agent-ui

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill documentation includes instructions to download a component registry file from 'ui.inference.sh' using the shadcn CLI. These resources are hosted on the vendor's official domain.
  • [CREDENTIALS_UNSAFE]: The setup guide references an 'INFERENCE_API_KEY' environment variable required for the proxy route. The documentation uses a safe 'inf_...' placeholder and does not contain hardcoded secrets.
  • [COMMAND_EXECUTION]: The skill provides standard installation commands using 'npm' and 'npx' for the author's SDK and related UI skills. These commands are intended for the initial setup of the developer environment.
  • [PROMPT_INJECTION]: The skill defines a surface for processing agent outputs within a UI component. This includes rendering widgets and executing client-side tools like form filling based on model responses.
  • Ingestion points: Agent responses and tool definitions are received via the configured 'proxyUrl'.
  • Boundary markers: Not explicitly detailed in the component configuration snippets.
  • Capability inventory: The component supports client-side form interaction ('scan_ui', 'fill_field') and UI widget generation.
  • Sanitization: Data sanitization and validation are handled by the '@inferencesh/sdk' and the internal component logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 09:26 PM