Write an AGENTS.md file in your repository root with build, test and style rules, because many coding agents read it. Plan before you build with GitHub Spec Kit or Kiro specs, and follow each builder's own prompting guide: Lovable and Bolt both advise small, specific prompts and planning first. Use Supabase's agent skills for safer database code.
AI coding goes wrong when the model guesses. A rules file tells it how your project works, a spec tells it what to build, and a good prompt tells it the next small step. These free resources cover all three.
The 12 picks
Swipe or use the arrows.
Free, MIT licenceAGENTS.md
A simple open format: a Markdown file at your repo root with build, test and style instructions. Free and MIT licensed, and the site says Codex, Jules, Aider, Cursor, VS Code, Gemini CLI and GitHub Copilot read it. Support elsewhere varies.
Free, MIT licenceGitHub Spec Kit
GitHub's open-source toolkit for spec-driven development: you write a spec and a plan, then the agent implements them. Free under MIT, needs Python tooling, and its commands change often, so follow the current README.
Free tier, 50 creditsKiro specs
Kiro, an AI editor, turns an idea into requirements, design and task files you can track. The IDE is proprietary, with a free tier of 50 credits a month. Check its models and any student offer on the pricing page.
Free, MIT licenceBMAD-METHOD
An open framework of role-based prompts and workflows (analyst, product manager, architect, developer) for planning and building software with AI. Free under MIT. Heavy for small apps, and version 7 is in preview.
Free to readLovable prompting guide
Lovable's own advice: plan first, ask the AI clarifying questions, prompt by component rather than whole page, use real content and iterate with version history. Specific to Lovable's workflow and credits.
Free to readBolt prompting guide
Bolt's help page on prompting well: start with the architecture, clear context regularly, add features one at a time and keep prompts small and specific.
Free to readv0 prompting guides
Vercel's guide to prompting v0 by product surface, context of use and constraints. The blog post is older, so check the current v0 documentation for new features.
Free to readAnthropic prompting best practices
Anthropic's official guide to prompting current Claude models: clarity, examples, structure, tool use and agentic work. It replaced the old copy-paste prompt library, which now redirects here.
Free, MIT licenceSupabase agent skills
Official instruction packs an agent loads for Supabase Auth, storage, realtime, row-level security and Postgres practice. Free under MIT and still pre-1.0. They guide the agent but do not replace reviewing your policies.
Free, mixed licencesAnthropic Skills repository
Anthropic's example Agent Skills that show how to package instructions for Claude. Many are Apache-2.0, and the document skills are source-available. The README calls them demonstrations, so test before relying on them.
Free, CC0awesome-cursorrules
A community collection of rules files per framework, released under CC0. Quality is uneven and many predate AGENTS.md, so read a rule before pasting it.
Free to readPublished Claude system prompts
Anthropic publishes the system prompts used on claude.ai and its apps in its release notes. This is the legitimate way to study how a real product instructs a model, instead of leaked collections.
Compare in a table
Show the full comparison table
| Tool | Best for | Free plan |
|---|---|---|
| AGENTS.md | One instruction file for many agents | Free, MIT licence |
| GitHub Spec Kit | Planning before letting AI code | Free, MIT licence |
| Kiro specs | A spec-first workflow | Free tier, 50 credits |
| BMAD-METHOD | Structured multi-step projects | Free, MIT licence |
| Lovable prompting guide | Getting more from Lovable credits | Free to read |
| Bolt prompting guide | Prompting Bolt effectively | Free to read |
| v0 prompting guides | Prompting v0 for UI | Free to read |
| Anthropic prompting best practices | Clear prompts for Claude | Free to read |
| Supabase agent skills | Safer Supabase code | Free, MIT licence |
| Anthropic Skills repository | Learning to write skills | Free, mixed licences |
| awesome-cursorrules | Starter rules per framework | Free, CC0 |
| Published Claude system prompts | Studying a real system prompt | Free to read |
Get started this week
- Create AGENTS.md with how to run, test and build your project.
- Write a one-page spec: users, three features and what is out of scope.
- Give the agent one small task from the spec and review the diff.
How to choose
- Put the commands to build, test and run in AGENTS.md first; that is what agents need most.
- Write a short spec before the first prompt, even a page, and keep it in the repository.
- Prompt one small change at a time and read the diff before the next one.
- Use rules from community lists as a starting point and delete what you do not need.
Free plans change. Limits above were checked on 3 Nov 2026; confirm on each tool's pricing page before relying on them.
Frequently asked questions
What is AGENTS.md?
A plain Markdown file at the root of a repository that tells coding agents how to build, test and style the project. It is an open format and many agents read it.
What is spec-driven development?
Writing down what you want, requirements and a plan, before the AI writes code, then having the agent implement from that spec. It reduces guessing and makes changes easier to review.
Do I need a rules file for a small app?
A short one still helps. Ten lines covering how to run the app, how to test it and what never to do, such as committing secrets, saves many wasted prompts.
Where can I find good prompts for AI coding?
Start with the official guides from Lovable, Bolt, v0 and Anthropic. They are written for those exact tools and are more reliable than copied prompt lists.
Learn this with mentors
Programmes for builders and engineers with admissions open: live classes, mentors and project work.
