tensorrt-llm

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

• Third-party content exposure detected (high risk: 0.80). The skill's server/docs explicitly state "automatic model download and compilation" and link to public model hubs (e.g., https://huggingface.co/models and GitHub), so it fetches and loads untrusted, user-hosted models from third-party sites which the agent then uses for inference.