The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local CLI commands including gbrain advisor, gbrain backup status, and gbrain advisor --apply to inspect and manage the environment.
[PERSISTENCE]: The skill provides instructions for setting up a weekly scheduled task via a cron-scheduler. While this is the core intended functionality for a recurring advisor, it establishes a persistent automated workflow for the agent.
[INDIRECT_PROMPT_INJECTION]: The skill processes JSON output from the gbrain advisor command, which originates from the brain database. This creates a surface where malicious content stored in the database (e.g., job names, findings) could influence the agent's summary or decision-making.
Ingestion points: Data from gbrain advisor --json (SKILL.md).
Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded within the JSON data.
Capability inventory: The skill can execute file system checks and run fix commands via the gbrain CLI.
Sanitization: No sanitization or escaping mechanisms are defined for processing the external tool output before it is summarized for the user.
[DYNAMIC_EXECUTION]: The skill dynamically constructs shell commands by interpolating a dispatch_id from tool output into the command gbrain advisor --apply <dispatch_id>. This creates a risk of command injection if the dispatch_id value is malicious, although the instructions require explicit user confirmation before execution.

gbrain-advisor