Title: Introducing: Postgres Best Practices

Supabase has released Agent Skills for Postgres Best Practices—a collection of 30 machine-readable rules that guide AI agents to write correct, high-quality database code. AI coding agents are generally capable but often lack understanding of complex systems like Postgres, leading to queries that work but cause performance issues, missed security policies, or database locking problems in production.

The skills are organized into eight priority categories, from critical concerns like query performance and security to medium-priority items like monitoring. Each rule follows a standardized format with explanations, code examples of incorrect versus correct approaches, and impact ratings. This approach treats best practices as explicit guidance rather than relying on agent training data.

The framework builds on the open Agent Skills standard developed by Anthropic and now adopted across the industry. Vercel and Cloudflare have used similar approaches to package years of specialized knowledge into machine-readable instructions. Supabase's effort applies this same model to database expertise drawn from supporting hundreds of thousands of projects.

These skills work alongside Supabase's Model Context Protocol (MCP) server, which gives agents the ability to execute database operations. While the MCP server is the capability layer, the best practices are the judgment layer—agents can be given guardrails before they execute potentially damaging changes. Installation is straightforward through npm or Claude Code plugins, and the community is invited to contribute additional rules based on real-world Postgres gotchas.