tensorrt-llm

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

• Third-party content exposure detected (high risk: 0.70). The skill's runtime (SKILL.md and serving docs) explicitly states trtllm-serve performs "automatic model download and compilation" and points to public model repositories (e.g., HuggingFace), so it clearly fetches untrusted, user-hosted models from the open web which the agent ingests and which can materially alter its outputs/behavior.