test-mobile-app（测试移动应用程序）

Mobile App Testing Skill

This skill enables Claude to perform end-to-end mobile application testing:

1. Analyze the app structure and infer user-facing functionality
2. Generate use cases from an end-user perspective
3. Write concrete test scenarios with expected results
4. Execute tests via Appium + Android emulator (or interpret results statically)
5. Produce a structured HTML/Markdown test report

Phase 1 — App Analysis

What to collect

Before generating use cases, gather as much context as possible:

• Source code (Android/Java/Kotlin, iOS/Swift, React Native, Flutter)
• APK file — use androguard to extract Activity list, permissions, Manifest
• Screenshots — analyze UI from images
• Description — what the app does, target audience

APK Analysis (Android)

Read scripts/analyze_apk.py for full script. Quick usage:

python3 scripts/analyze_apk.py path/to/app.apk

Outputs: package name, activities, permissions, strings → feeds into use case generation.

Source Code Analysis

If source is available, scan for:

• Screen/Activity/Fragment/Page names → each is a potential use case surface
• Navigation graphs (React Navigation, NavController)
• API endpoints called (network requests)
• Form fields, validation logic
• Authentication flows

Phase 2 — Use Case Generation

Methodology

Think from the perspective of a real end user — not a developer. Ask: *"What would a person actually do with this app?"*

Use case format:

UC-<N>: <Short Title>
Actor: End User
Precondition: <What must be true before this action>
Steps:
  1. <action>
  2. <action>
  ...
Expected outcome: <what the user sees/gets>
Priority: High / Medium / Low

Use Case Categories to Always Cover

1. Onboarding — first launch, tutorial, permissions prompt
2. Authentication — registration, login, logout, password reset
3. Core Feature Flow — the primary value action of the app (1-3 flows)
4. Data Entry — any form: required fields, validation, error states
5. Navigation — bottom nav, back button, deep links
6. Empty States — what happens when there's no data
7. Error Handling — no internet, server error, invalid input
8. Settings / Profile — change preferences, update data
9. Notifications — if the app uses push notifications
10. Accessibility — basic: is text readable, are tap targets big enough

Aim for 15–30 use cases depending on app complexity.

Phase 3 — Test Scenario Writing

For each use case, write a test scenario:

TEST-<N>: <Title>
Related UC: UC-<N>
Type: Functional | UI | Regression | Smoke
Steps:
  1. Launch app
  2. <specific action with exact input data>
  3. ...
Assertions:
  - Element <locator> is visible
  - Text "<expected>" is displayed
  - Screen navigates to <ScreenName>
  - No crash / error dialog
Expected Result: PASS / FAIL criteria

Test Types to Include

Expo / React Native — Local Backend Setup

When the user's backend runs on the same machine as the development environment, the mobile app cannot use localhost or 127.0.0.1 — on a physical device or emulator, those addresses resolve to the device itself, not the host computer.

Find the host machine's IP

# macOS / Linux
ifconfig | grep "inet " | grep -v 127.0.0.1

# Windows (PowerShell)
ipconfig

Use the LAN IP (e.g. 192.168.1.15). The device and the machine must be on the same Wi-Fi network.

Android emulator shortcut: 10.0.2.2 is a special alias that always points to the host machine, so you don't need the LAN IP when testing on the built-in Android emulator.

Configure the Expo app

Expo supports .env files with the EXPO_PUBLIC_ prefix:

1. Create .env.local in the project root: EXPO_PUBLIC_API_URL=http://192.168.1.15:5000
2. Add .env.local to .gitignore (machine-specific setting).
3. Use the variable in code: ` const response = await fetch(${process.env.EXPO_PUBLIC_API_URL}/users); `

For multi-environment setups, use app.config.js:

// app.config.js
export default ({ config }) => ({
  ...config,
  extra: {
    apiUrl: process.env.EXPO_PUBLIC_API_URL || 'http://localhost:5000',
  },
});
// Access via: Constants.expoConfig.extra.apiUrl  (expo-constants)

Configure the backend

Two things must be set on the server side:

Example for Express:

const cors = require('cors');
app.use(cors()); // or restrict to: { origin: 'http://192.168.1.15:8081' }
app.listen(5000, '0.0.0.0', () => console.log('listening on all interfaces'));

Tunneling fallback (ngrok / Expo tunnel)

If direct LAN access fails (VPN, restrictive router, office firewall), use tunneling:

# Expo built-in tunnel (wraps ngrok)
npx expo start --tunnel

This creates a public HTTPS URL that forwards traffic to the local server. It's slower than LAN but works through any network. Update EXPO_PUBLIC_API_URL to the tunnel URL while using it.

Testing checklist for local backend scenarios

• Backend bound to 0.0.0.0, not 127.0.0.1
• CORS configured on the server
• EXPO_PUBLIC_API_URL set to LAN IP (or http://10.0.2.2:<port> for Android emulator)
• Device and machine on the same Wi-Fi (for physical device)
• No firewall blocking the backend port on the host machine
• API endpoints respond to direct curl http://<host-ip>:<port>/health from terminal before running app tests

Phase 4 — Test Execution

Environment Setup

Read references/setup-appium.md for full Appium + emulator setup.

Quick check:

python3 scripts/check_environment.py

This verifies: adb, emulator, Appium server, Python client.

Running Tests

# Run all tests
python3 scripts/run_tests.py --apk path/to/app.apk --output results/

# Run smoke tests only
python3 scripts/run_tests.py --apk path/to/app.apk --suite smoke --output results/

# Run on specific device
python3 scripts/run_tests.py --apk path/to/app.apk --device emulator-5554 --output results/

Test Execution Without Emulator (Static Mode)

If no emulator is available (which is common — most users won't have Appium set up), Claude can still provide significant value:

1. Analyze source code / screenshots / APK statically
2. Generate use cases and write all test scenarios
3. Mark execution status as MANUAL_REQUIRED
4. Generate a comprehensive report with all test cases ready to be run manually
5. Provide step-by-step manual testing instructions the user can follow

This is the most common execution path — don't treat it as a fallback. Make the static report just as polished and detailed as the automated one.

Use --static flag:

python3 scripts/run_tests.py --static --tests tests.json --output results/

Phase 5 — Report Generation

python3 scripts/generate_report.py --results results/ --output test_report.html

Report includes:

• Summary: total tests, passed, failed, skipped
• Per-test details: steps, assertions, actual vs expected, screenshots
• Use case coverage matrix
• Issues found (with severity: Critical / Major / Minor)
• Environment info (device, OS, app version)

Read references/report-template.md for report structure details.

Workflow Summary

1. Receive app (APK / source / description / screenshots)
        ↓
2. Run analyze_apk.py OR inspect source code
        ↓
3. Generate use cases (UC-1...UC-N) — show to user, ask for feedback
        ↓
4. Write test scenarios (TEST-1...TEST-N) — derive from use cases
        ↓
5. Check environment (check_environment.py)
        ↓
6a. Emulator available → run_tests.py → capture results
6b. No emulator → static mode → mark for manual execution
        ↓
7. generate_report.py → HTML report → present to user

Important Notes

• Always show use cases to the user before writing tests — they know their app best.
• Locators: Prefer accessibility id > resource-id > xpath. Never use index-based xpath.
• Waits: Always use explicit waits (WebDriverWait), never time.sleep.
• Screenshots: Capture on every assertion failure automatically.
• Crash detection: After every interaction, check for crash dialogs (the check_for_crash() function in scripts/run_tests.py handles this automatically).
• Language: Generate use cases and reports in the language the user is using.