The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes system commands (kubectl, etcdctl) to inspect cluster configuration and etcd security. This is the primary function of the skill and is implemented using safe subprocess calls with list-based arguments to avoid shell injection vulnerabilities.
Evidence: Found in scripts/agent.py (e.g., check_etcd_encryption, check_etcd_access) and scripts/process.py (e.g., run_command, check_etcd_tls).
[INDIRECT_PROMPT_INJECTION]: The skill represents an indirect prompt injection surface as it ingests and processes resource names and arguments from the Kubernetes API which could be manipulated by a malicious actor in the cluster.
Ingestion points: kubectl JSON output processing in scripts/agent.py and scripts/process.py.
Boundary markers: Absent.
Capability inventory: List-based subprocess.run calls for kubectl and etcdctl binaries.
Sanitization: Uses standard JSON parsing and string inclusion checks for configuration flags.

performing-kubernetes-etcd-security-assessment