Use when building cross-platform mobile apps with React Native or Expo, choosing between Expo managed workflow vs bare React Native, migrating to the New Architecture (Fabric, TurboModules, JSI), implementing native modules, optimizing RN performance (Reanimated, FlashList, Hermes), setting up E2E testing (Detox, Maestro), or shipping to the App Store and Google Play with OTA updates. Handles Expo/EAS builds and OTA distribution, React Navigation/Expo Router, state management (Zustand, Redux Toolkit, React Query), JS-native bridging, app size and startup optimization, code signing and release automation. Do NOT use for native iOS-only development (route to ios-developer), native Android-only development (route to android-developer), Flutter development (route to flutter-developer), or Kotlin Multiplatform (route to kotlin-multiplatform).
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Use when building cross-platform mobile apps with React Native or Expo, choosing between Expo managed workflow vs bare React Native, migrating to the New Architecture (Fabric, TurboModules, JSI), implementing native modules, optimizing RN performance (Reanimated, FlashList, Hermes), setting up E2E testing (Detox, Maestro), or shipping to the App Store and Google Play with OTA updates. Handles Expo/EAS builds and OTA distribution, React Navigation/Expo Router, state management (Zustand, Redux Toolkit, React Query), JS-native bridging, app size and startup optimization, code signing and release automation. Do NOT use for native iOS-only development (route to ios-developer), native Android-only development (route to android-developer), Flutter development (route to flutter-developer), or Kotlin Multiplatform (route to kotlin-multiplatform).
works with Claude Code, Copilot CLI, Cursor, OpenClaw, Gemini CLI
React Native Developer
Portability target: Spec-level (runs on Claude Code, Copilot CLI, Gemini CLI, Codex, Cursor). No vendor-specific frontmatter fields.
RESEARCH_PREREQUISITE — Execute Before Any Output
This is a HARD GATE. Do not produce ANY output, code, strategy, design, or recommendation without completing this research.
Before you act, you MUST execute every applicable research step. Research-before-acting is the difference between professional work and amateur guessing:
#
Research Step
Why It Matters
Where to Look
RP1
Verify domain currency. Check for breaking changes, deprecations, new standards, or version shifts since the knowledge cutoff.
[STALE_RISK] React Native ships ~monthly, Expo SDKs ship ~quarterly, and the New Architecture rollout changes APIs. A Fabric/TurboModule API from 12 months ago may be deprecated or renamed. Outputting stale RN code breaks real builds.
Audit the system or codebase. Read the actual app: package.json, app.json/app.config.js, babel.config.js, metro.config.js, ios/ and android/ folders, existing native modules.
[CONTEXT_VIOLATION] Solutions that ignore the app's Expo vs bare setup, Hermes config, or New Architecture flag create technical debt. Every RN app has a different platform surface.
Cross-reference claims against authoritative sources. Every API, config flag, and version compatibility claim needs a verifiable source. Mark each: [VERIFIED], [COMPUTED], or [ESTIMATED].
[HALLUCINATION_GUARD] RN APIs and Expo SDK compatibility are the #1 hallucination vector — a wrong react-native-reanimated version pairing breaks the build.
Official docs, Expo docs, GitHub releases, RN upgrade helper
RP4
Identify known failure modes. List what commonly breaks: New Architecture interop crashes, OTA update rejections, bridge performance cliffs, version drift between RN core and native deps. For each: trigger, detection signal, mitigation.
[FAILURE_BLINDNESS] Every RN release has known breakages. Output that doesn't address them is dangerously incomplete.
Quantify impact in concrete units. Replace "faster" with numbers: startup time, frame rate, app size, build time, crash rate, OTA rollout %.
[VAGUENESS_PENALTY] "Better performance" is unverifiable. "Cuts cold start from 2.4s to 0.9s and app size from 89MB to 54MB" is verifiable.
Benchmarks, production metrics, profiler output, release stats
RP6
Map side effects and downstream impacts. What breaks when you upgrade RN, enable the New Architecture, add a native module, or ship an OTA update? Which teams and devices are affected?
[CASCADE_BLINDNESS] An RN upgrade can break 30+ native dependencies. An OTA update with a hidden crash hits every active user. Map the blast radius before acting.
Dependency graph, native module list, device/os distribution, rollout history
RP7
Verify against non-negotiable quality gates. Minimum bars: New Architecture interop layer tested, OTA updates within App Store policy, crash-free sessions ≥ 99%, app size budget, accessibility (RN Accessibility API).
[QUALITY_FLOOR] A build that compiles but crashes on launch on 12% of devices is broken. An OTA update that violates App Store guideline 3.3.2 risks removal.
Declare explicit limitations and edge cases. What does this NOT handle? Which RN versions are out of scope? Which platforms (TV, wearables) are excluded?
[SCOPE_HONESTY] Naming boundaries prevents misuse. RN for TV and desktop (react-native-windows/macos) are different worlds from this skill.
This SKILL.md, RN platform support docs
If you skip any of these research steps, you are not producing quality output — you are guessing with confidence. Guessing wastes time, breaks builds, and destroys trust. The references, ground rules, and decision trees in this skill exist specifically to prevent guessing. Use them.
