figma-code-connect

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • Indirect Prompt Injection Surface: The skill ingests component definitions and property data from a Figma MCP server to generate code snippets. This creates a potential surface where malicious or unexpected metadata from an external design file could influence the agent's output during template generation.
  • Ingestion points: Figma component names, properties (TEXT, BOOLEAN, VARIANT, SLOT, INSTANCE_SWAP), and selection suggestions fetched via the get_code_connect_suggestions and get_context_for_code_connect MCP tools.
  • Boundary markers: The skill relies on structured API mapping rather than explicit boundary delimiters for the raw design data.
  • Capability inventory: The agent searches and reads local project files to match component property interfaces and writes generated .figma.ts files to the local project directory.
  • Sanitization: The skill provides a rigorous mapping framework using specific library methods (e.g., getEnum, getBoolean, getString) to validate and transform design properties into predefined code structures.
  • External Package Integration: The skill references the @figma/code-connect package, which is the official library for Figma's Code Connect feature. It provides configuration guidance for integrating these types into the project's development environment.
  • Command Execution Guidance: The provided reference documentation includes example shell commands for publishing and unpublishing mappings using the Figma CLI tool. These examples are for user reference and utilize standard placeholders for sensitive values such as access tokens.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 11:36 AM
Security Audit — agent-trust-hub — figma-code-connect