axiom-testflight-triage（Axiom 试飞分类）

TestFlight Crash & Feedback Triage

Overview

Systematic workflow for investigating TestFlight crashes and reviewing beta feedback using Xcode Organizer. Core principle: Understand the crash before writing any fix — 15 minutes of triage prevents hours of debugging.

Red Flags — Use This Skill When

• "A beta tester said my app crashed"
• "I see crashes in App Store Connect metrics but don't know how to investigate"
• "Crash logs in Organizer aren't symbolicated"
• "User sent a screenshot of a crash but I can't reproduce it"
• "App was killed but there's no crash — just disappeared"
• "TestFlight feedback has screenshots I need to review"

Decision Tree — Start Here

"A user reported a crash"

1. Open Xcode Organizer (Window → Organizer → Crashes tab)
2. Select your app from the left sidebar
3. Find the build version the user was running
4. Is the crash symbolicated?

- YES (you see function names) → Go to Reading the Crash Report - NO (you see hex addresses like 0x100abc123) → Go to Symbolication Workflow

1. Can you identify the crash location?

- YES → Go to Common Crash Patterns - NO → Go to Claude-Assisted Interpretation

"App was killed but no crash report"

Not all terminations produce crash reports. Check for:

1. Jetsam reports — System killed app due to memory pressure

- Organizer shows these separately from crashes - Look for high pageOuts value

1. Watchdog termination — Main thread blocked too long

- Exception code 0x8badf00d ("ate bad food") - Happens during launch (>20s) or background tasks (>10s)

1. MetricKit diagnostics — On-device termination reasons

- Requires MetricKit integration in your app

→ See Terminations Without Crash Reports

"I want to review TestFlight feedback"

1. Xcode Organizer → Feedback tab (next to Crashes)
2. Or App Store Connect → My Apps → [App] → TestFlight → Feedback

→ See Feedback Triage Workflow

Xcode Organizer Walkthrough

Opening the Organizer

Window → Organizer (or ⌘⇧O from Xcode)

UI Layout

┌─────────────────────────────────────────────────────────────────┐
│ [Toolbar: Time Period ▼] [Version ▼] [Product ▼] [Release ▼]   │
├──────────┬──────────────────────────┬───────────────────────────┤
│ Sidebar  │     Crashes List         │       Inspector           │
│          │                          │                           │
│ • Crashes│  ┌─────────────────────┐ │  Distribution Graph       │
│ • Energy │  │ syncFavorites crash │ │  ┌─────────────────────┐  │
│ • Hang   │  │ 21 devices • 7 today│ │  │ ▄ ▄▄▄ v2.0          │  │
│ • Disk   │  └─────────────────────┘ │  │ ▄▄▄▄▄ v2.0.1        │  │
│ Feedback │  ┌─────────────────────┐ │  └─────────────────────┘  │
│          │  │ Another crash...    │ │                           │
│          │  └─────────────────────┘ │  Device Distribution      │
│          │                          │  OS Distribution          │
│          ├──────────────────────────┤                           │
│          │     Log View             │  [Feedback Inspector]     │
│          │  (simplified crash view) │  Shows tester feedback    │
│          │                          │  for selected crash       │
└──────────┴──────────────────────────┴───────────────────────────┘

Key Features

Crash Entry Badges

Crashes in the list show badges indicating origin:

The Triage Questions Workflow

Before diving into code, ask yourself these questions (from WWDC):

Question 1 — How Long Has This Been an Issue

→ Check the inspector's graph area on the right → Graph legend shows which versions are affected → Look for when the crash first appeared

Question 2 — Is This Affecting Production or Just TestFlight

→ Use the Release filter in toolbar → Select "Release" to see App Store crashes only → Select "TestFlight" for beta crashes only

Question 3 — What Was the User Doing

→ Open the Feedback Inspector (right panel) → Check for tester comments describing their actions → Context clues: network state, battery level, disk space

Using the Feedback Inspector

When a crash has associated TestFlight feedback, you'll see a feedback icon in the crashes list. Click it to open the Feedback Inspector.

Each feedback entry shows:

Example insight from WWDC: A tester commented "I was going through a tunnel and hit the favorite button. A few seconds later, it crashed." This revealed a network timeout issue — the crash occurred because a 10-second timeout was too short for poor network conditions.