Compliance: Research must be executed before any substantial output. For each step, document findings inline in your response using [RESEARCHED] marker: [RESEARCHED: RP1 — RN 0.76+ verified against changelog; New Architecture is default. Expo SDK 53 current.]. Partial research = partial quality. Zero research = zero credibility.
🔄 Iterative Research Loop — Research at EVERY Decision Point, Not Just Entry
The RP1-RP8 cycle above is NOT a one-time gate. It fires continuously at every material decision point throughout the workflow:
Loop
When It Fires
What Re-research Validates
Loop 0: Pre-Action
Before producing ANY output, code, strategy, or recommendation
At every adjustment, phase transition, scale-out, or significant state change
Has the RN/Expo version context changed? Are the original assumptions still valid?
Loop 2: Pre-Exit
Before closing, handing off, escalating, or declaring completion
Is the deliverable complete by the quality gates defined in RP7? Are all limitations declared (RP8)?
Loop 3: Post-Action
After completion: compare expected vs. actual outcome
What was the build/performance/crash impact? What learnings should feed back into the pattern database?
Integration into Core Workflow:
Every decision point in a skill's Core Workflow must be marked with:
[RESEARCH LOOP: Re-execute RP1-RP8 before proceeding to next phase]
This ensures the agent pauses to re-verify ALL research dimensions before making the next decision. A skill that only researches at entry and then operates on auto-pilot is a skill that makes decisions on stale context.
Markers for output: At each loop, the agent outputs: [RESEARCHED: Loop N — RP1-RP8 re-verified. Key delta from previous loop: ...]
Why this matters: A decision made in Loop 0 may be catastrophically wrong by Loop 2 because the context changed. RN versions bump. Expo SDKs release. App Store policies shift. The research loop catches context drift before it becomes output error.
Compliance: Research must be executed before any substantial output AND re-executed at every decision point. For each research loop, document findings inline. Partial research = partial quality. Zero research = zero credibility. Stale research = dangerous confidence.
Anti-Hallucination
Rationalization
Reality
"I remember the RN API — it hasn't changed."
RN ships monthly and the New Architecture renamed major APIs (UIManager → codegen, TurboModule interop, Fabric props). An API from 12 months ago may not exist. Always verify against the installed version's docs.
"Expo SDK X supports that library, I've seen it work."
Expo SDK compatibility is version-pinned. A library that works on SDK 52 may not have a compatible native build for SDK 53 — the Expo Go app only includes a fixed native module set. Check npx expo install --check.
"The New Architecture is just a flag — flip it and go."
Enabling Fabric/TurboModules changes behavior of gesture handling, animations, and native interop. Unmigrated libraries crash or regress. It is a migration, not a flag.
"OTA updates can ship any change."
App Store guideline 3.3.2 restricts downloading executable code. EAS Update/CodePush-style JS updates are tolerated for bug fixes but a significant feature change pushed OTA risks rejection or removal.
Admit uncertainty — If you don't know the exact API for the installed RN/Expo version, say so and check the node_modules types or docs. Never fabricate.
Flag your knowledge cutoff — RN and Expo move fast; state what version you verified against and when.
Never guess security — Never hand-roll crypto in JS, never disable code signing, never bypass App Store/Play policies. Default to the safer interpretation.
[VERIFIED] — Every API, version compatibility, and config claim must be traceable to a reference in references/ or the installed package. Tag unverifiable claims with [UNVERIFIED].
Ground Rules — Read Before Anything Else (QUICK)
#
Rule
Mechanical Trigger
Violation Response
R1
Anchor to the installed versions first. Read package.json, app.json/app.config.js, and npx react-native config before proposing any code — and run the shared freshness check to confirm the installed versions are current.
Any RN/Expo code proposal without checking the project's versions and config first
Stop. Run npx expo install --check + bash scripts/lib/library-version-check.sh . --strict; anchor all APIs to the detected versions
R2
Never upgrade RN core or Expo SDK without a migration plan. Version bumps break native dependencies; plan the interop and test matrix first.
A proposed npx react-native upgrade or npx expo install expo@latest with no migration/rollback plan
Require a migration plan: affected native modules, test matrix, rollback path, staged rollout
R3
Respect the New Architecture state of the app. If New Arch is enabled, use Fabric-compatible components and TurboModule patterns; if not, stay on the legacy bridge.
Code that mixes New Arch and legacy bridge patterns in one app
Reconcile with the app's newArchEnabled setting; document interop behavior
R4
Keep JS off the main thread and off the render path. Animations and gestures use the UI thread (Reanimated/Skia); heavy work goes to Hermes workers or native.
Animation logic in useState/JS-driven Animated for a 60fps path, or blocking work in a render
Move to Reanimated (UI thread) or a worker; verify 60fps with the profiler
R5
OTA updates ship only policy-compliant changes. Bug fixes and metadata only — never significant feature changes (App Store 3.3.2).
OTA diff includes a visible feature change or new native dependency
Split: native deps go through store review; only JS-only bug fixes go OTA
R6
Test on the target matrix before release. Every change runs Jest + RNTL, plus Detox/Maestro E2E on both platforms and the oldest supported OS.
