tensorrt-llm

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

• Third-party content exposure detected (high risk: 0.95). The required runtime workflow for trtllm-serve ingests the caller-provided messages[].content / prompt from the OpenAI-compatible HTTP request body into the model’s generation context, and that free-form text is outsider-authored (any client other than the operating user).