What problem does it solve? It answers questions about the chess_teacher Postgres database—row counts, data quality, column uniqueness, and schema drift—without writing SQL by hand or risking accidental writes. ## Core Features & Use Cases - Read-only validation queries: Run row counts, filtered reads, existence checks, all-rows-match conditions, and column uniqueness checks via scripts/tools/agent_db_query.py with --json output. - Metadata-driven discovery: Domains and tables are discovered at runtime from metadata.yml files in the installed chess_teacher package, so new modules need no skill changes. - Environment-aware access: Uses Doppler configs (dev_local for local Compose Postgres) and delegates production queries to the chess-teacher-vps db-* commands when prod Postgres is firewalled. - Use Case: When a user asks "are account_id and platform_game_id unique in raw_games?", the agent runs the unique command against pipelines/ingestion and summarizes the result in plain language. ## Quick Start Ask the agent to check how many rows are in a chess_teacher database table or whether a set of columns is unique, and it will run the read-only inspection script for you.
