ai-elements
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [Component Installation via Remote Registry]: The skill recommends installing UI components using the
npx shadcn@latest addcommand pointing to the officialai-sdk.devregistry. This pattern involves fetching component definitions and source code from a remote server, which is the standard distribution method for this ecosystem. - [Execution of CLI Tools]: Instructions include the use of
npx ai-elements, which downloads and runs a package from the npm registry. This is a common developer workflow for component management and scaffolding. - [Indirect Prompt Injection Surface]: This skill provides tools to process and render untrusted data such as AI-generated text and tool outputs.
- Ingestion points: Untrusted content enters the application context via properties like
message.content,event.briefing, ornarrativeTextas described inSKILL.md. - Boundary markers: The skill encourages the use of specialized components like
<MessageResponse>to wrap content, which provides better structure than raw text rendering. - Capability inventory: The identified capabilities include shell command execution for installation (
npx) and component-based UI rendering. - Sanitization: The recommended
<MessageResponse>component uses a dedicated markdown renderer (Streamdown), providing a layer of processing between raw text and the DOM.
Audit Metadata