figma-to-code
Pass
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill files and instructions contain no malicious patterns, obfuscation, or security vulnerabilities. The logic is consistent with the stated purpose of assisting in Figma-to-code development workflows.\n- [PROMPT_INJECTION]: The skill proactively addresses Indirect Prompt Injection risks associated with processing external design data. In
references/verification-and-boundaries.md, the 'Prompt-Injection Boundary' section provides a mandatory defense strategy: it defines Figma data (layer names, text content, comments) as untrusted, warns against following instructions embedded in the design, and mandates that repository instructions take precedence over generated content. This follows best practices for handling untrusted input from external platforms.\n- [DATA_EXFILTRATION]: No unauthorized data exposure or exfiltration patterns were identified. The skill correctly utilizes established Figma MCP tool protocols to fetch design context and assets. Sensitive files such as SSH keys or environment variables are not accessed, and network operations are limited to the intended Figma and local asset sources.\n- [COMMAND_EXECUTION]: Command execution is limited to standard development practices, including repository-specific checks (lint, build, tests) and MCP tools. No privilege escalation attempts, persistence mechanisms, or unauthorized system modifications were detected. Instructions emphasize inspecting local project conventions (e.g., package metadata, styling setup) before executing tools, which further grounds the agent in the local environment's security context.
Audit Metadata