The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local Node.js wrapper scripts (scripts/ha.mjs and scripts/ha-refresh-context.mjs) to execute commands and interact with the Home Assistant REST API. These scripts facilitate actions like querying entity states, calling services, and listing devices.
[DATA_EXPOSURE]: The files references/inventory-nice.md and references/inventory.md provide the agent with a detailed snapshot of the user's home automation environment. This includes entity IDs, device names, and network-related switches (e.g., FRITZ!Box Wi-Fi controls, vacuum internet access), exposing infrastructure details to the agent's context.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates external data from Home Assistant into the agent's prompt context.
Ingestion points: Entity names, areas, and states are read from references/inventory-nice.md and references/inventory.md which are generated from the Home Assistant API.
Boundary markers: Absent; the data is presented as standard markdown lists without explicit separators or instructions to ignore embedded commands.
Capability inventory: The agent can call services via scripts/ha.mjs, which allows it to change device states, run scripts, or toggle network access for local devices.
Sanitization: No evidence of sanitization or validation of the entity names or states retrieved from the API is present in the skill documentation.

homeassistant