The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the gh CLI for its core functionality. The scripts/github.py script wraps the CLI using subprocess.run to execute commands, and the skill's instructions guide the agent to use gh directly for repository actions.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes untrusted content from GitHub, which could contain instructions intended to compromise the agent's behavior.
Ingestion points: Fetches issue and pull request titles, bodies, and comments via scripts/github.py (e.g., in format_issue_summary and format_pr_summary).
Boundary markers: The skill lacks explicit delimiters or instructions for the agent to ignore potentially malicious content within retrieved GitHub data.
Capability inventory: The agent is authorized to perform sensitive actions such as merging pull requests, running workflows, and modifying repository settings via the gh CLI.
Sanitization: Content fetched from GitHub is formatted into Markdown and presented to the agent without filtering or sanitization.
[EXTERNAL_DOWNLOADS]: The skill enables downloading external resources from GitHub, including cloning repositories (gh repo clone) and fetching build artifacts (gh run download).

github