ai-native-ui-design
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill promotes Generative UI patterns where AI-generated content is directly rendered as the application interface. This model ingests untrusted data from an LLM into the presentation layer.
- Ingestion points: The skill utilizes the streamUI function and components such as GeneratedOutput to render AI responses in SKILL.md and references/animation-patterns.md.
- Boundary markers: While the skill suggests using Contextual Trust badges, it does not explicitly define sanitization boundaries or delimiters for the generated UI elements.
- Capability inventory: The examples demonstrate the use of the tools parameter within the streamUI configuration, enabling tool execution based on model output.
- Sanitization: The provided UI implementation snippets do not include explicit logic for sanitizing or escaping the content streamed from the model.
Audit Metadata