GitSkills: A Dataset of Agent Skills on GitHub

An agent skill is a folder containing a SKILL.md file with instructions for a language-model agent, optionally accompanied by scripts and reference files. The agent loads the skill when it judges that a task matches the skill description. Anthropic introduced the format in October 2025 as an open specification. Nine months later, skill files in the millions sit in public GitHub repositories — yet no dataset recorded this population before GitSkills.

GitSkills records 3,797,117 SKILL.md files collected from 282,200 public GitHub repositories, owned by 195,841 accounts, in July 2026. The files are grouped by content hash into 1,877,981 distinct contents. One representative file per group is enriched with its full text, parsed front matter, folder contents, repository metadata, and — for 458,548 files in standard locations plus a size-stratified sample of the rest — the commit history of the file, with anonymised first- and last-commit author accounts. Every occurrence is retained with its own repository and path, so both unique contents and their copies can be studied.

The dataset is distributed as a single self-contained SQLite file (agent_skills_release.db) with four tables:

• artifacts (3,797,117 rows) — one row per discovered file: repository, path, basename, location class, content hash; for representatives, also full text, parsed front matter, and commit history.
• artifact_siblings (7,264,865 rows) — scripts and reference files stored alongside a representative skill.
• repos (282,200 rows) — repository metadata: owner, stars, language, fork status, license, dates.
• mining_runs (7 rows) — provenance log of each collection run.

Collection is read-only, against the GitHub REST, GraphQL, and code-search APIs and the raw-content CDN. Discovery partitions the search space by file size to work around the code-search API's 1,000-result cap; deduplication groups files by content hash and selects one representative per group for enrichment, while retaining every occurrence with its own repository and path. The dataset covers public repositories only and should be read as a lower bound on the population.

Commit author accounts are replaced by keyed one-way codes, identical for the same account throughout the dataset, so authorship can be traced without identifying anyone; the key used to generate the codes is not distributed. Bot accounts keep their login. Email addresses and personal names in commit messages are redacted.

A Parquet mirror (sharded by table, for browsing and use with the datasets library) is available on Hugging Face at huggingface.co/datasets/mvaccargiu/gitskills. A sample accompanying the original submission is at github.com/giuseppedestefanis/gitskills-sample.

This dataset accompanies a paper accepted at the MSR 2027 Mining Challenge, also available as a preprint at arXiv:2608.10906. If you use the GitSkills dataset, please cite:

G. Destefanis, D. Graziotin, M. Vaccargiu, and M. Ortu, “GitSkills: A Dataset of Agent Skills on GitHub,” in Proceedings of the 24th IEEE/ACM International Conference on Mining Software Repositories (MSR ’27). IEEE, Piscataway, NJ, USA, 2027, 3 pp. To appear. DOI: 10.48550/arXiv.2608.10906