craft-system
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes project-level files like
DESIGN.mdand content from external design tools (Figma, Paper) to inform its design system decisions. This represents an attack surface where malicious instructions could be embedded in untrusted data. - Ingestion points:
DESIGN.md(project root), Figma variables/components, Paper artboards. - Boundary markers: None specified in the skill instructions to distinguish between data and instructions.
- Capability inventory: Shell command execution (headless browser), file system read/write (HTML/CSS generation), and design tool manipulation via MCP.
- Sanitization: No explicit sanitization or validation of the design brief content or design tool data is mentioned.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to generate HTML and CSS files based on design tokens and then render them using a headless browser for visual verification.
- Execution method: Running headless Chrome or Playwright on locally generated HTML files to capture screenshots.
- [COMMAND_EXECUTION]: The
references/adapters/html.mdfile provides specific shell command templates for the agent to execute, such as using headless Chrome to take screenshots of the generated design boards ("<chrome>" --headless=new --screenshot=board-1280.png --window-size=1280,900 file://$PWD/board.html).
Audit Metadata