The Agent Skills Directory

[SAFE]: The skill provides instructions for installing official packages from well-known sources (NVIDIA and the Python Package Index). Dependencies include tensorrt-llm and torch.
[SAFE]: Technical documentation includes security-conscious implementation details, such as utilizing environment variables for API keys and configuring Kubernetes NetworkPolicies to restrict access to the inference service.
[SAFE]: External references are directed exclusively to established technology providers and open-source repositories, including NVIDIA's GitHub organization and the HuggingFace model hub.
[SAFE]: No evidence of obfuscation, unauthorized data exfiltration, privilege escalation, or persistence mechanisms was found across the skill's instructions or supporting guides.
[SAFE]: The instructions incorporate human-in-the-loop safeguards, advising the agent to confirm resource costs and hardware requirements before execution.

tensorrt-llm