The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes gbrain advisor --json to gather diagnostic information about the environment.
[COMMAND_EXECUTION]: The skill provides instructions to execute specific fix commands (e.g., gbrain upgrade, gbrain apply-migrations) but explicitly mandates that the agent must ask for user permission before running them.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from the gbrain advisor tool, which could theoretically contain malicious instructions if the tool's data source is compromised. However, the risk is mitigated by the skill's instruction to only summarize findings and the requirement for user approval for any resulting actions.
Ingestion points: Output of gbrain advisor --json (SKILL.md)
Boundary markers: Present; the skill instructs the agent to read and summarize findings rather than directly obeying them.
Capability inventory: Shell command execution via gbrain CLI (SKILL.md)
Sanitization: Not explicitly mentioned, but the workflow requires human-in-the-loop confirmation for all mutations.

gbrain-advisor