The ToolkitShips with every AppLighter template

A toolkit your agent
already knows how to use.

Every AppLighter template ships with skills and slash commands for the work agents do over and over. Tested recipes, not improvised guesses. Run them in Claude Code, Codex, OpenCode, or Cursor.

~/taxi-booking · type /
any agent
$ claude
/
Available commands:
/add-screenscaffold a screen with router + state
/add-provideradd a new auth/payment/map/AI provider
/swap-backendmigrate Vibecode DB to a new adapter
/audit-securityrun the public security checklist
/rebrandtokens, logos, copy, splash, icon
/shipeas build + submit, iOS & Android
… 18 commands total. Type / to see them in your editor.
Works first-class with
Claude CodeCodexOpenCodeCursorWindsurfAider
Why the toolkit

Agents guess.
Tested recipes don’t.

An agent without a toolkit is an agent improvising. It re-derives where the auth lives, re-invents how the payment provider should plug in, re-decides which way to wire a new screen. Each guess burns context and creates drift.

A skill is a tested recipe an agent can run. /add-screen means “here’s exactly how this codebase adds a screen — same patterns, same conventions, same tests.” Less drift. Faster shipping. Smaller diffs.

Templates ship with the toolkit pre-installed. You get the foundation and the way to extend it consistently.

The catalog

18 commands.
6 categories.

Updated monthly. New commands ship with each release.

Extend

Add new screens, features, providers
/add-screen
Scaffold a new screen with Expo Router, design tokens, and state wired up.
args: <name> [--tab] [--modal]
/add-provider
Add a new auth, payment, map, or AI provider behind the existing abstraction.
args: <type> <name>
/add-feature
Generate a full feature — screen + state + API + tests — following the architecture.
args: <name> [--with-ai]

Migrate

Swap backends, upgrade frameworks
/swap-backend
Migrate Vibecode DB from Supabase to Drizzle, Neon, or your own backend.
args: <adapter>
/upgrade-expo
Walk through an Expo SDK upgrade with the AppLighter release-notes checklist.
args: [--to=<sdk>]
/upgrade-rn
Bump React Native + native dependencies with rollback-safe migrations.
args: [--to=<version>]

Verify

Security, types, and code health
/audit-security
Run the public security checklist locally. Flags leaked keys, missing RLS, weak edges.
args: [--strict]
/typecheck-all
Type-check rider, driver, and edge function packages in one pass.
args: [--watch]
/lint-fix
Apply repo-standard lint + format rules. Agent-safe (no logic changes).
args: [path…]

Brand

Make the template yours
/rebrand
Walk the agent through tokens, copy, palette, and typography for your brand.
args: [--from=<brief.md>]
/update-icons
Generate iOS + Android icons in every required size from one SVG.
args: <path-to-svg>
/splash-screen
Build a splash screen with motion that matches your brand tokens.
args: [--animated]

Ship

Build, submit, and update
/eas-build
EAS build with the right env, version bump, and credential check for both platforms.
args: [--profile=<name>]
/eas-submit
Submit to App Store and Play with store metadata, screenshots, and release notes.
args: [--platform=ios|android]
/ota-update
Push an Expo Updates OTA with safe-rollback if crash rate spikes.
args: [--channel=<env>]

AI

Build, tune, and swap AI features
/add-ai-feature
Add a streamed AI feature behind an edge function. Tool use + retries built in.
args: <name>
/swap-ai-provider
Switch between Anthropic, OpenAI, or local. Streaming and tool use re-tested.
args: <provider>
/test-prompts
Run the repo's prompt regressions against the current model and report deltas.
args: [--against=<model>]
How it works

One folder.
Many agents.

Each skill is a single Markdown file with a tested prompt and step list. The same skill compiles to a Claude Code command, a Codex skill, an OpenCode workflow, and a Cursor rule. Write once, run anywhere.

  • Version-controlled in the repo
  • Tested against multiple models
  • Editable — fork and adapt for your team
  • Shareable — commit a new skill, PR it back
skills/add-screen.md
---
name: add-screen
description: Scaffold a new screen
args: ["name"]
---

# Add a new screen named {{name}}

1. Create app/{{name}}.tsx with the
   standard layout from components/layouts.
2. Register the route in app/_layout.tsx
   following the existing order.
3. Wire state via lib/state/use{{Name}}()
   following the hook pattern.
4. Add a placeholder test in
   tests/screens/{{name}}.test.tsx.
5. Run pnpm typecheck & pnpm test.

## Don't
- Don't introduce a new state library.
- Don't bypass the layout component.
- Don't skip the test file.
Per-agent setup

Configured for every agent
in the repo.

No setup. Clone the repo, open it in your agent, the commands appear.

Claude Code
.claude/commands/
├─ add-screen.md
├─ swap-backend.md
└─ … 16 more
CLAUDE.md
Codex
.codex/skills/
├─ add-screen/
├─ swap-backend/
└─ …
AGENTS.md
OpenCode
.opencode/
├─ commands.yml
└─ workflows/
AGENTS.md
Cursor
.cursorrules
.cursor/commands/
├─ add-screen.md
└─ …
Windsurf
.windsurfrules
.windsurf/workflows/
└─ …
Use another agent?
Request adapter →
Community

Share a skill.
Earn the credit.

The toolkit grows with the community. Found a recipe that works for your team? PR it. We review, test against multiple agents, and ship it in the next release. Your name on the changelog.

CHANGELOG.md · this month
  • NEW/test-promptsby @priya
  • UPD/swap-backend+ Neon adapter
  • NEW/add-ai-featureby @adit
  • UPD/audit-security+ OWASP-25 checks
  • NEW/ota-updateby @core
5 commands this month. Next release: 2 weeks.

The toolkit ships with
every template.

Buy any template. Get all 18 commands. Get every command added next month, automatically.

Stay Updated on the Latest UI Templates and Features

Be the first to know about new React Native UI templates and kits, features, special promotions and exclusive offers by joining our newsletter.