Skills that make AI better at Next.js and Supabase apps
What an agent skill is, seven real skills for databases, design, testing, review and debugging, and how to install them
You ask an AI coding tool to add a table to your Supabase project and it works, but it forgets row-level security. You ask for a new page and it looks like every other AI-made page. You report a bug and it patches the symptom instead of the cause. None of this means the model is weak. It means the model does not know your standards, and you are repeating the same corrections in every session.
Skills fix that. This post explains what a skill is, then spotlights seven real skills from the AI Khazna catalogue that make an AI agent noticeably better at building Next.js and Supabase apps.
What a skill is
Anthropic describes Agent Skills as modular capabilities that package instructions, metadata and optional resources such as scripts and templates, which Claude uses automatically when they are relevant. In practice a skill is a folder with a SKILL.md file. At the top of that file sit a name and a description that say what the skill does and when to use it. Below them are the instructions.
The clever part is how skills load. The agent sees only the short name and description of each installed skill until a task matches one. Then it reads the full instructions, and it opens any extra reference files only if the task needs them. So you can install several skills without filling the model's context with text it does not need.
A skill is different from a rules file. Your CLAUDE.md or Cursor rules describe your project and apply to every session. A skill carries know-how for one kind of job, such as writing Postgres migrations or debugging, and switches on when that job comes up. Our guide to writing a rules file for Claude Code and Cursor covers the first; this post covers the second.
Seven skills for Next.js and Supabase apps
Each of these is listed in the AI Khazna catalogue, and the descriptions below come from their listings.
1. Supabase
Published by Supabase, Supabase is meant for any task that involves Supabase: Database, Auth, Edge Functions, Realtime, Storage, Vectors, Cron and Queues, plus client libraries such as supabase-js and @supabase/ssr, the CLI, migrations and security audits. Its guidance is practical: check current docs because Supabase changes often, run a test query after every change, enable row-level security on every table in exposed schemas, never use user metadata for authorisation, and keep service-role keys out of public clients.
Use it when: you add sign-in, wire Supabase into Next.js server components, or create tables.
2. Supabase Postgres Best Practices
Supabase Postgres Best Practices is described as Postgres best practices maintained by Supabase, for Postgres running anywhere. It covers table and column design, migrations, RLS policies and the tests that verify them, indexes, triggers, functions, queues, scheduled jobs and vector search. Rules are grouped into eight categories ordered by priority, with query performance, connection management, and security and RLS at the top, and each rule shows incorrect and correct SQL.
Use it when: the agent writes anything that lives in the database, especially migrations and policies.
3. Frontend Design
From Anthropic, Frontend Design gives guidance for distinctive, intentional visual design when building new UI or reshaping existing UI. It asks the agent to understand the product and audience first, choose typography deliberately, and avoid common signs of templated AI design such as identical rounded cards and decorative labels that carry no information.
Use it when: you build landing pages, dashboards or any screen users will judge your product by.
4. Test Driven Development
Test Driven Development, by obra, is for implementing any feature or bugfix before writing implementation code. It enforces red-green-refactor: write one failing test, confirm it fails for the right reason, write the minimum code to pass, then run the full suite. Its core idea is simple: if you did not watch the test fail, you do not know it tests the right thing.
Use it when: you touch payments, permissions or any logic you cannot afford to break silently.
5. Webapp Testing
Anthropic's Webapp Testing is a toolkit for interacting with and testing local web applications using Playwright. It helps the agent verify frontend behaviour, capture screenshots and read browser logs, and it includes a helper that starts your servers before a test and stops them after. For dynamic apps it recommends waiting until the page has settled before inspecting it.
Use it when: you want the agent to click through sign-up, forms or checkout in a real browser instead of assuming they work.
6. Code Review
Code Review, from Anthropic, reviews code changes for security, performance and correctness. Its security checks include SQL injection, XSS and CSRF, authentication and authorisation flaws, secrets in code, path traversal and SSRF. It also flags performance problems such as N+1 queries and missing indexes, and ends with a verdict: approve, request changes or needs discussion.
Use it when: before you merge a feature or launch. Pair it with our list of common AI security mistakes.
7. Systematic Debugging
Systematic Debugging, also by obra, is for any bug, test failure or unexpected behaviour, before proposing fixes. Its rule is no fixes without root-cause investigation first. It walks the agent through four phases: investigate and reproduce, compare with working code, test one hypothesis at a time, then fix the cause with a failing test first.
Use it when: the agent has tried two quick fixes and the bug is still there.
How to install a skill
The exact steps depend on your tool, but the idea is the same everywhere: put the skill folder where the agent looks for skills.
- Open the skill's page in the catalogue. Each listing shows its source repository and an install command, often in the form
npx skills add owner/repo. - Or copy the folder by hand. In Claude Code, Anthropic's documentation says personal skills go in
~/.claude/skills/and project skills in.claude/skills/inside your repository. Each skill is a folder containing itsSKILL.md. - Prefer project skills for team work. Committing them to the repository means everyone, and every agent session, gets the same standards.
- Read before you install. A skill is instructions and sometimes scripts that your agent will follow. Open the
SKILL.md, check what any scripts do, and install only from sources you trust. - Test it. Give the agent a small task that should trigger the skill, such as adding a table, and check that it follows the new rules.
Many listings note compatibility beyond Claude Code, including Cursor and other agents, so check the listing for your tool.
A sensible starter set
If you are starting a Next.js and Supabase project today, begin with three: Supabase and Supabase Postgres Best Practices for the data layer, and Systematic Debugging for the moments things go wrong. Add Frontend Design when you reach the interface, Test Driven Development and Webapp Testing when real users depend on the app, and Code Review before every launch. Fewer, well-chosen skills beat a long list the agent rarely needs.
Next step: browse the skills catalogue for your stack, or start from the Vibe Coding hub to set up a project with the right skills from day one.