check
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-controlled source code (CSS, JSX, TSX, etc.) and external design configuration files to generate compliance reports and code fixes. This constitutes a data ingestion surface where malicious instructions could theoretically be embedded in comments or strings within the scanned files.
- Ingestion points: Processes content from
.design-lock,.design-cache/, and source files matching extensions like.css,.scss,.tsx,.jsx,.html,.vue, and.svelte. - Boundary markers: No specific delimiters or safety warnings are defined to separate ingested file content from the agent's internal instructions.
- Capability inventory: The skill performs file system read operations across the project and offers to modify source files via an auto-fix feature.
- Sanitization: The instructions do not specify any sanitization or validation of the ingested content beyond identifying design-related tokens (hex codes, font names).
Audit Metadata