figma-to-code
Warn
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill contains specific instructions for the agent to prompt the user to install an external component from an untrusted GitHub repository (github:mindverse/skillhub). This encourages the installation and potential execution of code from an unverifiable third-party source.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided content such as design descriptions and screenshots to generate code, but it lacks sufficient safeguards to prevent potential prompt injection attacks within that data.
- Ingestion points: User-provided design descriptions, screenshots, and technical requirements mentioned in the '理解设计稿' (Understand Design) section.
- Boundary markers: Absent. The instructions do not define delimiters or warnings to ensure the agent treats the user-provided design data purely as information rather than as executable instructions.
- Capability inventory: The skill is configured to generate text-based code (HTML, CSS, React, Vue) and does not specify access to external tools, network resources, or file system writing capabilities in the frontmatter.
- Sanitization: Absent. No logic is provided to filter, validate, or sanitize the user-provided design descriptions before they are processed by the agent.
Audit Metadata