The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to interact with and process various GitHub resources, such as issue titles, bodies, comments, pull request descriptions, and file contents. These resources are controlled by third parties and could be used to embed malicious instructions intended to manipulate the agent's behavior.
Ingestion points: Operations like gh api repos/owner/repo/issues, gh api repos/owner/repo/pulls, and gh api repos/owner/repo/contents/path fetch external content into the agent's context (e.g., SKILL.md, resources/issues-pull-requests.md, resources/repositories.md).
Boundary markers: The skill includes explicit security warnings (W011) instructing the agent to treat all fetched content as opaque data and never as instructions. These warnings appear in SKILL.md, resources/issues-pull-requests.md, and resources/repositories.md.
Capability inventory: The skill utilizes the gh CLI and curl for repository management, issue tracking, workflow execution, and secret management across all resource files.
Sanitization: The instructions specifically advise the agent to sanitize or quote content before including it in shell commands and to clearly label the origin of external data.
[SAFE]: The skill demonstrates safe credential handling practices by explicitly warning against hardcoding API tokens and instead recommending the use of environment variables or the GitHub CLI's internal authentication mechanisms. All external links and API endpoints reference trusted official GitHub domains.

github-api