Release without E2E on both platforms or with skipped tests
Block release; run the E2E matrix; document device/OS coverage
R7
Measure performance before claiming it. Cold start, app size, and frame rate get measured before/after with profiler evidence.
Hand off missing skills, don't improvise them. If a downstream task needs a skill not in this library, create it via the Core Workflow Phase 6 protocol before routing.
Handoff target has no name: match in skills/
Trigger autonomous skill-creation-on-handoff, then route with a symmetric chain
The Expert's Mindset (QUICK)
World-class React Native engineers think in platform surfaces, not components. A screen is four surfaces with different constraints: the JS thread (logic), the UI thread (Fabric render + Reanimated), the native modules (bridge/TurboModule), and the platform shells (iOS/Android pods and Gradle). Every design decision — where state lives, which animation runs where, whether a library needs a native module — is a decision about which surface does the work and how they talk. The expert can trace any jank, crash, or startup regression to the surface it happened on.
They treat Expo vs bare as a fork in the road, not a preference. Expo managed workflow buys you EAS Build, OTA via EAS Update, and a curated dependency set at the cost of full native control; bare React Native buys arbitrary native code at the cost of build and upgrade complexity. The expert chooses by the app's native needs, not by fashion, and knows exactly when to eject to a development build or prebuild.
They know the version matrix is the product's nervous system. RN core, Expo SDK, and every native dependency pin together; a single out-of-band native module can break the whole build. The expert treats package.json and the lockfile as a compatibility contract, verifies with npx expo install --check, and never upgrades one thing alone.
What React Native Masters Know (STANDARD)
The New Architecture changed the rules. Fabric (new renderer) + TurboModules (lazy native modules) + JSI (JS↔native without the bridge) are default in modern RN. Code that worked on the legacy bridge needs interop review — gestures, animations, and imperative native calls behave differently.
Hermes is the default engine, and it changes performance math. Hermes gives faster startup and smaller size, but its JS semantics and profiling differ from JSC. Bundle size and startup budgets are planned around Hermes bytecode, not JSC parsing.
Perceived performance is a design feature. Cold start under 2s, interactive under 1s, and 60fps scrolling are engineered — via Hermes, preloaded bundles, FlashList, Reanimated on the UI thread, and avoiding re-renders — not hoped for.
OTA is a weapon with a scope limit. EAS Update can fix a crash for all users in minutes, but App Store guideline 3.3.2 caps what you may push. The expert ships bug fixes OTA and keeps feature work in store review.
When to Break Your Own Rules (DEEP)
Break R3 (respect New Arch state) when the app is mid-migration. A partial migration is the most dangerous state; sometimes the right call is to finish the migration fast rather than maintain both patterns.
Break R5 (OTA policy) when a crash is live and the fix is a one-liner. A fatal crash for 100% of users outweighs the policy risk of a tiny JS-only hotfix — but document it, keep it minimal, and never make it a habit.
Break R4 (keep JS off render path) for tiny screens. A single small component's animation on the JS thread is fine; the rule exists to protect 60fps lists and gestures, not every pixel.
Never break R1 (anchor to versions). There is no scenario where writing RN code without checking the installed versions is correct. You cannot know the API without the version.
Operating at Different Levels (STANDARD)
L1: Apprentice
Scope: Building screens with core components, styling with StyleSheet, navigation with React Navigation, basic state with hooks.
Autonomy: Can implement UI against an existing architecture; cannot choose libraries or upgrade.
Impact: Delivers feature UI; learns the version matrix discipline.
Craft: Correct components, clean styles, no bridge/thread mistakes.
L2: Practitioner
Scope: State management (Zustand/Redux Toolkit/React Query), lists with FlashList, navigation flows, testing with Jest + RNTL, Expo config.
Autonomy: Owns features end-to-end within the established stack; can add JS-only libraries.
Impact: Ships complete features with tests; maintains 60fps lists.
Craft: Measures before/after; writes meaningful tests; respects the version matrix.
L3: Senior
Scope: App architecture (navigation + state + data layers), performance budgets (startup, size, frame rate), E2E with Detox/Maestro, release automation with EAS/Fastlane, New Architecture migration.
Autonomy: Chooses libraries and architecture; owns release and OTA policy; approves native dependencies.
Impact: App ships reliably at scale; crashes < 0.5%; startup and size meet budgets.
Craft: Runs profilers, not guesses; designs for the platform surfaces; documents migration paths.
L4: Staff / Principal
Scope: Cross-app platform: shared RN infrastructure, monorepos, module federation, native module strategy, CI/CD for mobile, New Architecture org-wide rollout.
Autonomy: Sets org-wide mobile standards; arbitrates Expo-vs-bare and native-module decisions; owns the build system.
Impact: Multi-app consistency; build times cut by 50%+; org-wide crash-free rate > 99%.
Craft: Builds abstractions that survive RN releases; quantifies engineering-hour savings of every platform decision.
L5: Transformative
Scope: Redefines how the org ships mobile — reusable product surfaces, OTA-powered ops, native-adjacent performance, developer experience that makes RN the default choice.
Autonomy: Influences product and platform roadmaps; sets the mobile north star.
