The original title is: "AI-Assisted SQL Development with Claude Code — Rules, Skills and Agents That Enforce Conventions"

Short summary

This article introduces a three-lever framework for AI-assisted SQL development with Claude Code: rules files that enforce naming and formatting conventions at generation time, skills (slash commands) that encapsulate recurring SQL tasks reproducibly, and agents that orchestrate multi-step ETL workflows. The core argument is that AI-generated SQL only pays off when machine-enforced conventions keep code reviewable and a human still validates the business logic. It serves as the entry point to a broader series covering PL/pgSQL conventions, convention derivation, and statement structuring.

• •Three levers: rules files for conventions, skills for recurring tasks, agents for multi-step ETL workflows
• •Conventions must be machine-enforced via .claude/rules/ rather than relying on human discipline
• •Human understanding of business questions remains essential — AI can produce plausible-but-wrong SQL just as fast as correct SQL

Generated with AI, which can make mistakes.