mistral-core-workflow-b
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements a Retrieval-Augmented Generation (RAG) pipeline that creates a vulnerability surface for indirect prompt injection by ingesting untrusted data.
- Ingestion points: The
ragChatfunction inSKILL.md(Step 6) and the implementation inreferences/implementation.mdincorporate retrieved document content (context) and user input (query) into the model's chat messages. - Boundary markers: The implementation uses basic text headers (e.g., "Context:", "Question:") but lacks robust delimiters or explicit instructions for the model to ignore potential commands embedded within the retrieved documents.
- Capability inventory: The skill is configured with powerful tools including
Write,Edit, andBash(npm:*), which increases the potential impact if an injected instruction is successfully executed by the agent. - Sanitization: The provided code snippets do not include logic for sanitizing or validating the retrieved document content or user queries before they are included in the prompt.
Audit Metadata