The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the kdiag command-line interface and several internal Bash scripts (scripts/pod-triage.sh, scripts/connectivity-check.sh, scripts/eks-health.sh) to troubleshoot clusters. These scripts take user-provided parameters like pod names and namespaces to automate diagnostic sequences.- [COMMAND_EXECUTION]: Provides high-privilege access tools such as kdiag shell for interactive container and node-level debugging, and kdiag capture for packet sniffing. These operations involve deep interaction with the host environment and network stack.- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it retrieves and interprets data from the Kubernetes cluster that could be controlled by an external actor.
Ingestion points: The agent reads and summarizes output from kdiag logs, kdiag inspect (cluster events), and kdiag health (resource statuses).
Boundary markers: The instructions do not specify any delimiters or warnings to prevent the agent from executing instructions found within logs or event messages.
Capability inventory: The skill has access to interactive shells (kdiag shell), network captures (kdiag capture), and cluster configuration through the CLI.
Sanitization: No validation or sanitization is performed on the data fetched from the cluster before it is presented to the language model.

kdiag