skill-installer

To install agent skills from a GitHub repository, you provide the repository path or URL, and the installer handles direct archive downloads or sparse Git checkouts directly into your local environment.

Curated agent skills are pre-selected collections available from default or experimental OpenAI repositories. You list them to discover available skills, see installation status annotations, and select experimental skills for your environment.

Yes, you can install skills from private GitHub repositories by providing environment-based authentication using Git credentials or GitHub tokens to authorize access during the repository cloning process.

Git is not strictly required because the installer supports direct archive downloads alongside sparse Git checkouts. The skill installer relies on Python standard-library tooling and network access to fetch and validate skill directories.

A sparse checkout is a Git technique used to fetch only specific directories from a repository. It is used during skill installation to download only the required skill folder containing SKILL.md, reducing download time and storage.

You validate a downloaded agent skill directory by checking for the presence of a SKILL.md file inside the target folder. The skill installer performs this validation automatically to ensure the directory is a functional agent skill.