The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches the kube-bench binary and Kubernetes configuration manifests from Aqua Security's official GitHub repository.\n- [COMMAND_EXECUTION]: The scripts/agent.py file uses the subprocess module to execute kube-bench and kubectl commands. These calls use argument lists rather than shell strings, which follows security best practices for preventing shell injection.\n- [PRIVILEGE_ESCALATION]: The skill requires cluster-admin access to the Kubernetes cluster and sudo privileges for local installation. The provided Kubernetes Job and CronJob manifests use high-privilege settings (e.g., hostPID: true, host path mounts for /etc/kubernetes and /var/lib/kubelet) which are necessary for the tool to inspect node-level configuration files as per the CIS Benchmark requirements.\n- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted data from the cluster environment and tool outputs.\n
Ingestion points: Data enters the skill via json.loads in scripts/agent.py (parsing kube-bench and kubectl output) and scripts/process.py (parsing saved scan results).\n
Boundary markers: The skill relies on the structured nature of JSON but does not implement specific delimiters or 'ignore' instructions for data that might contain natural language strings (e.g., pod names or remediation text).\n
Capability inventory: The skill can execute subprocesses (kube-bench, kubectl) and write report files to the filesystem.\n
Sanitization: No specific sanitization or escaping is performed on the data before it is included in reports or processed by the agent.

performing-kubernetes-cis-benchmark-with-kube-bench