web-prototype
Pass
Audited by Gen Agent Trust Hub on Jul 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted content from external web pages (URLs or screenshots) to generate project code, creating a surface for indirect prompt injection.
- Ingestion points: Target pages are accessed via
mcp__Claude_in_Chrome__read_page,WebFetch, and screenshot analysis as described inSKILL.md. - Boundary markers: The instructions lack explicit delimiters or "ignore" directives to prevent the agent from following instructions potentially embedded in the source page's text or accessibility tree.
- Capability inventory: The skill is capable of writing multiple files to create a Next.js project and suggests user-executed commands like
npm install. - Sanitization: There is no evidence of sanitization or filtering of text extracted from the source page before it is processed by the model to generate the prototype.
Audit Metadata