Opening Crash in Project

1. Select a crash in the list
2. Click Open in Project button
3. Xcode opens with:

- Debug Navigator showing backtrace - Editor highlighting the exact crash line

Sharing Crashes

1. Select a crash
2. Click Share button in toolbar
3. Options:

- Copy link to share with team - Add to your to-do list

1. When teammate clicks link, Organizer opens focused on that specific crash

Symbolication Workflow

Why Crashes Aren't Symbolicated

Crash reports show raw memory addresses until matched with dSYM files (debug symbol files). Xcode handles this automatically when:

• You archived the build in Xcode (not command-line only)
• "Upload symbols to Apple" was enabled during distribution
• The dSYM is indexed by Spotlight

Quick Check: Is It Symbolicated?

In Organizer, look at the stack trace:

Manual Symbolication

If automatic symbolication failed:

# 1. Find the crash's build UUID (shown in crash report header)
#    Look for "Binary Images" section, find your app's UUID

# 2. Find matching dSYM
mdfind "com_apple_xcode_dsym_uuids == YOUR-UUID-HERE"

# 3. If not found, check Archives
ls ~/Library/Developer/Xcode/Archives/

# 4. Symbolicate a specific address
xcrun atos -arch arm64 \
  -o MyApp.app.dSYM/Contents/Resources/DWARF/MyApp \
  -l 0x100000000 \
  0x0000000100abc123

# 5. Symbolicate an entire crash log (parse + symbolicate in one step)
# Note: crashlog is an LLDB Python script, not a compiled tool.
# Works via xcrun but may not be available in all Xcode configurations.
xcrun crashlog MyCrash.ips

atos vs crashlog: Use atos for individual addresses (always available). Use crashlog to parse and symbolicate an entire crash report at once — it handles the full Binary Images section and resolves all addresses automatically.

Common Symbolication Failures

Preventing Symbolication Issues

# Verify dSYM exists after archive
ls ~/Library/Developer/Xcode/Archives/YYYY-MM-DD/MyApp*.xcarchive/dSYMs/

# Verify UUID matches
dwarfdump --uuid MyApp.app.dSYM

Reading the Crash Report

Key Fields (What Actually Matters)

Reading the Stack Trace

The crashed thread's stack trace reads top to bottom:

• Frame 0 = Where the crash occurred (most specific)
• Lower frames = What called it (call chain)
• Look for your code = Frames with your app/framework name

Thread 0 Crashed:
0   libsystem_kernel.dylib    __pthread_kill + 8        ← System code
1   libsystem_pthread.dylib   pthread_kill + 288        ← System code
2   libsystem_c.dylib         abort + 128               ← System code
3   MyApp                     ViewController.loadData() ← YOUR CODE (start here)
4   MyApp                     ViewController.viewDidLoad()
5   UIKitCore                 -[UIViewController _loadView]

Start at frame 3 — the first frame in your code. Work down to understand the call chain.

Example: Interpreting a Real Crash

Exception Type:  EXC_BAD_ACCESS (SIGSEGV)
Exception Codes: KERN_INVALID_ADDRESS at 0x0000000000000010

Thread 0 Crashed:
0   MyApp    0x100abc123 UserManager.currentUser.getter + 45
1   MyApp    0x100abc456 ProfileViewController.viewDidLoad() + 123
2   UIKitCore 0x1a2b3c4d5 -[UIViewController loadView] + 89

Translation:

• EXC_BAD_ACCESS with KERN_INVALID_ADDRESS = Tried to access invalid memory
• Address 0x10 = Very low address, almost certainly nil dereference
• Crashed in currentUser.getter = Accessing a property that was nil
• Called from ProfileViewController.viewDidLoad() = During view setup

Likely cause: Force-unwrapping an optional that was nil, or accessing a deallocated object.

Common Crash Patterns

EXC_BAD_ACCESS (SIGSEGV / SIGBUS)

What it means: Accessed memory that doesn't belong to you.

Common causes in Swift:

// Before (crashes if user is nil)
let name = user!.name

// After (safe)
guard let user = user else {
    logger.warning("User was nil in ProfileViewController")
    return
}
let name = user.name

EXC_CRASH (SIGABRT)

What it means: App deliberately terminated itself.

Common causes:

Debug tip: Look at "Application Specific Information" section — it usually contains the actual error message.

Watchdog Termination (0x8badf00d)

What it means: Main thread was blocked too long and the system killed your app.

