The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python and shell scripts to execute various 'kubectl' commands, including pod status checks, log retrieval, and disruptive operations like pod deletion or rollout restarts.
Evidence: 'scripts/pod_diagnostics.py' wraps 'kubectl' calls to describe pods and fetch logs; 'SKILL.md' documentation includes emergency operations such as 'kubectl delete pod ... --force' and 'kubectl rollout undo'.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves and displays pod logs, events, and resource configurations which are controlled by external applications and could contain malicious instructions designed to manipulate the agent.
Ingestion points: 'scripts/pod_diagnostics.py' captures container and init container logs; 'scripts/cluster_health.sh' captures cluster-wide event messages.
Boundary markers: None identified; the diagnostic output does not explicitly delimit untrusted content from the rest of the agent's context.
Capability inventory: The skill can execute shell commands and modify Kubernetes workloads.
Sanitization: No filtering or escaping is applied to ingested log data or event strings before they are printed.
[DATA_EXFILTRATION]: The network diagnostic script accesses sensitive Kubernetes service account tokens from within target pods to verify API server reachability.
Evidence: 'scripts/network_debug.sh' executes 'cat /var/run/secrets/kubernetes.io/serviceaccount/token' via 'kubectl exec' to authenticate its connectivity probe.
[DYNAMIC_EXECUTION]: Shell scripts in the skill execute command strings dynamically using the shell's evaluation capabilities.
Evidence: 'scripts/cluster_health.sh' and 'scripts/network_debug.sh' use 'bash -c' to run piped command sequences for event filtering and component health checks.

k8s-debug