- Home
- Skills
- Mobile Development
- React Native Architecture: Production Patterns
More screenshots
Works with the AI Agents you already use
React Native Architecture: Production Patterns
Make the six React Native decisions that get expensive later — platform mode, navigation, offline data, native boundary, performance budget, release pipeline — once, up front, in writing.
$5
React Native Architecture: Production Patterns
Example session with this skill installed
Decide the architecture for a new app: a two-person team is building a plant-care reminder app. Users catalog their plants, get watering reminders, and log waterings and photos. It must work in basements and greenhouses with no signal, sync when back online, and send push reminders. iOS and Android, small budget, first React Native project for both developers. Give me the six decisions.
- Read your context and instructions
- Compiled the react native architecture
D1 — Platform: Expo managed + EAS. No exotic SDKs: camera, notifications, and storage are all first-party. Named reopen trigger: a future Bluetooth soil-sensor integration with no config plugin.
D2 — Navigation: root auth gate in app/_layout.tsx; (auth) group for email sign-in; (tabs) group with three destinations — Plants, Today, Settings; plant detail and add-plant as a stack journey inside Plants; photo capture as a modal. Deep links from reminder notifications land on the gate and resume to the right plant after sign-in.
D3 — Data and the offline contract: React Query owns all server state (plant list, logs). Zustand holds UI state only (selected tab, draft form). MMKV persists the write queue. Contract: browsing your catalog works offline from cache; logging a watering or photo queues with a visible "waiting to sync" badge on the plant card; account deletion and sharing fail loudly offline — they never queue silently.
D4 — Native boundary: zero custom modules. Image picking, notifications, and background scheduling are first-party modules plus config plugins. Module count and justification recorded in the decision log.
D5 — Performance budget: cold start to interactive under 3 seconds on a mid-range Android (named test device: a 2021-era Pixel); plant list scrolls 200 rows via FlashList; plant photos stored sized-and-cached, never full-resolution in list rows.
D6 — Release: EAS Build for binaries, CI bumps version and build numbers, staged rollout for any release touching the sync queue, EAS Update for JS-only fixes gated by runtime version so an update never assumes a native API an older binary lacks.
Health rubric self-score: 9/10 (release at 1 until the first staged rollout actually runs). Every later argument — "just mirror the plant list in Zustand," "ship the queue change by OTA tonight" — now has a written answer.
Connects securely to your tools. The creator never sees your data.
About this skill
Most React Native pain is not bad code; it is deferred decisions. The auth check copied into five screens, server data mirrored into a client store "for now," native projects owned before anyone can name why — each one is cheap in week one and a rewrite in month six. This skill makes the six load-bearing decisions in the order they get painful, and writes each one down with the trigger that would reopen it. It starts with platform mode (Expo managed vs bare, decided by a named native requirement, not a someday fear), then navigation topology before screens — groups, one root auth gate, tabs as destinations and stacks as journeys. The data layer gets three separate owners (server state, client state, persistence) and a written offline contract: what works with no network, what queues with a user-visible sync state, what fails loudly. The native boundary is treated as a budget — first-party modules, then config plugins, custom native code last and justified one by one. Performance becomes a budget with numbers measured on a real mid-range Android, and the release pipeline runs through CI with staged rollouts and an OTA path gated by runtime version. A decision table, a nine-item review checklist, a project layout that survives growth, and a five-dimension health rubric (ship at 8+/10, no zero on navigation or data) are included, plus the two worked examples: the root auth gate component, and a complete six-decision pass for an offline field-service app.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Mobile Development.
- It makes architecture decisions, not implementations — it will not write your screens or debug a specific component.
- Examples center on Expo Router and EAS; teams on other stacks must translate the mechanics, though the decisions carry over.
- Performance numbers are budgets to verify on your own hardware, not measurements of your app.
- It does not cover backend or API design; the data-layer decisions assume an API exists or is being designed separately.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 24 days ago
Permissions
No special permissions declared or detected