Time limits:

Common causes:

• Synchronous network request on main thread
• Synchronous file I/O on main thread
• Deadlock between queues
• Expensive computation blocking UI

// Before (blocks main thread — will trigger watchdog)
let data = try Data(contentsOf: largeFileURL)
processData(data)

// After (offload to background)
Task.detached {
    let data = try Data(contentsOf: largeFileURL)
    await MainActor.run {
        self.processData(data)
    }
}

Jetsam (Memory Pressure Kill)

What it means: System terminated your app to free memory. No crash report — just gone.

Symptoms:

• App "disappears" without any crash
• Jetsam report in Organizer (separate from crashes)
• High pageOuts value in report
• Often happens during photo/video processing or large data operations

Investigation:

1. Profile with Instruments → Allocations
2. Look for memory spikes during the reported operation
3. Check for image caching without size limits
4. Look for large data structures kept in memory

Common fixes:

• Use autoreleasepool for batch processing
• Implement image cache with memory limits
• Stream large files instead of loading entirely
• Release references to large objects when backgrounded

Terminations Without Crash Reports

When users report "the app just closed" but you find no crash:

The Terminations Organizer

The Terminations Organizer (separate from Crashes) shows trends of app terminations that aren't programming crashes:

Window → Organizer → Terminations (in sidebar)

Key insight: Compare termination rates against previous versions to find regressions. A spike in memory terminations after a release indicates a memory leak or increased footprint.

Check for Jetsam

1. Organizer → Select app → Look for "Disk Write Diagnostics" or "Hang Diagnostics"
2. These aren't crashes but system-initiated terminations

Check for Background Termination

Apps can be terminated in background for:

• Memory pressure (jetsam)
• CPU usage while backgrounded
• Background task timeout

Ask the User

If no reports exist:

1. "Was the app in the foreground when it closed?"
2. "Did you see any error message?"
3. "What were you doing right before it happened?"
4. "How long had the app been open?"

Enable Better Diagnostics with MetricKit

MetricKit crash diagnostics are now delivered on the next app launch (not aggregated daily). This gives you faster access to crash data.

import MetricKit

class MetricsManager: NSObject, MXMetricManagerSubscriber {

    static let shared = MetricsManager()

    func startListening() {
        MXMetricManager.shared.add(self)
    }

    func didReceive(_ payloads: [MXMetricPayload]) {
        // Process performance metrics
    }

    func didReceive(_ payloads: [MXDiagnosticPayload]) {
        for payload in payloads {
            // Crash diagnostics — delivered on next launch
            if let crashDiagnostics = payload.crashDiagnostics {
                for crash in crashDiagnostics {
                    // Process crash diagnostic
                    print("Crash: \(crash.callStackTree)")
                }
            }

            // Hang diagnostics
            if let hangDiagnostics = payload.hangDiagnostics {
                for hang in hangDiagnostics {
                    print("Hang duration: \(hang.hangDuration)")
                }
            }
        }
    }
}

When to use MetricKit vs Organizer:

Claude-Assisted Interpretation

Using the Crash Analyzer Agent

For automated crash analysis, use the crash-analyzer agent:

/axiom:analyze-crash

Or trigger naturally:

• "Analyze this crash log"
• "Parse this.ips file: ~/Library/Logs/DiagnosticReports/MyApp.ips"
• "Why did my app crash? Here's the report..."

The agent will:

1. Parse the crash report (JSON.ips or text.crash format)
2. Check symbolication status
3. Categorize by crash pattern (null pointer, Swift runtime, watchdog, jetsam, etc.)
4. Generate actionable analysis with specific next steps

Effective Prompts

Basic interpretation:

Here's a crash report from my iOS app. Help me understand:
1. What type of crash is this?
2. Where in my code did it crash?
3. What's the likely cause?

[paste full crash report]

With context (better results):

My TestFlight app crashed. Here's what I know:

- User was [describe action, e.g., "tapping the save button"]
- iOS version: [from crash report]
- Device: [from crash report]

Crash report:
[paste full crash report]

The relevant code is in [file/class name]. Help me understand the cause.

