agent-ui
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to download component registry data from the developer domain at ui.inference.sh.
- [COMMAND_EXECUTION]: Instructions include using the npx shadcn utility to download and install components from a remote URL, which is a standard procedure for this UI library.
- [INDIRECT_PROMPT_INJECTION]: The component serves as a chat interface for AI agents, which is an inherent surface for indirect prompt injection. 1. Ingestion points: Processes user messages and outputs from client-side tools (e.g., scan_ui and fill_field) within the Agent component. 2. Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore embedded commands in the data. 3. Capability inventory: The skill enables client-side tool execution, file/image uploads, and API proxying. 4. Sanitization: There is no mention of input validation or sanitization for data processed by the agent.
Audit Metadata