agent-ui
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides commands to download configuration files and software packages from
ui.inference.shandnpm. It also references thebelt-sh/cliskill. These resources are directly related to the stated purpose of the skill and come from infrastructure consistent with the tool's vendor. - [INDIRECT_PROMPT_INJECTION]: The skill describes a UI component that processes user-defined configurations and prompts, creating a surface for potential instruction injection.
- Ingestion points: The
agentConfigandconfigproperties in theAgentcomponent withinSKILL.md. - Boundary markers: None identified in the provided code examples.
- Capability inventory: The component includes capabilities for browser-based tool execution (e.g.,
scan_ui,fill_field), human-in-the-loop approval workflows, and file/image handling. - Sanitization: Not demonstrated in the examples; however, the component is a development building block meant to be integrated into secured applications.
Audit Metadata