Working with AI assistants
AI coding assistants write better code when they know a project’s rules. The boilerplate ships those rules as plain Markdown in .skills/, so Claude Code, Cursor, Copilot, Codex or Aider can follow the same conventions your team does.
The guides
Section titled “The guides”| File | Covers |
|---|---|
.skills/api-hook-creation.md |
Adding API hooks: folder layout, react-query-kit patterns, what the shared client already handles, pagination, mutations |
.skills/ui-creation.md |
Building screens and components: where files go, theme tokens, text styles, scaling, ScreenWrapper rules, a review checklist |
.skills/localization.md |
Adding and using translations, typed keys, language switching and RTL |
ui-creation.md is rewritten by the CLI for the styling option you choose, so the assistant gets Unistyles, StyleSheet or Tailwind instructions to match your app.
How to use them
Section titled “How to use them”Point your assistant at the relevant file when you ask for something:
Read
.skills/api-hook-creation.md, then add a hook that fetches a user’s orders fromGET users/:id/orders.
Or add them to your tool’s project rules (Cursor rules, Copilot instructions, CLAUDE.md or AGENTS.md) so they’re always loaded.
Why it works well with AI
Section titled “Why it works well with AI”- One place for everything. Each feature is a single folder with a clear public API, so an assistant can change one feature without touching others.
- Guard rails catch mistakes. Typed routes, typed translation keys, strict TypeScript, ESLint rules and Zod-validated env vars turn most AI mistakes into immediate errors instead of runtime bugs.
- Fail-loud tooling.
yarn lint && yarn type-check && yarn testis a quick, reliable check after any AI-generated change. AGENTS.mdat the root reminds assistants to read the docs for the exact Expo SDK version in use, since Expo APIs change between releases.