agentic-mobile-kit

One setup for AI-assisted development of React Native apps, bare or Expo. Every AI tool on the team reads the same rules, project knowledge and workflows. Git hooks and CI check the work, whichever person or tool wrote it. A task started in one tool, or on one machine, carries on in another.

Tested with Claude Code, Codex and Antigravity. --tools also sets up Cursor, OpenCode and Kiro.

Install

1. In your app's root folder, on a clean git branch, run:

   npx agentic-mobile-kit

   It asks what to do, one step at a time: install with the recommended setup (Claude Code, Codex and Antigravity, no CI), or choose the AI tools and CI. There's nothing to npm install: the kit copies its files into the app and isn't a dependency. If you ran npm i agentic-mobile-kit, commit or undo that first, because the kit needs a clean working tree.

   Starting a new app? Run it in an empty folder. It offers to create a React Native app with the React Native CLI (with its iOS pods) or an Expo app with create-expo-app, commits it, and installs the kit in it.

2. Run it again and choose Check this project and machine. It lists each problem with its fix.

3. In your AI tool, run m-onboard. It reads the project, drafts docs/amk/product.md, tech.md, structure.md, conventions.md, glossary.md and design-system.md, asks what the code can't show, and writes them once you approve. Then commit and open a PR.

Each choice prints the command it runs, for scripts and CI: init (with --ci github, --ci codemagic or --tools claude,codex,antigravity, and --new bare or --new expo to create the app first), sync, doctor and uninstall. --help lists every option.

init never overwrites your files, is safe to run again, and can be undone with git. The team guide lists everything it changes.

How work flows

| You want to | Run | What happens | |---|---|---| | Describe the app for agents | m-onboard | Reads the whole project, drafts the product, tech, structure, conventions, glossary and design-system files in docs/amk/, asks what the code can't show, and writes them once you approve. Run it after installing, and again after big changes. | | Think a problem through | m-explore | The agent reads the code, explains how it works today and lays out options with trade-offs. It never edits code. | | Agree on an idea before building it | m-prototype | Concepts as clickable HTML mocks when how it should work is still open, then real screens with mock data on a throwaway branch, with screenshots and a video for the demo. The agreed requirements go to m-feature | | Build a feature | m-feature | Requirements → you approve → design and tasks → you approve → one task at a time with tests → device checks with screenshots → living spec updated → review → PR | | Make a small change | m-change | Checks it's small, then shows the code, living-spec and docs changes for one approval, makes them together and runs the checks | | Fix a bug | m-bugfix | Makes the bug fail on demand (a test, or a Maestro flow on the device), lists the likely causes, finds the root cause, then makes the smallest fix | | Stop, or switch tool or machine | m-pause | Writes a handoff in the spec, commits and pushes | | Carry on anywhere | m-continue | Reads the spec and handoff, then picks up at the next step | | Review a branch | m-review | Checks the work against the spec (complete, correct, coherent) and covers both platforms, error states, security and native changes | | Ship a release | m-release | Changelog from real commits, store and client notes, versions, release check, release PR, tag, signed builds | | Stop a mistake repeating | m-learn | Turns a correction into a lasting check, a known-issues entry or a rule |

Invoke a skill with /m-feature in Claude Code and Antigravity, or $m-feature in Codex. The skills are set to run only when you invoke them.

Alongside other AI workflows

Apps that already use Superpowers, BMAD, proAgents or similar keep them as they are. The kit adds its section next to theirs and links its skills into an existing .claude/skills folder. A task started with an m- skill follows the kit's steps, and everything else keeps working as before.

Expo apps

init recognizes Expo apps that generate ios/ and android/ with npx expo prebuild. Agents get Expo's rules and skills, and never edit the generated folders. The kit's scripts generate them before each build, so device checks, release checks and signed builds work as they do in bare apps, with no Expo account. Teams that build with EAS run eas build and eas submit themselves. Expo apps that keep ios/ and android/ in git are set up like bare apps.

Update or remove

Run npx agentic-mobile-kit in an app that has the kit. It offers to update it, add CI, check the setup or remove the kit, and tells you when a newer version is out. The same as commands:

npx agentic-mobile-kit@latest sync   # update the kit's files, merging in your own changes
npx agentic-mobile-kit uninstall     # remove the kit; files you changed are kept and listed

Both need a clean working tree and leave the result for you to review with git diff.

• Report: sync lists what it updated, added, removed and merged.
• Your docs: the product, tech, structure and conventions files your team filled in are never changed.
• Adding CI later: sync --ci github or --ci codemagic does it.
• From 0.6 or earlier: the update moves the kit's files from docs/ai/ and scripts/ai/ to docs/amk/ and scripts/amk/, with your edits. Your own files in the old folders stay where they are, and it lists any of your files that still use the old paths.

Requirements

• Git and Node.js 20 or later, on macOS, Windows or Linux. The kit's tests run on all three for every change.
• The app at the root of its git repository.
• Android work needs the Android SDK and Java. iOS work needs a Mac with Xcode and CocoaPods. doctor checks the rest.
• On Windows, turn on Developer Mode (Settings > System > For developers) and run git config --global core.symlinks true. Then init can link the Claude Code skills folder, and clones keep the link.

More

• What the kit adds to an app: every file, script, hook and check it installs.
• Team guide: day-to-day use, what's enforced, troubleshooting and a 1-hour training plan.
• Branch protection: the GitHub settings that make the CI checks required.
• Tool matrix: how the kit was tested in each AI tool and on a real app.
• Published from GitHub Actions with npm provenance.

License

MIT