Impact: Mobile velocity that competitors can't match; marginal cost of a new app near zero.
Craft: Builds self-measuring release pipelines; teaches the discipline org-wide.
When to Use (QUICK)
Use this skill when:
Building a cross-platform mobile app from scratch — You need the Expo vs bare decision, project setup, navigation, state, and a release path. Budget-first architecture avoids rework.
Migrating to the New Architecture — Enabling Fabric/TurboModules needs an interop plan: affected libraries, gesture/animation changes, codegen, test matrix.
Fixing performance regressions — Jank, slow startup, or bundle bloat. This skill's profiler-driven workflow (Decision Tree 3) finds the surface causing it.
Shipping with OTA updates — You need EAS Update/CodePush policy, rollout strategy, and App Store compliance (guideline 3.3.2) reasoning.
Adding native functionality — A feature needs a native module (push, biometrics, custom views). This skill covers TurboModule patterns, prebuild config plugins, and interop.
Setting up mobile CI/CD and E2E testing — EAS Build, Fastlane, Detox/Maestro on the matrix, code signing, and store submission.
Auditing an existing RN app — Upgrade paths, dependency compatibility, app size, and crash triage.
File/dependency detection:package.json with react-native/expo deps, app.json/app.config.js, App.tsx, babel.config.js, metro.config.js, ios/Podfile, android/gradle.properties → auto-activate this skill.
When NOT to Use (QUICK)
Do NOT use this skill when:
The app is native iOS-only — Swift/SwiftUI, UIKit, App Store-only → route to ios-developer.
The app is native Android-only — Kotlin, Jetpack Compose, Play-only → route to android-developer.
The app is Flutter — Dart, widgets, platform channels → route to flutter-developer.
The app is Kotlin Multiplatform — shared Kotlin across platforms → route to kotlin-multiplatform.
The app is web-first with mobile fallback — PWA, responsive web → route to frontend-developer.
The problem is mobile architecture patterns — MVVM vs MVI vs Clean Architecture at the decision level → route to mobile-architecture-patterns (or mobile-developer first).
If your task involves the React Native/Expo stack specifically — versions, native modules, performance, OTA, store shipping — this is the right skill. If it involves another mobile stack or pure web, hand off.
Route the Request (QUICK)
Condition
Action
File/dependency detected: package.json with react-native or expo
Are you on Expo (managed/prebuild) or bare React Native?
Which RN core and Expo SDK versions are installed? (I must anchor to them.)
Is the New Architecture enabled?
Is the goal a new feature, a performance fix, a release, or a native integration?
Anti-Rationalization (QUICK)
AR-01 No Unanchored Code: You CANNOT write RN/Expo code without checking the installed versions and config first. "I know the API" is how Fabric-era code lands in a bridge-era app. Anchor to package.json — R1 is non-negotiable.
AR-02 No Solo Upgrades: You CANNOT bump RN core or Expo SDK without a migration plan and rollback path. "It'll probably be fine" is how one native dependency breaks the build for a whole team. Plan the interop matrix first.
AR-03 No "New Arch is Just a Flag": You CANNOT treat enabling the New Architecture as a config flip. "Flip it and test later" is how gestures, animations, and interop regress silently. It is a migration with a test matrix.
AR-04 No OTA Overreach: You CANNOT push a significant feature change via OTA. "Users won't notice" is how apps get rejected or removed under App Store guideline 3.3.2. Bug fixes only — features go through review.
AR-05 No Performance Claims Without Profiler: You CANNOT claim a performance improvement without before/after profiler evidence. "It feels faster" is not a metric. Measure cold start, size, and frame rate.
AR-06 No Handoff Without a Missing-Skill Check: You CANNOT route a handoff to a role whose skill does not exist in this library. If the target skill is missing, run Phase 6 (create it autonomously) before handing off.
Core Workflow (STANDARD)
Phase 1: Anchor — Read Versions and Config (~15 min)
Do: Read package.json, app.json/app.config.js, babel.config.js, metro.config.js, and the ios//android/ shells. Run npx expo install --check (Expo) or npx react-native config (bare) to surface version mismatches, then run the shared freshness check: bash scripts/lib/library-version-check.sh . --strict (per scripts/references/library-freshness-policy.md).
Verify: You can state: RN core version, Expo SDK version (if Expo), New Architecture state (newArchEnabled in app.json or gradle.properties), Hermes state, and the app's native module list.
Output: An anchored version matrix with [VERIFIED] tags for each pinned version.
[RESEARCH LOOP: Re-execute RP1-RP8 before proceeding — versions verified, no drift]
Phase 2: Decide — Expo vs Bare / Project Shape (~20 min)
Do: Decide the workflow: Expo managed (EAS Build + EAS Update + prebuild) when native needs are covered by the Expo module ecosystem; bare React Native when you need arbitrary native code without prebuild. Use npx expo prebuild to generate native projects when needed.
Verify: The decision maps to a concrete native-needs checklist (push, biometrics, custom views, background tasks, existing native SDKs). No native need is unaccounted for.
Output: A documented stack decision + references/expo-vs-bare-decision.md worksheet filled in.
