The Agent Skills Directory

[Command Execution]: The skill uses standard administrative tools like kubectl and make to manage Kubernetes resources. These operations are conducted within the user's existing cluster context to deploy controllers and inference providers, which is typical for Kubernetes-based automation.
[Credential Management]: When deploying gated models that require a HuggingFace token, the skill employs a secure pattern to handle the secret. It reads the token interactively and stores it in a temporary file with restricted permissions (chmod 600) before creating a Kubernetes secret and immediately deleting the file. This approach prevents sensitive tokens from being recorded in shell history or persistent logs.
[Validation and Testing]: The inclusion of a smoke test using curl allows users to verify that the model endpoint is correctly serving requests. This is a standard validation step and does not involve exfiltration of sensitive data.
[Hardware Compatibility]: The skill includes a detailed assessment phase that checks for GPU capabilities and provides specific configuration advice (e.g., for T4 or V100 GPUs) to ensure successful model deployment without unexpected runtime errors.

airunway-aks-setup