agent-ui
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download component registries and image assets from ui.inference.sh and cloud.inference.sh using npx shadcn and standard markdown links.
- [COMMAND_EXECUTION]: The documentation includes instructions for users to run commands like npx shadcn@latest add, npm install @inferencesh/sdk, and npx skills add to set up the development environment.
- [INDIRECT_PROMPT_INJECTION]: The Agent component renders declarative UI widgets directly from JSON data in agent responses, which is an ingestion point for untrusted content.
- Ingestion points: The component processes JSON data from agent responses to generate UI widgets (SKILL.md).
- Boundary markers: Boundary markers or delimiters are not explicitly defined in the provided code snippets.
- Capability inventory: The agent component supports client-side tools (scan_ui, fill_field) and streaming tokens through a proxy route.
- Sanitization: The documentation does not specify sanitization methods, implying reliance on the internal logic of the @inferencesh/sdk and the Agent component.
Audit Metadata