The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs read operations on sensitive file paths including ~/.kube/config, /etc/kubernetes/manifests/kube-apiserver.yaml, and /etc/kubernetes/manifests/etcd.yaml. These files contain cluster authentication credentials and core configuration details. While this access is aligned with the skill's primary purpose of security auditing, it involves sensitive data exposure.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources that could be influenced by an attacker.
Ingestion points: Data enters the agent context through the output of kubectl get commands (which fetch objects from the live cluster) and through file scanning of manifest repositories via ripgrep (rg).
Boundary markers: There are no explicit instructions or delimiters defined to isolate untrusted command output from the agent's core instructions.
Capability inventory: The skill utilizes kubectl, curl, and shell access, which provides a high-capability environment for interacting with infrastructure.
Sanitization: The skill relies on standard tools like jq for parsing, but lacks specific sanitization or validation for content within Kubernetes object metadata (names, labels, annotations) which could contain malicious instructions.
[COMMAND_EXECUTION]: The skill provides extensive bash scripts and shell command batteries for performing audits. It uses tools such as kubectl, jq, curl, and ripgrep to gather configuration data and test API endpoints. The audit includes behavioral tests like checking anonymous API reachability using curl.

aegis-kubernetes-cluster-hardening