What to Include

What Claude Can Help With

• Interpreting exception types and codes
• Identifying likely cause from stack trace
• Explaining unfamiliar system frames
• Suggesting where to add logging
• Proposing fix patterns

What Requires Your Judgment

• Whether the suggested fix is correct for your architecture
• How to reproduce the crash locally
• Priority relative to other bugs
• Whether it's a regression or long-standing issue

Feedback Triage Workflow

Where to Find Feedback

Xcode Organizer (recommended): Window → Organizer → Select app → Feedback tab

App Store Connect: My Apps → [App] → TestFlight → Feedback

What's in Each Feedback Entry

Triage Workflow

1. Sort by recency — Newest first, unless investigating specific issue
2. Scan screenshots — Visual issues are immediately apparent
3. Read comments — User's description and context
4. Check version — Is this fixed in a newer build?
5. Categorize:

Limitations

• No direct reply — TestFlight doesn't support responding to feedback
• Screenshots only — No video recordings
• Limited context — Users often don't explain what they were trying to do

MCP-Powered Feedback Access

If asc-mcp is configured, you can access crash diagnostics and tester data programmatically:

Limitation: TestFlight text feedback and screenshots are NOT available via the App Store Connect API. Use Xcode Organizer or the ASC web dashboard for feedback content.

Setup: /skill axiom-asc-mcp

Getting More Context

If feedback is unclear and the tester is reachable:

• Contact through TestFlight group email
• Add in-app feedback mechanism with more detail capture
• Include reproduction steps prompt in your TestFlight notes

Pressure Scenarios

Scenario 1: "VIP user says app crashes constantly, but I can't find any crash reports"

Pressure: Important stakeholder, no evidence, tempted to dismiss with "works for me"

Correct approach:

1. Verify they're on TestFlight (not App Store, not dev build)
2. Confirm they've consented to share diagnostics (Settings → Privacy → Analytics)
3. Check for jetsam reports (kills without crash reports)
4. Check crash reports for their specific device/OS combination
5. Ask for specific reproduction steps
6. If still nothing: request screen recording of the issue

Response template:

"I've checked our crash reports and don't see crashes matching your description yet. To help investigate: (1) Could you confirm you're running the TestFlight version? (2) What exactly happens — does the app close suddenly, freeze, or show an error? (3) What were you doing right before? This will help me find the issue."

Why this matters: "Works for me" destroys trust. Investigate thoroughly before dismissing.

Scenario 2: "Crash rate spiked after latest TestFlight build, need to fix ASAP"

Pressure: Time pressure, tempted to guess at fix based on code changes

Correct approach:

1. Open Organizer → Crashes → Filter to the new build
2. Group crashes by exception type (look for the dominant signature)
3. Identify the #1 crash by frequency
4. Symbolicate and read the crash report fully
5. Understand the cause before writing any fix
6. If possible, reproduce locally
7. Fix the verified cause, not a guess

Why this matters: Rushed guesses often introduce new bugs or miss the real issue. 15 minutes of proper triage prevents hours of misdirected debugging.

Scenario 3: "Crash report is symbolicated but I still don't understand it"

Pressure: Tempted to ignore it or make random changes hoping it helps

Correct approach:

1. Paste full crash report into Claude with context
2. Ask for interpretation, not just "fix this"
3. Research exception type if unfamiliar
4. If still unclear after research, add logging around the crash site:

func suspectFunction() {
    logger.debug("Entering suspectFunction, state: \(debugDescription)")
    defer { logger.debug("Exiting suspectFunction") }

    // ... existing code ...
}

1. Ship instrumented build to TestFlight
2. Wait for reproduction with better context

Why this matters: Understanding beats guessing. Logging beats speculation. It's okay to say "I need more information" rather than shipping a random change.

Quick Reference

Organizer Keyboard Shortcuts

Common Exception Codes

Crash Report Sections

Resources

WWDC: 2018-414, 2020-10076, 2020-10078, 2020-10081, 2021-10203, 2021-10258

Docs: /xcode/diagnosing-issues-using-crash-reports-and-device-logs, /xcode/examining-the-fields-in-a-crash-report, /xcode/adding-identifiable-symbol-names-to-a-crash-report, /xcode/identifying-the-cause-of-common-crashes, /xcode/identifying-high-memory-use-with-jetsam-event-reports

Skills: axiom-memory-debugging, axiom-xcode-debugging, axiom-swift-concurrency, axiom-lldb (reproduce and investigate interactively), axiom-asc-mcp (programmatic ASC access)

Agents: crash-analyzer (automated crash log parsing and analysis)