The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the exec tool to run the gbrain CLI for diagnosing and repairing database issues. The instructions specify hardcoded command sequences (e.g., gbrain db-repair --json, gbrain db-repair --yes) to perform these tasks, ensuring the agent does not generate arbitrary shell commands.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to respond to error markers like GBRAIN_DB_ACCESS <reason> found in tool outputs. It implements a critical security boundary by instructing the agent to never parse or execute data from these markers, treating them only as triggers for its hardcoded diagnostic logic.
Ingestion points: Untrusted data enters the agent context via gbrain CLI stderr or MCP tool result envelopes containing the GBRAIN_DB_ACCESS marker.
Boundary markers: None present in the raw data, but the skill instructions act as a logical boundary by defining a strict protocol for response.
Capability inventory: The skill has access to the exec tool, allowing it to perform shell operations.
Sanitization: The skill enforces sanitization by mandating the use of fixed command strings and ignoring all dynamic content within the trigger markers during command construction.
[DATA_EXPOSURE]: The skill instructions acknowledge the sensitivity of database connection strings and specify that the gbrain CLI tool output is pre-redacted by the vendor, preventing accidental leakage of credentials when the agent reports results to the operator.

db-repair