ai-core/structured-outputs
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents patterns that ingest untrusted user input through the
messagesparameter in thechat()anduseChat()functions. - Ingestion points: The
messagesarray and thesendMessagefunction inSKILL.mdaccept raw user input that is passed to language models. - Boundary markers: The skill uses
outputSchema(via Zod, ArkType, or Valibot) to constrain the LLM's output into a specific JSON structure, which acts as a structural boundary. - Capability inventory: The skill utilizes network capabilities to communicate with LLM providers (OpenAI, Anthropic, Gemini, etc.) and performs UI rendering of model outputs.
- Sanitization: Standard Schema validation (Zod, etc.) is employed to validate the structure of the model's response, though content-level sanitization is left to the developer.
- [EXTERNAL_DOWNLOADS]: The skill references documentation and source code from the official TanStack GitHub repositories (
TanStack/ai). These are vendor-owned resources and are used for instructional purposes.
Audit Metadata