design-reference-first-ui
Pass
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The instructions encourage high autonomy, advising the agent to proceed with design and implementation without pausing for user approval gates ("Do not add a user-approval pause", "build the requested first screen without adding an approval gate"). Additionally, the skill's workflow involves researching and ingesting patterns from external "production references" (live web content and screenshots), which establishes an attack surface for indirect prompt injection. These elements are part of the intended UI design workflow and do not attempt to bypass core safety constraints.
Audit Metadata