The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local commands using the gbrain CLI tool (e.g., gbrain advisor --json) to retrieve system state and apply fixes. These operations are limited to the gbrain environment and follow a user-approval workflow.
[SAFE]: The skill implements strong safety boundaries, including a 'read-only' contract for its primary diagnostic tool and explicit instructions to never execute mutations (via --apply or specific fix commands) without the user's direct consent.
[DATA_EXPOSURE]: The skill accesses local brain metadata to provide coaching. There are no patterns suggesting data exfiltration or transmission to external domains.

gbrain-advisor