The Agent Skills Directory

[SAFE]: The skill performs legitimate security auditing functions for Kubernetes environments. All instructions and scripts follow standard operational procedures for cloud security assessments and access control validation.
[DATA_EXFILTRATION]: The skill accesses Kubernetes configuration (via the KUBECONFIG environment variable or the standard ~/.kube/config path) and cluster metadata to evaluate permissions. This access is a requirement for auditing the cluster's security state. The results are stored in a local file (k8s_rbac_audit.json), and no evidence of external data exfiltration was found.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Kubernetes API, such as resource names and specifications, which could be manipulated by an attacker who already possesses cluster write permissions. The risk is considered minimal as it only affects the resulting security report.
Ingestion points: Metadata from ClusterRoles, Roles, RoleBindings, and Pod security contexts read through kubectl and the kubernetes Python library.
Boundary markers: None present in the data processing scripts.
Capability inventory: Read access to Kubernetes API; write access to local report files.
Sanitization: The skill performs standard JSON processing but does not apply specific sanitization to Kubernetes resource names or labels before logging them to the console or report file.

auditing-kubernetes-cluster-rbac