weaviate

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In scripts/ask.py, the outsider-provided --query string is passed to QueryAgent.ask(query) and the agent then retrieves/reads Weaviate objects from the user-specified --collections, so arbitrary attacker-controlled content already stored in those collections becomes LLM-ingested free text at runtime.