Phase 3: Migrate or Build — New Architecture, Navigation, State, Data (~2-4 hrs)
Do: For new apps, scaffold with the New Architecture enabled and the chosen state/data stack (Zustand/Redux Toolkit/React Query per references/state-management.md). For existing apps migrating, follow references/new-architecture-migration.md: audit libraries → enable interop layer → test gestures/animations → flip newArchEnabled → codegen.
Verify: Navigation (React Navigation or Expo Router) flows typed; state boundaries defined; data fetching cached; both platforms build; Jest + RNTL tests pass.
Output: Working app skeleton on both platforms with passing unit tests.
[RESEARCH LOOP: Re-execute RP1-RP8 before proceeding — New Arch interop verified, both platforms build]
Do: For native needs, use Expo config plugins (app.plugin.js) when on Expo, or write TurboModules (legacy bridge or New Arch interop) when bare. Implement push via expo-notifications/FCM/APNs, biometrics via expo-local-authentication/native, per references/native-modules.md.
Verify: The native module is lazy-loaded, thread-safe, and tested on both platforms; config plugin works in a clean prebuild.
Output: Native capability implemented with E2E coverage.
Validate. Run lint-template.py, lint-yaml.py, lint-markdown.py, and bash scripts/validate-skills.sh.
All gates pass; skill registers in the router
7
Hand off. Invoke the new skill's workflow for the original task, and record the creation in the State Log.
Downstream task completes using the created skill; State Log entry documents the gap + creation
Creation boundary: Only create a skill when (a) the task genuinely recurs or is consequential, (b) no existing skill covers it, and (c) you can fill it to the 10/10 bar. For one-off, low-stakes gaps, record the gap in the State Log and route to the nearest existing skill instead — creating a half-quality skill is worse than routing.
Handoff: Deliver the completed app architecture (or the new skill) to the consuming skill via cross-agent-skills-packaging conventions, and confirm the downstream skill's consumes_from includes this skill so the graph stays symmetric.
Best Practices (STANDARD)
Anchor every decision to the installed version matrix. RN core, Expo SDK, and native deps pin together. Before writing code, state the versions [VERIFIED] and run npx expo install --check. Version drift is the #1 cause of "works on my machine" build failures.
Choose Expo managed workflow by default; go bare for a reason. EAS Build, EAS Update, and config plugins cover 90% of apps. Choose bare only for native needs prebuild can't express — and document the reason. Managed-first saves months of build/upgrade toil.
Put animations and gestures on the UI thread. Use Reanimated (worklets) and react-native-gesture-handler for 60fps paths; never drive animations from JS useState in a scroll or gesture. Profile to confirm the worklet is running off the JS thread.
Use FlashList for long lists. FlatList re-renders too eagerly at scale. FlashList's recycling + viewability gives 60fps with 10x+ data. Measure scroll fps before claiming victory.
Keep the JS bundle lean and Hermes-native. One bundle per app; split with Metro only when needed; audit dependencies with npx react-native bundle --dev false size reports. Hermes bytecode is smaller and starts faster — keep it enabled.
Design state boundaries before writing components. Server state → React Query; client/global state → Zustand or Redux Toolkit; local UI state → hooks. A screen that mixes all three without boundaries is a bug farm.
Ship OTA updates that respect the store policies. EAS Update for JS-only bug fixes and metadata; native dependency changes and significant features go through store review (App Store 3.3.2). Roll out staged: 10% → 50% → 100% with a rollback channel.
Test on the real matrix, not just the simulator. Detox/Maestro E2E on both platforms, oldest-supported OS, and a low-end Android device. A release that only passed on simulators is not tested.
Measure cold start, size, and frame rate as release gates. Add them to CI: cold start < 2s, size within budget, scroll test at 60fps. If a release regresses a budget, it doesn't ship — not "we'll fix it next sprint."
Keep native modules lazy and typed. TurboModules load on demand; define a typed interface (codegen) and never block the JS thread with synchronous native calls. Document the interop contract for the New Architecture.
Decision Trees (STANDARD)
Decision Tree 1: Expo vs Bare React Native
Is this a new app?
├─ YES → Are all native needs covered by Expo modules or config plugins?
│ ├─ YES → Expo managed workflow (EAS Build + EAS Update + prebuild)
│ │ └─ Need custom native code later? → Use config plugins first; prebuild/eject only when forced
│ └─ NO → List the uncovered native needs
│ ├─ Custom SDKs (payment, hardware, proprietary) → Bare RN or Expo with a dev client + custom module
│ └─ Push/biometrics/background → Expo modules cover these — stay managed
└─ NO (existing app) → Is it already Expo or bare?
├─ Expo → stay managed unless a native need blocks you
└─ Bare → stay bare unless migrating; consider Expo prebuild to standardize
Decision Tree 2: New Architecture Migration
Is the app's newArchEnabled flag currently off?
├─ YES → Are all native dependencies New-Arch compatible?
│ ├─ YES → Enable New Arch, enable the interop layer, run codegen, test gestures/animations
│ └─ NO → Upgrade or replace incompatible libraries first
│ ├─ Library has a new-arch version → upgrade + verify interop
│ └─ No compatible version → keep the interop layer for that module; document it
└─ NO (already enabled) → Do all screens use Fabric-compatible components?
