The Agent Skills Directory

Standard Command Execution: The skill employs official CLI tools such as az, kubectl, and helm to manage Azure Kubernetes Service resources. These tools are used for their intended purpose, such as updating cluster settings, managing NodePools, and migrating self-hosted components. All modification tasks are gated by explicit instructions to seek user approval before execution.
Indirect Prompt Injection Surface: The skill ingests untrusted data from the Kubernetes cluster environment, including resource definitions (Pods, PVCs) and cluster events. This creates a potential surface where malicious content embedded in resource names, labels, or event messages could influence agent behavior.
Ingestion points: Data enters the context via kubectl get pods, kubectl get events, and az aks show (references/nap-evidence-collection.md, references/bin-packing-evidence.md).
Boundary markers: No explicit instruction-ignoring delimiters are used for the collected Kubernetes data.
Capability inventory: The skill can perform shell-based updates via az aks update and software removal via helm uninstall (references/nap-enablement-and-rollback.md, references/self-hosted-karpenter-migration.md).
Sanitization: No visible sanitization logic for resource-derived metadata is present in the instructions.
High-Privilege Operations: The skill includes procedures for uninstalling services (helm uninstall) and modifying cluster identity or networking configurations. While appropriate for a cluster administrator's workflow, these operations require careful user oversight during execution.

aks-cost-optimization