The Agent Skills Directory

Command Construction from Dynamic Input: The diagnostic workflow constructs shell commands using variables like {workload_name}, {cluster_name}, and {selector_labels}. These variables are sourced from the user prompt or the environment. This is a standard functional requirement for GKE troubleshooting, though it relies on the underlying platform to sanitize these inputs to prevent unintended command execution.
Diagnostic Data Access: The skill reads application logs, cluster events, and Kubernetes manifests to identify failure root causes. These sources may contain sensitive information such as environment variables or application stack traces. The skill handles this appropriately by presenting the analysis to the user and proposing manifest corrections via GitOps rather than applying changes directly to the cluster.
Indirect Prompt Injection Surface: The skill ingests data from external sources, specifically application logs and Kubernetes events. There is a potential surface where malicious data within these logs could attempt to influence the agent's diagnostic logic. However, the skill focuses on specific technical signatures (e.g., ExitCode 137, OOM signatures) and provides a structured workflow to mitigate this consideration.
Automated Source Control Interaction: The skill interacts with GitHub to check for existing Pull Requests and commit proposed fixes. These operations use standard Git commands to interact with trusted source control repositories, ensuring that all proposed changes are subject to human review through the GitOps process.

gke-workload-troubleshooting