ag-ui-a2ui-integration
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes a feedback mechanism where user interactions within the rendered UI are forwarded back to the agent as private user context starting with the header
A2UI user action:. While this creates an attack surface for indirect prompt injection from the client-side UI, the skill provides mitigation guidance by recommending the use of Zod schemas and runtime toolkit validation to ensure generated components and received actions match the defined catalog. - [DYNAMIC_EXECUTION]: The integration supports a 'Dynamic Schema Mode' where a primary agent uses a
generate_a2uitool to delegate to a sub-agent that streams UI components. This runtime generation of executable UI structures is a core functional feature and is implemented via framework-specific adapters that include validation layers. - [EXTERNAL_DOWNLOADS]: The documentation references legitimate, vendor-maintained packages and official resources from trusted or well-known ecosystems such as CopilotKit, LangGraph, and ADK. These references are used for standard dependency management and documentation purposes.
Audit Metadata