langchain-ai/deepagents

Overview

This skill helps write and execute SQL queries for tasks ranging from simple single-table lookups to complex multi-table JOINs and aggregations. It guides query planning, schema inspection, safe execution, and result formatting so you can deliver precise, auditable answers from a database.

How this skill works

The skill inspects table schemas using a schema inspection tool, drafts SQL with clear aliases and constraints, and executes read-only SELECT queries with a safe default LIMIT. For complex tasks it provides a step-by-step todo plan to map joins, choose aggregates, and validate GROUP BY and WHERE logic before execution.

When to use it

• Answer a question that requires fetching data from a database
• Compute aggregates like totals, averages, or counts across tables
• Combine related tables using JOINs to produce consolidated reports
• Validate assumptions about relationships or foreign keys before analysis
• Produce a reproducible, read-only query to share with others

Best practices

• Identify the relevant table(s) and inspect schemas before writing SQL
• Select only necessary columns; avoid SELECT * to reduce overhead
• Alias tables for readability and reference join conditions explicitly
• Apply WHERE filters before aggregation and include GROUP BY for non-aggregated columns
• Always use a LIMIT (default 5) when previewing results and never run data-modifying statements

Example use cases

• List top 5 customers by revenue with an Invoice -> Customer join and SUM aggregation
• Count active users per region using GROUP BY and proper WHERE filters
• Compare month-over-month sales by joining Orders and OrderItems and aggregating by date
• Validate foreign-key relationships by inspecting schemas and running small JOIN queries
• Extract a sample of rows from a table for quick schema verification