figma-roundtrip
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads metadata and content from Figma boards (e.g., fonts, weights, sizes, variables, and text layers) and ports those changes into local files like brand.json and component source code. This creates an indirect prompt injection surface where malicious design data could manipulate the agent's file-writing actions.
- Ingestion points: Data is ingested from Figma board layers and variables via the Figma MCP.
- Boundary markers: The skill lacks boundary markers or instructions to the agent to disregard instructions that might be contained within the design data.
- Capability inventory: The agent can read from external Figma files, modify local JSON configuration, and update component code files.
- Sanitization: No data sanitization or validation protocols are specified to ensure the Figma data is safe before it is applied to the codebase.
Audit Metadata