design-md
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch design systems from external sources and apply them as the "single source of truth" for UI generation. This creates a vulnerability where malicious instructions embedded within a design.md file could be executed by the agent.
- Ingestion points: Fetches content from arbitrary user-provided URLs and the gallery at
designsmd.vercel.app(SKILL.md). - Boundary markers: The instructions lack delimiters or warnings to ignore embedded natural language instructions within the design tokens.
- Capability inventory: The agent generates complex UI code (HTML/CSS) based on the fetched specifications.
- Sanitization: There is no requirement for the agent to validate or sanitize the fetched YAML/Markdown content before processing it.
- [EXTERNAL_DOWNLOADS]: The skill fetches design tokens and configuration from external repositories and services.
- Evidence: Automatically fetches design specifications from Vercel's official domain and the curated gallery at
designsmd.vercel.app. It also allows fetching from any URL provided by the user.
Audit Metadata