figma-preflight
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources including Figma file layer names, style names, and the CLAUDE.md configuration file. While this represents a theoretical ingestion surface for malicious instructions, the skill only catalogs these names for a status report, presenting minimal risk.
- [DYNAMIC_EXECUTION]: A JavaScript snippet is provided to be executed via the Figma API (likely via the figma-use skill) to inventory local styles and components. The script is static and defined within the skill instructions, representing a standard use of the agent's capability to interact with the Figma environment.
- [SAFE]: The skill implements data minimization by explicitly instructing the agent to store only the names of styles, variables, and components in the session context, rather than full object metadata or sensitive IDs.
- [SAFE]: Authenticated operations are performed through official MCP tool calls, and no unauthorized network exfiltration or obfuscation patterns were detected.
Audit Metadata