Skip to main content
Arka can import data files from public GitHub repositories without needing a custom skill for each repo. It downloads the repository archive, extracts common data formats, caches them locally, and writes a manifest with file previews.
Natural language routes to the same importer:

​
What Arka extracts

The generic importer scans the repo archive for:
  • .csv
  • .json
  • .jsonl / .ndjson
  • .tsv
  • .yaml / .yml
  • .xml
  • .parquet
It skips obvious build and dependency folders such as node_modules, .git, venv, dist, and build.

​
Cache location

Imported files are stored under:
The manifest includes source URL, branch, cached file paths, file sizes, and best-effort previews such as CSV columns or JSON keys.

​
Licensing

Arka does not infer a repository’s license or redistribution rights. Before using imported data in a product, dataset, demo, or public export, inspect the upstream repository’s license and README. For heavily curated sources, Arka may still provide a named preset such as arka exercises; those presets can add domain-specific normalization on top of the generic GitHub dataset importer.

Related topics

GitHub ActionsQuickstart: install Arka and run your first commandHow to code with ArkaSkills catalog: 70+ built-in Arka commandsArka — AI terminal agent documentation