hero-section
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes a verification step that uses
npx playwright screenshotto capture mobile and desktop previews of the generated HTML. This involves executing the Playwright package from a shell to perform visual audits of the code produced by the agent. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted or external data during its operation.
- Ingestion points: Reads project-specific design tokens from
.agents/design-context.mdand fetches design data from external Figma URLs using the Figma MCP (get_design_context). - Boundary markers: None explicitly defined for the ingestion of external design data.
- Capability inventory: The skill generates front-end code (HTML/CSS/JS) and executes shell commands for screenshot verification via Playwright.
- Sanitization: The skill uses a mapping process to transform external design tokens into internal CSS variables, which provides a layer of structured data handling.
Audit Metadata