k8s-coredns-troubleshooting

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the Kubernetes cluster, including pod logs (kubectl logs) and configuration files (kubectl get configmap). If an attacker has compromised a pod or service in the cluster, they could potentially inject malicious instructions into logs or configurations that the AI might interpret as commands.
  • Ingestion points: kubectl logs, kubectl get configmap, and kubectl exec outputs in SKILL.md and both reference files.
  • Boundary markers: No explicit delimiters are used when interpolating cluster output into the agent's context.
  • Capability inventory: The skill includes commands to modify cluster configurations (kubectl apply) and restart services (kubectl rollout restart).
  • Sanitization: No explicit sanitization or filtering of cluster data is implemented.
  • Mitigation: The skill instructions explicitly direct the agent to "ask before changing anything" and to hand proposed fixes to the user rather than executing them autonomously.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 11:22 PM