cross-modal-review
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external work products such as code diffs, brain pages, and analysis to be processed by a reviewer model. This introduces a surface for indirect prompt injection if the ingested data contains instructions designed to manipulate the secondary model's output.
- Ingestion points: Captures diffs, analysis, and pages from the project context as specified in the capture phase (SKILL.md).
- Boundary markers: The instructions do not explicitly specify the use of delimiters or 'ignore' instructions when interpolating the work product into the reviewer's prompt.
- Capability inventory: The skill utilizes
search,query, andget_pagetools and executes internal commands likegbrain eval cross-modal. - Sanitization: No sanitization or filtering logic is documented for the work product before it is sent for review.
Audit Metadata