The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents installation procedures for Linux and macOS that utilize elevated privileges and shell commands (e.g., sudo dd, sudo tee, sudo apt install). These are standard for system-level tool installation.
[EXTERNAL_DOWNLOADS]: Includes instructions for downloading the GitHub CLI from cli.github.com and installing tool extensions via gh extension install. These operations target well-known developer services.
[DATA_EXFILTRATION]: Documents commands that interact with sensitive credentials, such as displaying authentication tokens (gh auth token, gh auth status --show-token) and adding private SSH keys (gh ssh-key add ~/.ssh/id_rsa.pub). While these are intended features of the tool, they involve the handling of high-value secrets.
[PROMPT_INJECTION]: The skill describes an interface for processing untrusted data from external sources (GitHub issues, PRs, and comments) while possessing significant capabilities to modify the environment, which is a surface for indirect prompt injection.
Ingestion points: gh issue list, gh issue view, gh pr list, gh pr view in SKILL.md.
Boundary markers: Absent.
Capability inventory: gh api, gh run download, gh repo delete, gh pr merge, gh secret set, and gh extension install in SKILL.md.
Sanitization: Not specified in the documentation.

gh-cli