The Agent Skills Directory

Command Execution: The skill utilizes kubectl and make to perform cluster operations, such as installing controllers and deploying models. These commands are executed within the user's configured environment and are central to the skill's setup functionality.
Credential Management: Instructions are provided for securely handling HuggingFace tokens by creating Kubernetes Secrets using temporary files that are deleted after use. This pattern avoids hardcoding sensitive information and limits the exposure of credentials in shell history.
Indirect Prompt Injection Surface: The skill processes information from the Kubernetes cluster and test responses from deployed models. While these are standard operations, they represent an external data ingestion surface.
Ingestion points: Node metadata retrieval in step-1-verify.md and model response testing in step-6-summary.md.
Boundary markers: Delimiters are not explicitly used when processing these outputs.
Capability inventory: The skill employs kubectl, make, and curl across multiple reference files for system configuration and verification.
Sanitization: Standard Kubernetes CLI outputs and API responses are processed as returned by the system.

airunway-aks-setup