What problem does it solve? Working against a relational database through the rdb-mcp MCP server requires knowing its exact contract: three tools with runtime-dependent name prefixes, string-only result values, row caps with a truncation flag, single-statement classification rules, and engine-specific differences. This Skill prevents wrong assumptions, fabricated data, and unsafe writes by teaching the correct workflow for exploring schemas, running SQL, and interpreting results. ## Core Features & Use Cases - Schema Exploration: Use list_tables and describe_table to orient against an unknown schema before writing any SQL, including catalog queries for indexes and foreign keys the tools do not expose. - Safe SQL Execution: Learn execute_sql's read-versus-write classification, one-statement-per-call rule, rejected literal forms, and the confirmation discipline required before INSERT, UPDATE, DELETE, or DDL statements. - Honest Result Interpretation: Handle string-only values, the base64 prefix for binary data, empty columns on zero-row results, and the truncated flag that signals an incomplete answer. - Use Case: A user asks "what tables exist and show me recent orders". The Skill guides the agent to list tables, describe the orders table, run a bounded SELECT with ORDER BY and LIMIT, check the truncated flag, and report results without inventing rows. ## Quick Start Ask the AI to list the tables in the connected database and describe the schema of the table you care about before running any queries.