The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends the installation of an external GitHub CLI extension (abekdwight/gh-pr-review-check) from a source that is not verified as a trusted vendor.
[COMMAND_EXECUTION]: The skill executes multiple shell commands, including git push, gh api for GraphQL mutations, and local build tools such as npm or make, based on the content of external PR reviews. This provides a broad attack surface if malicious instructions are embedded in PR comments.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes PR descriptions and comments written by potentially untrusted external users to guide code fixes and repository modifications.
Ingestion points: PR metadata, comments, and review threads are ingested via the gh CLI and a bundled collection script.
Boundary markers: The instructions require the agent to maintain objectivity using "Fact-based analysis" and "Non-flattery" scores to mitigate bias, though these are guidelines rather than strict filters.
Capability inventory: The skill allows pushing code to the repository, interacting with GitHub APIs to resolve threads, and executing project-specific test commands.
Sanitization: It utilizes file-based variable passing for GraphQL queries to prevent injection during API communication, which follows security best practices for that specific vector.

review-pr-check