The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from GitHub via commands like gh issue view or gh pr list, creating a surface for indirect prompt injection.
Ingestion points: Commands such as gh issue view, gh pr view, and gh api fetch user-generated content from external, potentially attacker-controlled repositories.
Boundary markers: There are no instructions or delimiters provided to mitigate the risk of the agent executing commands embedded within fetched issue or PR descriptions.
Capability inventory: The documented commands provide extensive capabilities, including modifying repository settings, managing secrets, and executing code via extensions.
Sanitization: No sanitization or validation of external content is described in the documentation.
[EXTERNAL_DOWNLOADS]: Documents standard installation procedures that fetch GPG keys and repository configurations from the official cli.github.com domain.
[COMMAND_EXECUTION]: Includes a complete reference for commands that allow administrative tasks, such as gh repo delete, gh secret set, and workflow management.
[REMOTE_CODE_EXECUTION]: Describes the gh extension command group, which allows the installation and execution of third-party CLI extensions.
[CREDENTIALS_UNSAFE]: Lists commands for authentication and credential management, including displaying session tokens via gh auth token and managing SSH/GPG keys.

gh-cli