Agent skill · NVIDIA
tao-run-on-kubernetes
Kubernetes execution platform — submits TAO container jobs as single-pod k8s Jobs with NVIDIA GPU scheduling.
What it needs
About 8k tokens when loaded.
What this skill does
Kubernetes Standalone install? If this session was not initialized by the TAO skill bank plugin, run the tao-setup skill first (host preflight, credentials, cross-skill discovery). Submits TAO container jobs as Kubernetes Jobs. Works on any cluster reachable via kubeconfig (EKS / GKE / AKS / on-prem) or in-cluster service account (when the SDK runs inside a pod). Single-pod by default; opt into multi-node distributed training via numnodes > 1 (uses Indexed Job + headless Service, see Multi-node training below). Preflight Four checks: GPU host runtime ready, SDK installed, cluster reachable, GPU Operator/device plugin present. For self-managed nodes, if the selected model declares runtimerequirements.gpuhost, pass the corresponding --min--version flags to check and install commands. On managed clusters, require the provider node image or GPU Operator policy to satisfy that model profile. The GPU node runtime check is mandatory for self-managed nodes. For managed clusters where the client is not running on a GPU worker, verify the provider node image or GPU Operator policy and set TAOK8SSKIPNODERUNTIMECHECK=1 instead of running the installer on the client. The final GPU capacity check is a warning rather than a hard fail — kubectl isn't always installed. The SDK does a hard guard inside KubernetesSDK.createjob() that uses the kubernetes Python client to verify GPU capacity before submitting. Credentials & configuration Kubeconfig (one of): ~/.kube/config — default discovery path $KUBECONFIG — alternate path In-cluster service account — used when running inside a pod (no kubeconfig needed) TAOK8SNAMESPACE (optional): default namespace for Job submission. Defaults to default. TAOK8SCONTEXT (optional): kubeconfig context name to switch clusters. NGCKEY (optional): for nvcr.io image pulls. If you've pre-created an image-pull secret in the target namespace, pass its name to createjob via the imagepullsecret argument. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills NVIDIA/tao-run-on-kubernetes