figma-implement-design
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill connects to Figma's official Model Context Protocol (MCP) server to retrieve design specifications, metadata, and assets.
- Evidence: The tool configuration in
agents/openai.yamlreferenceshttps://mcp.figma.com/mcp, and the workflow inSKILL.mdinvolves fetching screenshots and assets from this remote source. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external content from Figma designs, which could theoretically contain instructions embedded in layer names, text nodes, or metadata intended to influence the agent's behavior during code generation.
- Ingestion points: The agent retrieves data using the
get_design_contextandget_metadatatools as described inSKILL.md(Step 2). - Boundary markers: There are no explicit delimiters or specific instructions provided to the agent to treat the ingested Figma data as untrusted text or to ignore potential instructions embedded within the design.
- Capability inventory: The agent has the capability to generate and write production-ready code to the local filesystem based on the translated design (Step 5 and 6).
- Sanitization: The skill does not implement explicit sanitization or validation of the text content extracted from Figma nodes before it is used to generate code.
Audit Metadata