DBHub skill teaches agents a progressive, token-efficient workflow for interacting with databases via DBHub MCP tools. It formalises an explore-then-query pattern: discover schemas, list tables, inspect columns, then execute precise SQL. This reduces failed queries and speeds up successful data retrieval.
Use when asked to inspect database schemas, find tables or columns, or run SQL through an MCP-connected DBHub server. It triggers on generic requests like "check the database" or explicit SQL tasks; it also applies whenever multiple databases are registered to ensure the correct tool variant is called.
search_objects and execute_sql usage, recommended detail_level values, error recovery patterns, and examples for multi-database setups.Compatible with agents that can call MCP tools (search_objects, execute_sql) such as Claude/Anthropic-based agents, Cursor, Codex/Code-Aligned agents, and other MCP-capable LLM connectors.
Pure guidance skill (no scripts) that teaches an explore-then-query workflow for DBHub MCP server tools. Well-structured with progressive disclosure pattern — start with names, drill to full detail only for tables you'll query. Good error recovery section and multi-database handling. Minor concern: no warning about destructive SQL statements (DROP/DELETE) in the execute_sql guidance.
Clean, well-written guidance skill. No scripts to test. The only security consideration is that it doesn't warn agents to avoid running destructive SQL, but the skill itself poses no risk.