for every agent

Discover, install, and publish reusable agent skills. One command to add any capability to any Claude agent.

Browse skills ↗Submit a skill
pdf-extract
Extract text and tables from any PDF document
v1.2.0↗ 8.4k
web-scrape
Structured scraping with CSS selectors + Markdown
v2.0.1↗ 6.1k
code-review
Review pull requests and surface actionable feedback
v1.0.4↗ 5.2k
Type to search >2k skills↵ to open
4+
verified skills
—
agent runs
2+
contributors
MIT
open license
Connect your agent

One config, every skill

Drop the Skillibary MCP server into your agent. It searches the hosted registry semantically, loads any skill on demand, and tracks usage automatically. No keys, no database, nothing to provision.

  1. 01
    Paste the config
    Drop the snippet on the right into your agent’s MCP settings.
  2. 02
    Restart the client
    Fully quit Claude Desktop / Cursor and reopen. Four skillibary tools appear.
  3. 03
    Ask the agent
    Try: “Find a Skillibary skill that writes git commit messages.”
Config file lives at:
  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
  • Cursor: Settings → Features → MCP

Full walkthrough with troubleshooting: docs/install →

claude_desktop_config.json
{
  "mcpServers": {
    "skillibary": {
      "command": "npx",
      "args": ["-y", "skillibary-mcp"]
    }
  }
}
After restart, try: "Find a skill that writes git commits."
Skill format

One Markdown file,
infinite capability

A skill is a plain Markdown file with structured front-matter. Any agent can read it, any team can maintain it.

  • Front-matter
    Name, description, version, tags — machine-readable metadata.
  • Parameters
    Typed inputs the agent passes when invoking the skill.
  • Returns
    What the skill produces — object shape, file, or stream.
  • Examples
    Optional prompt/response pairs that teach the agent how to use it.
SKILL.mdexample
---name: pdf-extractdescription: Extract text and tables from PDFsversion: 1.2.0tags: [documents, parsing, pdf]--- ## DescriptionExtracts structured text, tables, andmetadata from PDF files using pdfjs. ## Parameters- path (string) — file path or URL- pages (string?) — range e.g. "1-5" ## ReturnsObject with `text`, `tables`, `metadata`.
Publish a skill

Three steps to the registry

No account required to browse. GitHub login to publish. Verification is instant and automated.

01Write

Create a SKILL.md file with your name, description, parameters, and usage examples.

touch SKILL.md
02Validate

Run the CLI to check format, score quality, and catch issues before submitting.

npx skillibary validate
03Publish

Submit for review. Claude reads it, scores it, and publishes verified skills instantly.

npx skillibary publish
Submit a skill →Browse existing skills