simulator-tester

Automated simulator testing with visual verification for closed-loop debugging.

How to Use This Agent

Natural language (automatic triggering):

• "Can you test my app with location simulation?"
• "Take a screenshot to verify this fix"
• "Check if the push notification handling works"
• "Navigate to Settings and take a screenshot"
• "Record a video of the app running"
• "Test my app's permission dialogs"

Explicit command:

/axiom:test-simulator

What It Does

Testing Capabilities

1. Screenshot capture — Visual verification and debugging
2. Video recording — Capture complex workflows
3. Location simulation — Test GPS-based features
4. Push notifications — Test notification handling without a server
5. Permission management — Grant/revoke permissions programmatically
6. Deep link navigation — Navigate to specific screens
7. App lifecycle — Launch, terminate, install apps
8. Status bar override — Clean screenshots for documentation
9. Log analysis — Monitor console for crashes/errors

Test Scenarios Fixed

• Visual verification → Screenshot capture + Claude analysis
• Location testing → Set coordinates + grant permission
• Push notifications → Send test payload + capture result
• Permission flows → Reset + grant + verify state
• Crash reproduction → Navigate + log capture + analysis

How It Works

Core Principle: The simulator is a verification mechanism for closed-loop debugging.

Workflow:

1. Check simulator state (boot if needed)
2. Set up test scenario (location, permissions, etc.)
3. Capture evidence (screenshots, video, logs)
4. Analyze results (Claude is multimodal!)
5. Report findings with clear pass/fail

Time savings: 60-75% faster iteration on visual bugs

Example Usage

Screenshot Verification

# User: "Take a screenshot to verify the login button fix"

# Agent automatically:
1. Checks if simulator is booted
2. Boots iPhone 16 Pro if needed
3. Navigates to login screen (if deep link available)
4. Waits for UI to render
5. Captures screenshot
6. Analyzes image: "The login button appears centered and properly sized"

Location Testing

# User: "Test my app with location set to San Francisco"

# Agent automatically:
1. Sets location: xcrun simctl location booted set 37.7749 -122.4194
2. Grants location permission: xcrun simctl privacy booted grant location-always <bundle-id>
3. Launches app
4. Takes screenshot of map view
5. Verifies location marker appears at SF coordinates

Push Notification Testing

# User: "Send a test push notification and screenshot the result"

# Agent automatically:
1. Creates test payload JSON
2. Sends push: xcrun simctl push booted <bundle-id> payload.json
3. Waits for notification to appear
4. Captures screenshot showing notification
5. Checks logs for notification handling

Key Features

Visual Verification

• Claude can read screenshots (multimodal analysis)
• Before/after comparison for fixes
• UI state verification
• Error message capture

Test Automation

• Programmatic navigation via deep links
• Automated test scenario setup
• Background process management
• Log capture and analysis

Integration

• Works with /axiom:screenshot for quick captures
• Integrates with deep-link-debugging skill for navigation
• Referenced in enhanced debugging skills

Common Scenarios

Scenario	Capabilities Used
Visual bug fix	Screenshot + navigation + analysis
Location feature	Location simulation + permissions
Push handling	Push notification + log capture
Permission dialog	Permission management + screenshot
Crash reproduction	App lifecycle + log capture
App Store screenshots	Status bar override + navigation

Requirements

• Xcode with iOS Simulator installed
• App built for simulator
• Optionally: Deep links for navigation (see deep-link-debugging skill)

Related Tools

• /axiom:screenshot — Quick screenshot capture without full testing
• deep-link-debugging skill — Add debug-only deep links for navigation
• xcode-debugging skill — Environment-first debugging
• swiftui-debugging skill — SwiftUI-specific debugging with simulator verification
• memory-debugging skill — Memory leak detection with visual verification

Real-World Impact

Before: Make fix → rebuild → manually navigate 5 screens → visually check → 3 minutes After: Make fix → rebuild → agent navigates + screenshots → Claude verifies → 1 minute

Key insight: Automated visual verification enables closed-loop debugging.