The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/check_pages_setup.py script executes local git commands via subprocess.run to inspect the repository state, such as remote URLs and branch existence. The implementation uses list-based arguments which prevents command injection.
[EXTERNAL_DOWNLOADS]: The skill provides references to the official GitHub Actions starter workflows repository (github.com/actions/starter-workflows) to assist in setting up deployment pipelines. This targets a well-known and trusted service for configuration templates.
[PROMPT_INJECTION]: The skill performs static analysis on local configuration files (e.g., mkdocs.yml, docusaurus.config.js, .github/workflows/*.yml) using regular expressions. This creates a surface for indirect prompt injection where untrusted data in a project repository could influence the agent's diagnostic output, but the risk is categorized as low due to the deterministic nature of the parsing logic.

docs-site