lumos-import-figma
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to run a local Node.js script (
convert.mjs) to process design tokens and the Astro development server (astro dev) to verify builds.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect injection as it processes untrusted data from design files.\n - Ingestion points: Reads data from external Figma files (via MCP tools) and a user-provided
design.jsonfile.\n - Boundary markers: Absent. There are no explicit instructions for the agent to ignore potentially malicious text embedded within design names or metadata.\n
- Capability inventory: The skill can read local project files (
src/styles/base.css,package.json) and execute shell commands to run its conversion script and the local dev server.\n - Sanitization: The conversion script performs mathematical operations on numeric data but does not sanitize string values (such as 'name' or 'token' fields from the Figma JSON) before displaying them in reports to the agent, which could influence its next actions.
Audit Metadata