skills/jeremylongshore/tons-of-skills-marketplace/mistral-reference-architecture/Gen Agent Trust Hub
mistral-reference-architecture
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for building prompt templates and a service layer that processes external data, which introduces a surface for indirect prompt injection.
- Ingestion points: Data enters the system via the
ChatService.completeandChatService.streammethods (insrc/services/chat.service.ts) and theuserTemplatefunctions within thePROMPTSobject (insrc/mistral/prompts.ts). - Boundary markers: Some templates, such as
codeReview, use markdown code blocks as delimiters, but other templates lack explicit boundary markers or instructions to ignore embedded commands. - Capability inventory: The skill utilizes the
ReadandGreptools and facilitates network interaction with the Mistral AI API for chat and embedding services. - Sanitization: The implementation details in
SKILL.mdandreferences/implementation.mddo not demonstrate input sanitization or filtering of data before it is interpolated into prompt templates.
Audit Metadata