ClaudSkills — Open Catalog (auto-generated)

Machine-readable mirror of the ClaudSkills catalog. Daily auto-export from claudskills.com — the open registry of Claude Code / Claude Skills SKILL.md files.

What's here

File	Format	Use
skills.json	JSON array	General-purpose, easy to grep
skills.parquet	Apache Parquet	Fast analytics (pandas, DuckDB, Polars, Arrow)
skills.csv	CSV (tags |-joined)	Spreadsheets, ad-hoc tools
skills.ndjson	Newline-delimited JSON	Streaming consumers, jq, line-by-line
feed.atom	Atom 1.0	Readers (Feedly, Inoreader, NetNewsWire)
CITATION.cff	Citation File Format	Academic citation metadata

Quick start

# Grep for all "deployment"-related skills via jq:
curl -sL https://github.com/claudskills/catalog-public/raw/main/skills.ndjson \
  | jq 'select(.category == "ops" or (.tags | index("type:deployment")))'

# Load into DuckDB:
duckdb -c "SELECT category, COUNT(*) FROM read_parquet('https://github.com/claudskills/catalog-public/raw/main/skills.parquet') GROUP BY 1 ORDER BY 2 DESC"

# pip install + pandas:
python -c "import pandas as pd; df = pd.read_parquet('https://github.com/claudskills/catalog-public/raw/main/skills.parquet'); print(df.describe())"

Catalog snapshot

Field	Value
Total skills	69,369
Featured	20
Daily-eligible (quality ≥ 80)	7,297
Unique authors with ≥1 admitted skill	5,256
Categories	10
Rows dropped this run (no attribution)	0
Last refresh	2026-05-25 07:00 UTC

Top categories

Category	Skills
general	36,953
engineering	13,604
security	3,834
content	3,198
tools	3,189
science	2,872
product	2,542
growth	1,714
sales	1,256
ads	207

Top licenses

License	Skills
(unspecified)	47,442
MIT	17,344
Apache-2.0	3,379
AGPL-3.0	189
NOASSERTION	122
CC-BY-4.0	107
Complete terms in LICENSE.txt	91
MIT + Commons Clause	85
BSD-3-Clause	65
GPL-3.0	62

Top authors

Author	Skills
Jeremy Longshore jeremy@intentsolutions.io	3,304
mahipal	715
diegosouzapw	681
Pranav Nagrecha	664
majiayu000	643
aiskillstore	362
Klotzkette	359
Raishin	355
aipoch	334
onfire7777	287

Schema

Each row carries the following fields. Fields marked required are present on 100% of rows; the rest depend on what the upstream source disclosed.

Field	Type	Required	Description
slug	string	✓	URL-safe identifier, unique across the catalog
name	string	✓	Human-friendly display name
description	string	✓	One-paragraph summary (from SKILL.md frontmatter)
category	string	✓	Top-level taxonomy (engineering, sales, science, …)
subcategory	string	✓	Refined category (~118 sub-topics)
tags	string[]	✓	Orthogonal labels (lang:python, ai:claude, …)
featured	bool	✓	Editorial selection flag
daily_eligible	bool	✓	true if Pro quality_score ≥ 80 (top ~12%)
lastmod	YYYY-MM-DD	✓	Last meaningful change to the skill
catalog_url	URL	✓	Canonical home of this row on claudskills.com
author	string	—	Best-effort author/maintainer name
author_url	string	—	Canonical author page (GitHub profile, blog, etc.)
source_url	string	—	Where the SKILL.md was harvested from upstream
homepage	string	—	Project website if separate from the source repo
license	string	—	SPDX identifier or free-text license name

Attribution

Every row in this catalog carries catalog_url — the row's canonical home on claudskills.com — as the floor attribution. Where the upstream source disclosed it, we additionally preserve author, author_url, source_url, homepage, and license verbatim from the upstream SKILL.md frontmatter.

Current upstream-attribution coverage (rows with at least one of author/author_url/source_url/homepage):

Field	Rows	%
author	35,075	50.6%
source_url	34,120	49.2%
author_url	29,159	42.0%
license	21,927	31.6%

Upstream coverage grows organically each miner cycle as the SKILL.md-frontmatter backfill picks up newly-disclosed metadata.

If you're an author and want your credit corrected, added, or withdrawn, open an issue on this repo or email acreatorstore@translatea.com — we honour every opt-out within one miner cycle (~12 h).

License

This dataset is published under CC BY 4.0. You may use, redistribute, and build on it (including for training machine-learning models or commercial products) as long as you attribute ClaudSkills.com as the source. Per-skill licenses (visible in the license field of each row) are preserved verbatim from the upstream SKILL.md files and apply to the skill content itself — independently of the catalog's CC BY 4.0 wrapper.

How to cite

@misc{claudskills_2026,
  author = {ClaudSkills},
  title  = {ClaudSkills Open Catalog of Claude Code Skills},
  year   = {2026},
  url    = {https://claudskills.com},
  note   = {Dataset mirror: \url{https://github.com/claudskills/catalog-public}}
}

See CITATION.cff for the machine-readable version.

Refresh schedule

• Source auto-mined twice daily (02:00 + 14:00 local time) at claudskills.com.
• This mirror refreshed daily at 03:00 UTC (after the nightly miner pass settles).
• Mirrors on HuggingFace at huggingface.co/datasets/claudskills/skills refresh on the same cadence.

If the refresh has been stale for >36 hours, check GitHub Actions.

Privacy + opt-out

ClaudSkills is an open registry. Skills are mined from public sources only (GitHub, GitLab, awesome-lists, Reddit, Bluesky, etc.) and all attribution is preserved. If you are an author and want your skill removed (not just credit-corrected), open an issue on this repo or email acreatorstore@translatea.com — we honour every opt-out request within one miner cycle.

---

This file is regenerated on every export pass. Don't hand-edit — edit workflows/W1_export.py in the automation repo instead.