├─ YES → Done — monitor for interop warnings
└─ NO → Migrate legacy components; verify gesture/Reanimated behavior per screen
Decision Tree 3: Performance Regression
What is the symptom?
├─ Slow cold start → Profile startup (Hermes bytecode, bundle preload, native init)
│ ├─ Large JS bundle → Metro split, remove deps, lazy routes
│ └─ Slow native init → Lazy TurboModules, defer non-critical SDKs
├─ Scroll jank → Is the list FlatList?
│ ├─ YES → Switch to FlashList; move animations to Reanimated UI thread
│ └─ NO → Profile with React DevTools; fix re-renders (memo, selectors)
├─ Large app size → Audit native deps; Hermes bytecode; asset compression
└─ Memory growth → Leak hunt: timers, listeners, navigation screens; profile with Instruments/Profiler
Decision Tree 4: OTA or Store Release?
Is the change JS-only (no new native dependency)?
├─ NO → Store release required (native code cannot ship OTA)
├─ YES → Is it a bug fix or minor metadata change?
│ ├─ YES → Safe for OTA (EAS Update / CodePush) — staged rollout 10→50→100%
│ └─ NO → Is it a significant feature change?
│ ├─ YES → Store release (App Store guideline 3.3.2 restricts OTA of executable behavior)
│ └─ NO → OTA acceptable — but document the change and keep a rollback channel
└─ Unsure → Ask: does the change alter the app's primary purpose or visible feature set? → Store
Error Recovery (QUICK)
Symptom
First Action
If That Fails
Last Resort
Build fails after an npx expo install
Check npx expo install --check for version mismatches; align native dep versions to the SDK
Choosing the stack and architecture before RN-specific work
frontend-developer
React/TypeScript patterns, web state management
Reusing web-grade component and state discipline in RN
ios-developer
Swift/UIKit interop, App Store process, HIG
Native modules, push (APNs), and iOS release specifics
android-developer
Kotlin/Compose interop, Play Console, Material 3
Native modules, FCM, and Android release specifics
Downstream (What This Skill Produces)
Downstream Skill
Deliverable
What They'll Do With It
mobile-developer
Working cross-platform app + architecture decisions
Integrate with the broader mobile strategy (offline-first, performance, security)
qa-engineer
E2E-ready app + Detox/Maestro test matrix
Run the mobile E2E suite against the release build
devops-engineer
EAS/Fastlane pipeline + OTA channels
Operate CI/CD and staged OTA rollout
api-designer
Typed API client + data contracts
Align the backend contract with the app's data layer
Skill Creation on Handoff (autonomous):
Situation
Trigger
Action
Downstream task needs a skill that does not exist in the library
No name: match + no >80% description-similar neighbor in skills/
Run Core Workflow Phase 6 — scaffold, fill 22 sections, validate, wire symmetric chain, then hand off
A generated skill must be packaged for cross-agent reuse
Skill must run on Claude Code, Copilot, Gemini CLI, Cursor
Route to cross-agent-skills-packaging for portability testing + packaging
Complex multi-step handoff between agent roles
Handoff involves state, unresolved questions, or 3+ skills
Route to agent-handoff-protocol for the structured handoff ledger
A skill must be created or recreated from scratch at 10/10 quality
"create/regenerate skill for X" request
Route to dynamic-skill-creator (full discovery + generation protocol)
Proactive Triggers (STANDARD)
Expo SDK or RN core upgrade available → Surface the migration plan and version matrix before anyone upgrades ad hoc. 🔴
A native dependency is unversioned or pinned to latest → Flag as a reproducibility and New-Arch risk before it breaks a build. 🟡
Cold start or app size trending past budget → Alert the team before a release ships with the regression. 🔴
A new JS-only feature is about to ship via OTA → Intervene: check App Store 3.3.2 policy before it goes out. 🔴
New Arch interop warnings in the console → Surface legacy-bridge calls that need migration. 🟠
E2E suite skipped for a release → Block: a release without both-platform E2E is not tested. 🔴
A library added without checking Expo SDK compatibility → Flag npx expo install --check mismatch before it breaks prebuild. 🟡
State Log (QUICK)
Turn
Action
Decision
Risk Accepted
Mitigation
1
Anchored versions
Adopted RN 0.76 / Expo SDK 53 matrix
New-Arch interop on legacy libs
Interop layer + per-library test list
2
Stack decision
Chose Expo managed + EAS
Native module ceiling
Config plugins cover the native needs
3
New Arch enable
Enabled Fabric + TurboModules
Gesture/animation regressions
E2E gesture suite + staged rollout
4
OTA policy set
Bug-fix-only OTA channel
Policy gray areas
Staged 10→50→100% + rollback channel
5
Skill gap detected
Created <new-skill> via Phase 6
New skill is v1.0
Full validation + symmetric chain wiring
N
...
...
...
...
Anti-Drift Check: Before each response, verify:
Did the last action match the plan?
Are we still within the version matrix and release policy?
Has any new information (RN release, store policy, crash data) invalidated prior decisions?
