design-md
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external HTML and CSS source code, which creates an attack surface for indirect prompt injection.
- Ingestion points: The skill retrieves external code and images via
web_fetchorread_url_contentusing URLs provided by the Stitch MCP server (SKILL.md, step 5 in 'Retrieval and Networking'). - Boundary markers: The instructions do not specify any boundary markers or delimiters to isolate the untrusted external content or to explicitly tell the agent to ignore any embedded instructions.
- Capability inventory: The skill possesses file-writing capabilities (creating
DESIGN.md) and network read access. - Sanitization: There is no evidence of sanitization, escaping, or validation of the external content before it is parsed and synthesized into documentation.
Audit Metadata