mistral-sdk-patterns
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFE
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill correctly instructs users to manage sensitive credentials like
MISTRAL_API_KEYusing environment variables (process.envin TypeScript andos.environin Python), which is the recommended safe practice. - [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing external text (e.g., in
classifyTicket). It mitigates risks by implementing structured output with Zod validation and JSON Schema enforcement. - Ingestion points: The
textparameter inclassifyTicketandpromptsinprocess_batchrepresent external data entry. - Boundary markers: The skill uses JSON Schema to constrain LLM responses.
- Capability inventory: The skill utilizes the Mistral AI SDK for network-based inference.
- Sanitization: It uses Zod to validate the structure and content of LLM outputs at runtime.
- [EXTERNAL_DOWNLOADS]: The skill references official SDKs from Mistral AI, a well-known AI service provider. References to their GitHub repositories and official documentation are documented neutrally.
Audit Metadata