The Agent Skills Directory

[DATA_EXFILTRATION]: The skill utilizes the kubeconfig_view() tool, which accesses and displays the content of the local Kubernetes configuration file. This file typically contains sensitive information, including cluster authentication tokens, client certificates, and context metadata. Exposing these credentials to the model's context increases the risk of accidental disclosure or unauthorized access to the cluster.
[INDIRECT_PROMPT_INJECTION]: The skill features a vulnerability surface where it ingests untrusted data from the Kubernetes cluster which could be influenced by malicious actors.
Ingestion points: The skill reads external data using kubectl_logs, kubectl_get(resource_type="events"), and kubectl_describe. These sources can contain attacker-controlled strings (e.g., log lines, event messages, or resource annotations).
Boundary markers: No delimiters or specific instructions (e.g., "ignore embedded instructions") are present to separate untrusted data from the primary prompt.
Capability inventory: The skill possesses extensive cluster-wide read capabilities and resource comparison tools via the kubectl-mcp-server toolset.
Sanitization: There is no evidence of sanitization or filtering applied to the logs or resource descriptions before they are processed by the agent.

k8s-diagnostics