Want to list your augment (skill or MCP server) in the catalog? Open a pull request following the steps below. Contributions to this project are released to the public under the project's open source license.
catalog/
└── <publisher>/
└── <augment-name>.json
Each augment is a single JSON file located under your publisher directory.
gh repo fork github/agentfinder-catalog --clone
cd agentfinder-catalogYour publisher name should match your GitHub organization or username.
mkdir -p catalog/<your-publisher>Create a file at catalog/<your-publisher>/<augment-name>.json with the following schema:
{
"identifier": "urn:ai:github.com:<publisher>:<repo>:<augment-name>",
"displayName": "Human-Readable Name",
"mediaType": "application/ai-skill",
"url": "https://github.com/<publisher>/<repo>/blob/main/path/to/SKILL.md",
"description": "A short description of what this augment does.",
"tags": [
"optional-tag"
],
"metadata": {
"sourceSet": "<repo>",
"repoPath": "path/to/SKILL.md"
}
}| Field | Required | Description |
|---|---|---|
identifier |
✅ | A URN uniquely identifying the augment. Format: urn:ai:github.com:<publisher>:<repo>:<name> |
displayName |
✅ | A human-readable display name |
mediaType |
✅ | The type of augment (e.g., application/ai-skill) |
url |
✅ | URL to the augment's definition file (e.g., SKILL.md) |
description |
✅ | A brief description of the augment's purpose |
tags |
❌ | Tags used to categorize and filter the augment |
metadata.sourceSet |
✅ | The source repository name |
metadata.repoPath |
✅ | Path to the definition file within the repository |
Catalog entries for plugins whose functionality consists entirely of a GitHub Copilot
canvas keep the application/vnd.github.copilot-plugin media type and must include all
of these tags:
"tags": [
"canvas",
"canvas-only",
"github-copilot"
]Do not use canvas-only for a plugin that still provides useful agents, skills, hooks,
or MCP servers when its canvas is unavailable. Canvas-only entries must not include
the corresponding agent, skill, hook, or mcp-server capability tags.
Make sure your file is valid JSON, then regenerate and check the ARD ingestion catalog:
python -m json.tool catalog/<your-publisher>/<augment-name>.json
python3 scripts/generate_ai_catalog.py
python3 scripts/generate_ai_catalog.py --checkFiles under catalog/<publisher>/ remain the source of truth for contributor-managed entries. The root ai-catalog.json is generated for ARD ingestion, supplemented with missing entries from GitHub's public MCP catalog, and should not be edited by hand.
The pull request workflow regenerates ai-catalog.json automatically for branches in this repository. Fork workflows validate the catalog but cannot push changes; the main-branch fallback opens a follow-up pull request if regeneration is needed after merge. Hourly scheduled runs and manual runs on main also open or update that same refresh pull request for GitHub MCP Registry changes, even when no contributor files have changed. Manual runs on other branches regenerate and validate only, so the refresh branch's validation does not recursively publish another pull request.
git checkout -b add-<augment-name>
git add catalog/ ai-catalog.json
git commit -m "Add <augment-name> to catalog"
git push origin add-<augment-name>
gh pr create --title "Add <augment-name>" --body "Adds <augment-name> to the agentfinder catalog."- One file per augment — don't bundle multiple augments into a single file.
- Use kebab-case for file names (e.g.,
my-cool-skill.json). - Keep descriptions concise — one or two sentences.
- Ensure your URL is publicly accessible so reviewers can verify the augment definition.
Open an issue if you have questions or need help getting your augment listed.