The Agent Skills Directory

[SAFE]: The skill functions as an instructional guide for deploying inference infrastructure. It provides configuration templates for vLLM, Text Generation Inference (TGI), and NVIDIA Triton on Kubernetes.
[EXTERNAL_DOWNLOADS]: The skill references container images and model weights from established and well-known services including Hugging Face (ghcr.io), NVIDIA (nvcr.io), and Amazon (amazon/aws-cli). These are standard operational dependencies for the described architecture.
[COMMAND_EXECUTION]: The skill includes example commands for monitoring and observability, such as using kubectl to inspect GPU status and curl to fetch metrics from a local endpoint. These commands are routine administrative tasks.
[CREDENTIALS_UNSAFE]: The skill demonstrates secure practices by referencing Kubernetes Secrets for S3 access and utilizing a clearly marked placeholder (hf_...) for Hugging Face authentication tokens in its documentation examples.

coreweave-inference