tensorrt-llm
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS] (MEDIUM): The skill directs users to install tensorrt_llm and pull Docker images from nvidia/tensorrt_llm. Since NVIDIA is not in the specified trusted organization list, these are classified as unverifiable dependencies.
- [PROMPT_INJECTION] (MEDIUM): The skill configures an LLM inference server that processes untrusted data. 1. Ingestion points: User prompts via library calls and the trtllm-serve endpoint. 2. Boundary markers: Absent from the provided documentation and examples. 3. Capability inventory: High-throughput text generation and network API serving. 4. Sanitization: Absent; the implementation focuses on inference optimization.
Audit Metadata