The Agent Skills Directory

Identifier Validation: The skill instructs the agent to refuse cluster, node-pool, or namespace names that contain potentially malicious characters (e.g., quotes, pipes, or backticks), mitigating command injection risks when the agent suggests or runs commands.\n- Data Handling Best Practices: It includes a critical rule to treat all user-provided content, such as logs and YAML manifests, as data to be analyzed rather than instructions to be followed. This is an effective defense against indirect prompt injection.\n- Standard Tooling: The included diagnostic assets are shell scripts that utilize established command-line tools like kubectl, gcloud, and jq to fetch cluster metadata and logs from official Google Cloud endpoints.\n- Trusted External References: The skill points to official Google Cloud documentation and uses standard container images from trusted registries.

gke-cluster-autoscaler