review-security-k8s-agents-firewall

Auditing Kubernetes network policies for AI agent data exfiltration involves validating that containerized sandboxes enforce default-deny egress rules and restrict access to sensitive internal APIs. This skill automates that review to identify unauthorized external communication vectors.

Isolating AI agent pods from cloud metadata services in Kubernetes requires network policies that explicitly block egress to metadata endpoints. This skill audits your sandbox configurations to ensure autonomous agents cannot reach underlying cloud metadata services.

Yes, you can use this skill to check if your agent sandbox restricts access to the Kubernetes control plane. It audits network policies to ensure agent pods are restricted from accessing Kubernetes control plane endpoints and internal cluster APIs.

You need a default-deny egress policy for autonomous infrastructure agents to prevent unauthorized lateral movement and data exfiltration. This skill validates that agent sandboxes enforce these strict policies to maintain least-privilege networking compliance.

This skill works by auditing both standard Kubernetes network policies and firewall rules. It targets the identification of egress exfiltration vectors and internal API exposure within containerized sandboxes to secure AI agent execution environments.