generative-ui

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The repository provides a vocabulary of 27 default components for creating generative UIs inside assistant conversations. All rendering patterns are explicitly constrained via structural allowlists or compiled schemas without resorting to dynamic generation mechanisms or runtime execution strings like eval.
  • [SAFE]: External URLs and modules belong to the primary vendor (assistant-ui) or recognized well-known cloud/technology services and specifications (Slack Block Kit, Microsoft Teams Adaptive Cards, A2UI, and OpenUI). They are documented neutrally as native integration pathways.
  • [SAFE]: Input props received by components are validated against Zod schemas, mitigating risks of arbitrary code injection. The documentation explicitly advises developers to handle validation of complex input types, like URLs or raw structures, and warns against exposing untrusted data to unsafe primitives like dangerouslySetInnerHTML.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 04:49 PM
Security Audit — agent-trust-hub — generative-ui