The Agent Skills Directory

[SAFE]: Credential and Configuration Security: The skill manages sensitive Home Assistant credentials (base URL and access token) through the platform's configuration settings rather than hardcoding them, ensuring that user secrets are handled securely.
[SAFE]: File System Integrity: The ha_get_camera_snapshot tool includes logic to save images exclusively to designated workspace directories. It explicitly prevents writing to the sibling_journal directory, a shared communication folder, thereby protecting the integrity of text-based logs and preventing accidental file sprawl.
[SAFE]: Input Sanitization: The coerceServiceData utility function ensures that service parameters provided by the agent are correctly parsed from various formats (JSON or YAML-like strings) into the structured objects required by the Home Assistant API, reducing the risk of malformed requests.
[SAFE]: Defensive Tooling Design: The skill implements proactive checks to prevent common agent failures, such as service hallucination or incorrect entity identification. For example, it includes a redirection mechanism for PTZ camera controls that guides the agent to valid Home Assistant entities and services when it attempts to use non-existent ones.
[SAFE]: Context-Aware Safety Instructions: The documentation provided in SKILL.md includes explicit directives to the agent, such as the requirement to discover entity IDs before acting and the need for user confirmation before executing destructive actions (e.g., opening garage doors).

home-assistant