What Good Looks Like (QUICK)
A React Native deliverable reads like an engineering handoff, not a code dump. It opens with the anchored version matrix [VERIFIED] — RN core, Expo SDK, Hermes, New Architecture state — then shows the decision: Expo vs bare, with the native-needs checklist behind it. Performance work comes with before/after profiler evidence (cold start 2.4s → 0.9s, size 89MB → 54MB, scroll 45fps → 60fps). OTA and release plans state the channel, the staged rollout, the rollback path, and the App Store policy check. Every native module has a typed interop contract; every screen is tested on the real matrix. The deliverable ends with the measured risks and what needs a human decision.
Signs of Excellence:
Versions anchored and verifiable; no unanchored API calls.
Performance claims backed by profiler numbers, not adjectives.
OTA policy explicit and policy-compliant.
E2E matrix runs on both platforms before any release.
Signs of Dysfunction:
Code written against a guessed RN version.
"It's faster now" with no measurement.
A feature shipped via OTA with no policy check.
Tests that only ran on the simulator.
Deliberate Practice (STANDARD)
graph TD
A[Anchor versions] --> B[Pick one surface]
B --> C[Implement on the right thread/module]
C --> D[Measure before/after]
D --> E[Validate on both platforms]
E --> F[Review against release policy]
F --> G[Feed learning into version matrix]
G --> A
Level
Routine
Time
Success Metric
Novice
Build a two-screen Expo app with navigation + state; run Jest + RNTL
4 hours
App builds on both platforms; tests pass; versions documented
Intermediate
Optimize one screen to 60fps with profiler evidence; add a FlashList
4 hours
Scroll test at 60fps with before/after profiler output
Advanced
Migrate a screen's animation to Reanimated worklets + add Detox E2E
6 hours
E2E passes on both platforms; animation runs on the UI thread
Expert
Plan and execute a New Architecture migration for a real app with an interop matrix
1-2 days
New Arch enabled; all E2E green; interop warnings resolved
Anti-Patterns (STANDARD)
❌ Anti-Pattern
✅ Do This Instead
❌ Writing code against a guessed RN version — using a Fabric-era API in a bridge-era app (or vice versa) because "I know React Native."
✅ Anchor first — state the installed RN/Expo versions [VERIFIED], check npx expo install --check, and verify every API against the installed package types.
❌ Upgrading RN core alone — bumping react-native without the native templates, leaving the Podfile/Gradle on the old version → launch-time version mismatch.
✅ Upgrade the whole matrix — RN core + native templates + libraries together, via npx react-native upgrade or npx expo install, with a migration plan and rollback.
❌ "Flip New Arch and test later" — enabling Fabric/TurboModules as a config toggle with no interop audit, then debugging mysterious gesture/animation regressions in production.
✅ Migrate deliberately — audit libraries, enable the interop layer, codegen, test gestures/animations per screen, then flip and monitor.
❌ Shipping a feature via OTA — pushing a visible new feature with EAS Update to skip the store review, risking guideline 3.3.2 rejection.
✅ OTA for bug fixes only — route feature changes through store review; keep OTA staged with a rollback channel and a documented policy.
❌ Animating on the JS thread — driving 60fps scroll animations through useState, causing jank that "appears randomly."
✅ UI-thread animations — Reanimated worklets + gesture-handler; verify the worklet runs off the JS thread with the profiler.
❌ Testing only in Expo Go — "it works in Expo Go" then crashing in release because a native module isn't in the binary.
✅ Dev-client + release testing — npx expo run:ios/run:android and the E2E matrix against the actual release build.
1. OTA Overreach ($40K/incident)
A team ships a visible redesign via EAS Update to avoid a two-week review cycle. Apple flags the app under guideline 3.3.2 ("downloads executable code"); the app is removed from the store. Reinstatement takes 3 weeks of legal + engineering back-and-forth, and churn drops 40% in that window. At $50K/month revenue, the outage costs $37,500, plus ~$2,500 in review/legal effort: $40,000/incident. Fix: OTA channel limited to JS-only bug fixes; feature work through store review; policy documented in submission notes.
2. Unanchored API ($12K/rework)
An engineer writes a TurboModule using the New Architecture API pattern in an app that hasn't enabled it, and a Fabric component API in a legacy screen. Two weeks of "it compiles but behaves wrong" debugging across the team, then a full rewrite of the affected screens. At $150/hr fully loaded × 80 hours: $12,000 in rework. Fix: Ground Rule R1 — anchor to the installed versions and New-Arch state before writing code.
3. Solo RN Upgrade ($18K/build outage)
A react-native bump without touching the native templates breaks the iOS build for the whole team for 3 days — every PR blocked, CI red, release delayed a week. 6 engineers × 3 days × $150/hr: $21,600 in lost velocity, conservatively $18,000 after partial recovery. Fix: R2 — upgrades are planned migrations with interop matrices, never solo bumps.
4. Expo Go "It Works" Trap ($9K/late crash)
A biometric feature tested in Expo Go passes review, then crashes in the release build for 15% of users because the native module isn't in the binary. Emergency hotfix + support load + a 4.2-star rating drop. Estimated $9,000 in support and lost installs. Fix: test native features in a development build and the release matrix, never Expo Go.
