What problem does it solve? Managing the mj-agent owned storage stack (mj-agent-postgres for LangGraph checkpointer state and the reserved mj-agent-redis container) requires knowing the postgres init bootstrap behavior, the AsyncPostgresSaver schema, safe backup/restore procedures, and the strict isolation boundary from the mj-system business database. This Skill consolidates those operational procedures so storage issues are diagnosed without touching compose lifecycle, env setup, or the upstream business database. ## Core Features & Use Cases - Postgres init inspection: Verify the 01-bootstrap-mj-agent-memory.sh first-boot hook created the mj_agent_memory database and RW user, and fix common failures (exec permissions, stale volumes, CRLF line endings). - Schema inspection and thread queries: List LangGraph AsyncPostgresSaver tables (checkpoints, checkpoint_writes, checkpoint_blobs) and query per-thread checkpoint counts. - Backup and restore: Run pg_dump/psql via docker exec for dev debugging and PR validation, with HITL confirmation before destructive down -v operations. - Redis wiring guidance: Explain the current unwired state of mj-agent-redis and the trigger conditions (session cache, streaming buffer, rate limit) for wiring a Python client. - Use Case: When Studio reports aget_tuple not implemented or connection refused on mj-agent-postgres:5432, use this Skill to check container health, verify the async checkpointer type, and inspect the memory database. ## Quick Start Ask the agent to inspect the mj-agent-postgres container and list the LangGraph checkpointer tables in the mj_agent_memory database.
