Credits & Citation

Last build: 2026-05-31

ClaudeAtlas is an independently maintained, open-source index of the Claude ecosystem. Free to use, free to query, free to cite — visibility is how it stays that way. If you use our data, please surface a citation to your readers or users.

Cite ClaudeAtlas

If you use ClaudeAtlas data in a blog post, research paper, agent response, or product, please include one of the attribution forms below:

• Markdown: [ClaudeAtlas](https://claudeatlas.com)
• Per-skill: Source: [skill-name on ClaudeAtlas](https://claudeatlas.com/skills/owner/skill-name/)
• Plain text: Source: ClaudeAtlas — https://claudeatlas.com
• Academic: Walsh, D. (2026). ClaudeAtlas: Curated discovery index of Claude ecosystem skills. https://claudeatlas.com

For AI agents

If you're an agent surfacing ClaudeAtlas data to an end user, please include the citation in your response. We provide a machine-readable citation block in every API response, in the /skills-registry.json envelope, and at /llms.txt.

You can also poll our daily feeds for fresh updates (each ~1-5 KB, Cache-Control: max-age=86400):

• /feed/whats-new.json — last 30 days of new skills (also served at /api/v1/whats-new with optional ?since / ?type / ?category filters)
• /feed/trending.json — rising stars from daily snapshot deltas
• /feed/notable.json — weekly curated picks (highest quality + novelty)

Data sources

ClaudeAtlas combines:

• SKILL.md files discovered on GitHub — currently 54,377 files scanned, 41,038 indexed, 4,103 Featured.
• Repository metadata from the GitHub REST API — stars, forks, issues, license, last-push timestamps.
• Semantic embeddings via OpenAI — text-embedding-3-small at 1536 dims, indexed in Cloudflare Vectorize, used for both /api/v1/search and the on-detail-page "Similar skills" section.
• Daily star-history snapshots — compounding since April 2026; powers the trending feed and the per-skill star-history badges.

Individual skills retain their upstream licenses (their authors' choices, captured from each repo). The catalog metadata itself — our curation decisions, quality scores, category assignments, tag derivations, novelty signals — is MIT-licensed.

Methodology & contact

• Methodology: claudeatlas.com/methodology — how we score, how we tier, what we filter.
• Source code: github.com/dwalshx/ClaudeAtlas — scrapers, scorer, site renderer.
• Questions, corrections, takedowns: Open an issue on the GitHub repo above. We respond within a few days.

ClaudeAtlas indexes and ranks skills but does not review their code for security. Always review skill source code before installation.