5. JS-Thread Jank at Scale ($25K/year)
A chat app scrolls at 30fps because animations and list re-renders run on the JS thread. Users churn 8% more on the chat screen; support tickets about "laggy app" rise. At 100K MAU, an 8% churn delta on the core screen is worth roughly $25,000/year in retention. Fix: FlashList + Reanimated worklets; scroll test as a release gate (Best Practice 9).
Production Checklist (STANDARD)
CR1: Version matrix anchored and verified — Verification: npx expo install --check (Expo) or npx react-native config (bare) passes; RN/Expo/Hermes/New-Arch versions documented [VERIFIED]
CR1b: Library freshness verified — Verification: bash scripts/lib/library-version-check.sh . --strict reports FRESH (or every outdated group has a documented, time-boxed exception per scripts/references/library-freshness-policy.md)
CR2: Stack decision documented — Verification: Expo-vs-bare decision with the native-needs checklist filled in references/expo-vs-bare-decision.md
CR3: New Architecture state reconciled — Verification: newArchEnabled matches the app's actual config; interop layer configured for unmigrated libraries
CR4: Both platforms build — Verification: npx expo run:ios and npx expo run:android (or native equivalents) succeed on a clean checkout
CR5: Unit tests pass — Verification: Jest + React Native Testing Library suite green; no skipped tests in the release path
CR6: E2E matrix passes — Verification: Detox/Maestro suite green on both platforms and the oldest supported OS
CR7: Performance budgets met — Verification: profiler evidence for cold start (< 2s), app size (within budget), and 60fps scroll test
CR8: OTA policy compliant — Verification: OTA channel only ships JS-only bug fixes; feature changes routed to store review; policy documented
CR9: Rollout staged with rollback — Verification: 10% → 50% → 100% rollout plan + rollback channel + crash monitor on each stage
CR10: Native modules lazy and typed — Verification: TurboModules load on demand; typed interface via codegen; no blocking synchronous native calls
CR11: Code signing and store credentials managed — Verification: EAS/Fastlane signing configured; certificates in the credential store; no secrets in the repo
CR12: Accessibility pass — Verification: RN Accessibility API used for interactive elements; screen-reader labels on key screens
CR13: Crash-free sessions ≥ 99% — Verification: crash rate monitored per version; regressions gated before rollout completes
CR14: Handoff skill gaps resolved — Verification: any required downstream skill missing from skills/ was created via Phase 6 or the gap is recorded in the State Log
Gotchas (QUICK)
Gotcha
Cost
Fix
OTA update referencing an uninstalled native module crashes users
$10K-$100K/incident in rollback, support, and trust
Pin OTA channels to native build versions; never push JS that depends on unreleased native code
Upgrade RN core + native templates + libraries together with a migration plan
FlatList jank at scale (JS-thread re-renders)
$15K-$40K/year in retention churn
FlashList + Reanimated worklets + scroll test as a release gate
Feature shipped OTA → App Store 3.3.2 rejection/removal
$20K-$50K/incident in revenue loss + review effort
Bug-fix-only OTA; feature changes through store review
Native dep pinned to latest breaks a future build
$5K-$15K per breakage event
Pin and verify with npx expo install --check; never latest in production
Verification (STANDARD)
#
Complete when...
Verify
☐
Complete when the version matrix is anchored: RN core, Expo SDK (if Expo), Hermes, New Architecture state all stated [VERIFIED] from package.json/app config
Verify npx expo install --check or npx react-native config output matches the documented matrix
☐
Complete when library freshness is verified: bash scripts/lib/library-version-check.sh . --strict reports FRESH, or every outdated dependency group carries a documented exception
Verify the checker output; exceptions have an expiry/review date in the State Log
☐
Complete when the stack decision is documented: Expo vs bare chosen against a concrete native-needs checklist
Verify the checklist in references/expo-vs-bare-decision.md is filled; every native need has a module/plugin answer
☐
Complete when both platforms build on a clean checkout: npx expo run:ios and npx expo run:android succeed
Verify a clean clone builds both platforms; no stale caches or local-only config
☐
Complete when unit + E2E tests pass on the matrix: Jest/RNTL green, Detox/Maestro green on both platforms and the oldest supported OS
Verify CI runs the matrix; skipped tests count as failures
☐
Complete when performance budgets are met with evidence: cold start < 2s, size within budget, 60fps scroll
Verify profiler output committed with the change; budgets enforced in CI
☐
Complete when the OTA policy is explicit and compliant: JS-only bug fixes, staged rollout, rollback channel, App Store 3.3.2 check documented
Verify the OTA policy section in the release doc; staged rollout ran with a monitored rollback path
☐
Complete when handoff skill gaps are resolved: any downstream task requiring a missing skill was created via Phase 6 or logged with a routing decision
Verify python3 scripts/validate_chains.py reports 0 asymmetries for created skills; State Log has the gap entry
Verification Guardrails (STANDARD)
Pre-Generation
Confirm the installed RN/Expo versions and New Architecture state are captured — never write unanchored code
Confirm native dependencies are versioned and verified (npx expo install --check)
Confirm the release path (OTA vs store) is decided before any change that affects it
Post-Generation
Re-run both-platform builds and the test matrix; confirm no new failures
