---
title: Task recipes
description: Practical routing recipes for common Expo work.
---

# Task recipes

Ordered, skill-named procedures only—no large code samples. Stop conditions and
authorization boundaries are part of each recipe.

## Standard app audit

1. `$expo-development-foundation` → classify as existing consumer app and
   `audit_scope: consumer-app`.
2. `$expo-project-orientation` → emit the evidence-backed stack card.
3. `$expo-app-audit` at standard depth → select only relevant domain skills.
4. Add `$expo-design-review`, `$expo-native-ui`, and `$expo-a11y` when UI is in
   scope.
5. `$expo-verify` → safe local checks only; report exact evidence, skipped
   checks, remediation order, and residual risk.

**Stop:** the audit is read-only. Findings and recommendations do not authorize
fixes, native builds, EAS, store, OTA, telemetry, or production actions.

## Quick app health check

1. Orient the existing app.
2. Run `$expo-app-audit` at quick depth.
3. Check critical security/config/runtime stops, SDK/dependency health,
   verification availability, and highest-severity architecture risks.
4. Close with safe local `$expo-verify` evidence.

**Stop:** do not silently expand a quick check into a standard/deep audit or
implementation.

## Design and accessibility review

1. Orient and identify the UI surfaces and platforms.
2. `$expo-design-review` → hierarchy, tokens, responsive layout, complete
   states, interaction feedback, motion, and accessibility.
3. Compose `$expo-native-ui`, `$expo-ui`, `$expo-router`, and `$expo-a11y` only
   where their boundaries match.
4. Separate source inspection from screenshots/device evidence and report
   unavailable visual or assistive-technology checks as residual risk.

**Stop:** do not claim visual verification without screenshots/runtime evidence
and do not edit without authorization.

## Audit, then separately authorized remediation

1. Complete and hand off the read-only audit report.
2. Stop and obtain a second explicit request authorizing named fixes.
3. Reconfirm scope and prioritize critical/high findings in dependency order.
4. Implement only the authorized findings with their owning domain skills.
5. Close with `$expo-verify`; keep unaddressed findings and residual risk in the
   handoff.

**Stop:** an audit request is never the authorization for step 3.

## Bootstrap a new app

1. `$expo-development-foundation` → classify as new app
2. `$expo-bootstrap` → ordered chain
3. `$expo-project-structure` then `$expo-router`
4. `$expo-native-ui` / `$expo-ui` when screens need native-feeling UI
5. Data, DOM, modules, upgrade, EAS only when boundaries match

**Stop:** do not treat `showcase/` as the product app; no EAS without explicit
authorization.

## Onboard an existing app

1. `$expo-development-foundation` → classify as existing app
2. `$expo-project-orientation` (mandatory first)
3. Emit stack card + ordered shortlist
4. **Stop** unless implementation is authorized
5. Proceed only with shortlisted domain skills; never `$expo-project-structure`
   restructure of a mature layout

**Stop:** monorepo style-only flatten; inventing credentials; foundation audit
by default.

## Add a route

Use `$expo-router` for route groups, typed dynamic parameters, tabs, modals,
form sheets, headers, and not-found behavior. Keep route files focused and
move shared code into existing consumer folders (do not invent a new tree layout
on mature apps).

## Add data

Use `$expo-data-fetching` for a typed boundary with loading, error, cancellation,
cache, and offline behavior. Keep secrets out of the client bundle. Full session
/ OAuth lifecycle → `$expo-auth` (not data-fetching alone).

## Add an authenticated screen

1. Existing app → `$expo-project-orientation` first
2. `$expo-auth` for session, secure storage, Auth Session / PKCE, logout
3. `$expo-linking` when deep-link / redirect schemes are involved
4. `$expo-data-fetching` for authenticated API clients and cache invalidation
5. `$expo-examples` for vendor-specific SDK patterns (Clerk, etc.)

**Stop:** tokens in AsyncStorage; secrets in `EXPO_PUBLIC_*`; inventing OAuth
client secrets.

## Add camera / media (permissioned capability)

1. Orientation if existing app
2. `$expo-permissions` for timing, privacy strings, denial UX, Expo Go limits
3. Package APIs only after permission policy is clear
4. `$expo-dev-client` when Expo Go is insufficient

**Stop:** permission prompt on first frame; store submit is not part of this
recipe.

## Publish OTA or submit to stores

OTA updates and store submissions are outside this skill catalog. Do not run
paid EAS or store actions by inference. Confirm explicit authorization,
credentials, and JS-only OTA safety (decision tree 5) first, then follow the
Expo/EAS vendor docs for the authorized action.

## Works in Expo Go, fails in build

1. `$expo-project-orientation` + `$expo-debugging`
2. Re-run Expo Go vs development-client tree
3. `$expo-dev-client` / config plugins / native modules as indicated
4. Local static checks before any cloud build

## Add a config-plugin dependency

1. Orientation if existing app
2. `$expo-config-env` for app config / plugins surface
3. `$expo-permissions` when privacy strings are required
4. Mark development client required; `$expo-dev-client` for runtime
5. New binary for native-breaking changes—not OTA alone

## SDK upgrade dry-run

1. `$expo-project-orientation` on existing apps
2. `$expo-upgrade` for version matrix and breaking changes
3. Plan-only until implementation is authorized
4. Consumer typecheck / doctor; no store submit by inference

## Use web code incrementally

Use `$expo-dom` for an isolated DOM Component and `$expo-web-to-native` for an
end-to-end migration. Keep props serializable and actions explicit.

## Native capability or release

Use `$expo-module` and `$expo-dev-client` for custom native code. Use
`$eas-hosting` only for explicitly authorized hosting deploys. Store and OTA
releases are outside this catalog.
