The Agent Skills Directory

[COMMAND_EXECUTION]: The skill primarily operates by instructing the agent to use a local command-line interface tool (xgg). This tool is used to interact with the local gateway, manage device specifications, and compile automation rules. This is a core part of the skill's intended functionality for power users of the Xiaomi IoT ecosystem.
[REMOTE_CODE_EXECUTION]: The skill documents a workflow for migrating or cloning automation rules that involves generating shell scripts (e.g., replay.sh, clone.sh) using the xgg rule import command. While these scripts are executed locally, the skill provides extensive safety instructions, including mandatory preflight validation scripts, manual review requirements, and the use of snapshots to ensure integrity before and after execution.
[DATA_EXFILTRATION]: The skill processes sensitive household data, including device events and habit logs. However, the analysis shows that the data is stored in a private local directory (.xgg-private) and all network interactions are directed to a user-defined local gateway URL (--base-url). The instructions explicitly warn against including authentication codes in logs or shared repositories.
[SAFE]: The skill incorporates proactive security and stability measures, such as mandatory --spec-aware validation, strict linting, and a 'fail-closed' approach when device specifications or variable types do not match expectations. It also includes privacy guidelines for 'habit learning' to exclude sensitive devices like cameras or door locks unless explicitly authorized.

xgg-rule-authoring