kubernetes-attack
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPRIVILEGE_ESCALATIONPERSISTENCEDYNAMIC_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill provides instructions for identifying and extracting sensitive bearer credentials, including Kubernetes ServiceAccount tokens, kubelet client certificates, and cloud IAM tokens (EKS IRSA, GKE Workload Identity, AKS Managed Identity).
- [REMOTE_CODE_EXECUTION]: The skill contains instructions for installing third-party tools via piped shell execution (
curl | bash). Examples include the installation of Kubescape and the K3s cluster lab environment. It also documents patterns for executing remote payloads as part of offensive persistence and lateral movement techniques. - [EXTERNAL_DOWNLOADS]: The skill downloads numerous binary tools and scripts from external repositories and services, including
kubectl,CDK,peirates,kube-hunter,kube-bench,stratus-red-team, andkubeletctl. - [DATA_EXFILTRATION]: The playbooks describe explicit methods for exfiltrating sensitive data such as ServiceAccount tokens, cloud credentials, and environment variables. This includes specific techniques for DNS exfiltration (tunneling data through subdomains) and sending data to remote attacker-controlled endpoints.
- [PRIVILEGE_ESCALATION]: The skill is primarily focused on privilege escalation techniques within Kubernetes environments. It provides detailed manifests and commands for escaping containers to the host node (using
privileged: true,hostPath: /, andnsenter) and escalating RBAC permissions tocluster-adminvia verbs likebindandescalate. - [PERSISTENCE]: The skill provides detailed payloads and manifests for establishing long-term persistence in a cluster, such as malicious DaemonSets, CronJobs for reverse shells, and MutatingWebhook injections that can survive workload reinstalls.
- [COMMAND_EXECUTION]: The skill relies extensively on executing powerful shell commands for reconnaissance, exploitation, and lateral movement, including the direct manipulation of the
kubeletandetcdAPIs to bypass standard authentication and authorization controls. - [DYNAMIC_EXECUTION]: The skill describes the use of dynamic execution primitives to interact with host-level namespaces and processes from within containers. It also includes instructions for the runtime compilation and execution of kernel exploit PoCs (such as CVE-2022-0185 and CVE-2024-1086).
Audit Metadata