The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute high-privilege shell commands, including managing Docker containers and using the ha CLI for system administration. While these are necessary for the skill's primary purpose, they represent a significant capability tier.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, including Home Assistant entity states, history logs, and real-time WebSocket events. Because the agent uses this data to make decisions or generate output, and has access to powerful capabilities (shell execution, file writing), it is vulnerable to indirect prompt injection.
Ingestion points: Data is pulled from the REST API (/api/states, /api/history) and the WebSocket event stream (state_changed).
Boundary markers: None present; external data is not delimited or accompanied by instructions to ignore embedded commands.
Capability inventory: The skill can execute docker and ha commands, write to YAML configuration files on the host, and perform network operations via curl.
Sanitization: No explicit sanitization or filtering of entity states or event data is described before the data is processed by the agent.

home-assistant