github-pr-reviewer
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The setup process involves the agent executing local shell commands to package and upload the automation script to the platform's internal API. These commands are necessary for the skill's setup and are restricted to the local environment and authenticated platform endpoints.
- [EXTERNAL_DOWNLOADS]: The automation script communicates with the GitHub API to retrieve pull request data and post review comments. This interaction targets a well-known technology service and is fundamental to the skill's primary purpose.
- [CREDENTIALS_UNSAFE]: The skill manages a GitHub access token via the platform's built-in secret store (
get_secret). It follows security best practices by retrieving the token dynamically rather than hardcoding credentials or exposing them in logs. - [PROMPT_INJECTION]: The skill processes external data from pull request descriptions, which creates a surface for indirect prompt injection. This risk is mitigated through the use of delimiters in the AI prompt and specific instructions that restrict the reviewing agent's capabilities, such as preventing it from pushing code or approving reviews via the API.
Audit Metadata