The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs legitimate operations by invoking git and the GitHub CLI (gh) via the subprocess.run module. Technical analysis of the scripts/pages_publish.py file confirms that commands are constructed using argument lists rather than shell strings, which effectively prevents shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill interacts with GitHub APIs to clone repositories and resolve the latest versions of official GitHub Actions (such as actions/checkout and actions/deploy-pages). These network operations are directed at trusted GitHub infrastructure and are essential for the skill's stated purpose of managing repository workflows.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text or files to generate site content. This creates an attack surface where malicious input could influence the generated HTML. However, the skill provides basic sanitization by escaping HTML special characters in its fallback template and relies on an external dependency for primary design tasks, significantly reducing the risk of unintended behavior.
[SAFE]: High-risk actions, including the creation of new repositories, modifying repository visibility (making private repos public), and enabling GitHub Actions, are explicitly protected by 'confirmation gates'. These gates require the agent to pause and obtain manual user approval before proceeding, ensuring user control over sensitive account changes.

github-pages-publish