Introducing Supabase Agent Skills: an open-source set of instructions that teach AI coding agents how to build on Supabase correctly.

The Supabase Blog offers insights, tutorials, and updates on Supabase, an open-source alternative to Firebase for building scalable and real-time applications. Covering topics such as database management, authentication, and serverless functions, the blog provides resources for developers looking to leverage Supabase's features to build modern web and mobile apps. Developers can learn how to use Supabase to create secure, scalable, and performant applications that meet the needs of their users.

Supabase

Supabase has released Agent Skills, an open-source set of instructions that teach AI coding agents how to use Supabase correctly. The problem: agents already know about Supabase from training data, but they make critical mistakes — skipping RLS policies, hallucinating CLI commands, creating views without security_invoker=true, and ignoring up-to-date docs. The skill is a ~100-line SKILL.md file covering four areas: documentation access (teaching agents to fetch current docs via MCP, curl, or web search), security (inline checklist of RLS and auth gotchas), tooling (CLI --help discovery and MCP troubleshooting), and schema management (direct SQL edits during dev, then formalize migrations). Testing across Claude Code and Codex showed consistent improvement when using MCP + Skill vs. baseline or MCP alone, with task completion rates rising from 42-71% to 67-88%. Key insight: the bottleneck is context, not capability — agents knew how to implement security_invoker correctly when the skill was loaded, they just didn't know when to apply it.

AI Agents Know About Supabase. They Don't Always Use It Right.

Great move right there by supabase! Amazing!

This could become a broader pattern: instead of just building smarter models, we standardize “operational knowledge layers” for tools. In a way, we’re moving from teaching AI what to know - to teaching it how to behave correctly in context.