pr-review
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external pull requests and issues, which are treated as claims to be verified. This creates a surface for indirect prompt injection if the PR content contains instructions meant to manipulate the reviewer's verdict.
- Ingestion points: Pull request titles, descriptions, tests, and code at the head commit are explicitly processed (SKILL.md).
- Boundary markers: The instructions lack delimiters or explicit warnings to the agent to disregard instructions embedded within the PR content.
- Capability inventory: The skill identifies PRs and reads their content using repository tools to evaluate meaningfulness, correctness, and minimality.
- Sanitization: No sanitization or filtering of the external PR data is performed before the agent evaluates it.
- [INDIRECT_PROMPT_INJECTION]: The skill is configured for implicit invocation, allowing it to automatically analyze pull requests without a specific user command, which increases the exposure to untrusted data (agents/openai.yaml).
Audit Metadata