react-native

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The documentation provides instructions to initialize projects using npx assistant-ui@latest create and install dependencies via npx expo install. These are standard operations for project setup and package management using the vendor's official CLI tools.
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes a UI layer for chat applications, which inherently processes and renders untrusted content from AI models.
  • Ingestion points: MessagePrimitive.Content (SKILL.md) and ThreadPrimitive.MessagesFlatList (references/primitives.md).
  • Boundary markers: None are present in the UI primitives; message filtering and boundary enforcement are expected to be managed at the backend transport or model level.
  • Capability inventory: The skill provides components for network communication (AssistantChatTransport), file attachment handling (AttachmentAdapter), and clipboard operations (ActionBarPrimitive.Copy).
  • Sanitization: UI primitives render content as provided by the model; additional sanitization or custom rendering (e.g., Markdown) must be implemented by the developer.
  • [DYNAMIC_EXECUTION]: The skill utilizes a Metro bundler wrapper, withAui, to support a "use generative" directive for toolkit compilation. This is a build-time transformation used to generate tool schemas from source code, which is a standard feature of the assistant-ui framework.
  • [EXTERNAL_DOWNLOADS]: The skill fetches resources and packages from the official assistant-ui repositories and standard registries (npm). It also makes network requests to developer-configured API endpoints for chat functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 06:03 AM
Security Audit — agent-trust-hub — react-native