The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the GitHub CLI (gh) for managing pull request reviews. It includes explicit instructions to verify that the tool is installed and authenticated before proceeding, preventing errors or unexpected behavior.
[DATA_EXFILTRATION]: Network operations are performed via gh api targeting GitHub, which is a well-known and trusted service. No patterns of sensitive local file access or exfiltration to unauthorized domains were found.
[INDIRECT_PROMPT_INJECTION]: The skill processes external Pull Request data which could theoretically contain malicious instructions. However, it implements a mandatory security control requiring the agent to use AskUserQuestion to show the user exactly what will be posted and wait for explicit confirmation. This human-in-the-loop requirement mitigates risks associated with untrusted data ingestion.
Ingestion points: Pull request content and metadata read via gh api and gh pr view commands.
Boundary markers: Standard markdown blocks are used to structure drafted comments.
Capability inventory: gh api for reading and writing to GitHub repositories.
Sanitization: Mandatory user review and approval of all drafted content before any write operations occur.

github-pr-review