ai-provider-mistral-sdk
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides legitimate development patterns for integrating the
@mistralai/mistralaiSDK. It includes appropriate security warnings, such as advising against hardcoding API keys and recommending the use of environment variables. - [INDIRECT_PROMPT_INJECTION]: The skill demonstrates patterns for processing untrusted external data, such as user messages and articles, through LLM API calls.
- Ingestion points: Untrusted data enters the context via the
messagesarray in chat methods, string parameters in completion functions, and external image URLs in vision examples (e.g.,examples/chat.md,examples/structured-output.md). - Boundary markers: The instructions emphasize using the
systemrole to define model behavior and separate instructions from data. - Capability inventory: The provided code includes network requests to Mistral AI API endpoints and local file system access (
node:fs) for base64 image processing. - Sanitization: The skill recommends using
safePrompt: trueto enable provider-side safety filters and utilizingzodschemas for strict validation of structured model outputs.
Audit Metadata