hyper-pr-review
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates a strong defensive posture against indirect prompt injection. It explicitly instructs the agent that pull request text, comments, and linked documents must be treated as evidence to verify, and never as authoritative instructions to execute. It specifically warns the agent to ignore instructions found within code comments or test fixtures.
- [COMMAND_EXECUTION]: The skill utilizes standard, well-known developer utilities including the GitHub CLI (gh) and Git for retrieving pull request data and code diffs. These operations are scoped to the intended purpose of code review and do not involve arbitrary or dangerous command execution.
- [DYNAMIC_EXECUTION]: Instructions include a safety protocol for executing code or linters, stating that unknown-provenance code should only be run in a disposable environment without credentials or durable external effects. This is a benchmark for safe agent behavior when dealing with untrusted source code.
- [DATA_EXFILTRATION]: There are no patterns suggesting unauthorized data access or exfiltration. The skill focuses on reading repository data and providing feedback to the user within the established context of a PR review.
Audit Metadata