Role expert debug dervish

Systematic Debugging

Methodical approach to identifying and fixing software issues. Emphasises reproducibility, isolation, and verification over trial-and-error debugging. Covers browser DevTools, Node debugging, logging strategies, and Svelte-specific debugging techniques.

When This Skill Applies

Use this skill when:

• User reports something "not working"
• Investigating errors or unexpected behaviour
• Performance issues need diagnosis
• Setting up debugging infrastructure
• Explaining debugging approaches
• Teaching debugging techniques
• Questions about DevTools or debugging tools

The Debugging Methodology

Five-step process: Reproduce → Isolate → Diagnose → Fix → Verify

This applies to every debugging scenario without exception.

1. Reproduce

Make the bug happen reliably

Can't fix what you can't reproduce. First priority is finding reliable steps to trigger the issue.

Questions to ask:

• What exact steps produce the error?
• Does it happen every time or intermittently?
• What's the expected vs actual behaviour?
• What's the minimal reproduction case?

Document reproduction steps:

## To Reproduce

1. Navigate to `/dashboard`
2. Click "Load More" button
3. Scroll to bottom of page
4. Click "Load More" again

**Expected**: More items load
**Actual**: Page freezes, console shows error
**Frequency**: Happens every time on 2nd click

If intermittent:

• Look for race conditions
• Check network timing issues
• Consider state-dependent bugs
• Try multiple environments

2. Isolate (Root-Cause Depth)

Narrow down the cause — then go deeper.

Once reproducible, determine exactly where the problem originates. But don't stop at the first explanation. The first "cause" is often a symptom. Ask why repeatedly until you reach the structural root.

The Five Whys:

Bug: Users see stale data after updating their profile.

Why? → The cache isn't invalidated after the update.
Why? → The update function doesn't call cache.invalidate().
Why? → The caching layer was added after the update function was written.
Why? → There's no pattern ensuring new writes invalidate related caches.
Root: Missing cache invalidation convention. Fix the convention, not just this instance.

Depth vs speed: Not every bug warrants five whys. Use your judgement:

• Shallow fix appropriate: Typo, wrong variable name, missing null check
• Deep investigation warranted: Bug that could recur, affects multiple users, or reveals a pattern gap

Isolation techniques:

Binary search approach:

// Working at line 50?
console.log('Check 1:', data); // ✓ Data good here

// Working at line 75?
console.log('Check 2:', result); // ✗ Result undefined here

// Problem is between lines 50-75

Comment out code:

// Does removing this fix it?
// await someAsyncFunction();

// If yes, problem is in someAsyncFunction

Minimal reproduction:

// Strip away everything non-essential
// Original: 300 lines, complex state, multiple API calls
// Minimal: 20 lines that show the exact issue

async function minimalRepro() {
  const data = await fetch('/api/items');
  console.log(data); // undefined when expected array
}

Check assumptions:

// Assumption: API returns array
console.log(typeof data); // "object" - it's null!

// Assumption: User is logged in
console.log(user); // undefined - not logged in!

3. Diagnose

Understand why it's happening

• Runtime error: Type error, logic error, or environment issue?
• Test failure: Async issues, mock problems, environment differences, shared state?
• Logic bug: Off-by-one, wrong condition, state mutation, edge case?
• Performance: N+1 queries, large data, sync ops, memory leaks, inefficient algorithms?
• Production: Differences between local and production? Check logs, monitoring, external services.

4. Fix

Implement targeted solution

Now that you know the cause, fix it specifically.

Fix patterns:

Null/undefined checks:

// Before (crashes)
const count = items.length;

// After
const count = items?.length ?? 0;

Async timing:

// Before (race condition)
fetch('/api/data');
renderUI(); // Renders before data arrives

// After
const data = await fetch('/api/data');
renderUI(data);

State initialisation:

// Before (undefined on first render)
let items;

// After
let items = [];

Error boundaries:

// Before (crashes entire app)
const result = riskyOperation();

// After
try {
  const result = riskyOperation();
} catch (error) {
  console.error('Operation failed:', error);
  showErrorToUser('Something went wrong');
}

5. Verify

Confirm the fix works

Don't assume it's fixed. Test thoroughly.

Verification checklist:

• ✓ Original reproduction steps no longer produce error
• ✓ Expected behaviour now occurs
• ✓ No new errors introduced
• ✓ Edge cases still work
• ✓ Performance not degraded

Test edge cases:

// Fixed for normal case, but what about:
- Empty array
- Null values
- Very large datasets
- Network failures
- Simultaneous requests

Write regression test (see testing-obsessive skill):

it('should handle second "Load More" click', async () => {
  render(ItemList);

  await clickLoadMore();
  await clickLoadMore(); // This used to crash

  expect(screen.getAllByRole('listitem').length).toBeGreaterThan(10);
});

Additional resources

Tool-specific reference and worked bug patterns, loaded only when needed:

• tooling.md — runtime detection, Browser DevTools (Console/Network/Elements/Sources/Application/Performance tabs), Node.js debugging, Svelte-specific debugging (reactive statements, lifecycle, store debugging), logging strategies
• bug-patterns-and-performance.md — race conditions, stale closures, undefined reference errors, memory leaks, performance profiling, TypeScript debugging

Root-Cause vs Symptom Fixes

Symptom fix: Fixes the immediate problem but doesn't prevent recurrence. Root-cause fix: Addresses the structural issue that allowed the bug to exist.

// Symptom fix: Add null check where the crash happens
const name = user?.profile?.name ?? 'Unknown';

// Root-cause fix: Ensure profile is always populated at creation
async function createUser(data: CreateUserRequest): Promise<User> {
	return await db.users.create({
		...data,
		profile: { name: data.name } // Profile guaranteed at creation
	});
}

When to ship a symptom fix: When the root cause is expensive to fix and the symptom fix is safe. But always log the root cause as a follow-up task.

When to insist on root-cause fix: When the bug pattern could recur in other places, when data integrity is at risk, or when the symptom fix introduces its own complexity.

Debugging Checklist

Before:

• Can you reproduce it?
• Captured error/stack trace?
• Know when it started?

During:

• Isolated problem area?
• Checked assumptions?
• Reviewed recent changes?
• Documenting attempts?

After:

• Root cause fixed?
• Test added?
• Nothing else broken?
• Solution documented?