The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the exec tool to run gbrain CLI commands for detecting database engine types, initializing environments, and performing data migrations.
[PRIVILEGE_ESCALATION]: The agent is instructed to use flags such as --allow-docker and --allow-create-db, which grant the ability to manage host containers and execute administrative database operations like creating new databases.
[PERSISTENCE]: The skill manages a gbrain-postgres Docker container that is explicitly described as surviving reboots, providing a persistent database service on the host system.
[INDIRECT_PROMPT_INJECTION]: The skill ingests JSON output from gbrain engine status and gbrain migrate to make logical decisions, which presents an attack surface if the tool output can be influenced by untrusted data. * Ingestion points: Output from gbrain engine status --json (Step 1) and gbrain migrate --status --json (Step 4). * Boundary markers: No boundary delimiters are specified in the instructions for processing tool output. * Capability inventory: The skill uses the exec tool with the ability to modify system-level infrastructure and manage persistent containers. * Sanitization: Instructions define branching based on specific JSON keys (e.g., effective_engine, state) but do not mention sanitization for other data fields.
[CREDENTIALS_UNSAFE]: The skill interacts with high-value environment variables, including SUPABASE_ACCESS_TOKEN, SUPABASE_DB_PASSWORD, and PGPASSWORD. * Evidence: The instructions in Step 2 and Step 3 describe discovery and connection ladders that utilize these environment variables to authenticate with Supabase and local Postgres instances.

postgres-adopt