ui-multiple-design-experiment
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: Surface detected due to ingestion of untrusted UI code/screenshots (Step 1) and external web content (Step 1.5). Capabilities include web_search, web_fetch, and file writing. No explicit boundary markers or data sanitization logic are provided in the instructions.
- [EXTERNAL_DOWNLOADS]: The skill performs web searches and fetches content from design platforms for inspiration. It also allows for CDN-hosted fonts/icons in the generated output. These are standard development practices for the intended use case.
Audit Metadata