The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates downloading and installing third-party GitHub CLI extensions and agent skills. * Evidence: references/advanced-features.md contains instructions for gh extension install owner/gh-extension and gh skill install owner/repo skill-name.
[REMOTE_CODE_EXECUTION]: By teaching the agent to install and upgrade extensions, the skill enables a vector for executing arbitrary code from external sources. * Evidence: references/advanced-features.md documents gh extension install and gh extension upgrade --all.
[COMMAND_EXECUTION]: The skill relies extensively on executing shell commands through the GitHub CLI and git. * Evidence: Multiple files including SKILL.md and examples/*.sh use gh and git commands for repository management and automation.
[DATA_EXFILTRATION]: The skill provides the agent with capabilities to access sensitive information such as repository secrets, variables, and SSH/GPG keys. * Evidence: references/advanced-features.md documents commands like gh secret list, gh secret set, gh ssh-key list, and gh ssh-key add ~/.ssh/id_ed25519.pub.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub repositories, issues, and discussions which may contain malicious instructions. * Ingestion points: references/discussions.md (gh discussion list), references/search.md (gh search issues), and references/workflows-actions.md (gh run view --log). * Boundary markers: Absent; there are no instructions to the agent to treat fetched external content as data only or to ignore instructions embedded within it. * Capability inventory: Extensive capabilities including write operations, secret management, and extension installation. * Sanitization: Absent; the skill does not specify any methods for sanitizing or escaping the data fetched from GitHub before it enters the agent's context.

cli-gh