web-prototype
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from a user's brief and interpolate it into a generated HTML file.
- Ingestion points: Untrusted data enters the context via the "user's brief" and a
DESIGN.mdfile referenced in theSKILL.mdworkflow. - Boundary markers: The skill includes strict "Hard rules" and a detailed
references/checklist.mdthat provide guidance on output formatting and content constraints, acting as a safeguard against accidental instruction following. - Capability inventory: The skill uses tools to search for and fetch external images and writes the final prototype to a local file (
index.html). - Sanitization: The instructions do not specify a sanitization or escaping process for user-supplied strings before they are pasted into the HTML template, representing a potential injection surface.
Audit Metadata