The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates shell commands for kubectl, ssh, and git to audit and manage remote environments. These commands incorporate parameters such as namespaces, pod names, and host aliases retrieved from a local .deployment_info configuration file.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks (Category 8) as it processes untrusted data from Home Assistant runtime logs and repository files (YAML, Python scripts).
Ingestion points: Runtime logs via kubectl logs and ssh, repository configuration files (*.yaml, pyscript/*.py), and the .deployment_info metadata file.
Boundary markers: The instructions explicitly require the agent to 'report redacted evidence' and 'obtain explicit approval' before taking action, which serves as a procedural boundary.
Capability inventory: The agent can execute local Python scripts, git commands, kubectl exec/cp operations, and one-shot ssh commands.
Sanitization: The skill contains explicit prohibitions against reading or transmitting sensitive files like secrets.yaml, private keys, or authentication tokens.

home-assistant-config-ops