Fix

Fix bugs efficiently with intelligent routing — simple fixes applied immediately, complex issues diagnosed systematically before touching code.

What This Skill Does

The Challenge: Not all bugs are equal. A typo and a race condition require completely different approaches. Using heavy investigation for simple bugs wastes time; applying quick fixes to complex bugs creates more bugs.

The Solution: Fix skill analyzes bug reports, classifies complexity, and routes to the appropriate fixing strategy — direct fix for obvious issues, systematic investigation for complex ones. Prevents the common mistake of guessing at root causes for intermittent or systemic bugs.

Activation

Implicit: Activates when user reports a bug, error, or unexpected behavior.

Explicit: Activate via prompt:

/fix the checkout button throws a 500 error when cart is empty

Capabilities

1. Bug Classification

Assess complexity before deciding fix approach.

Complexity tiers:

Tier	Characteristics	Approach
Trivial	Typo, wrong value, obvious error	Direct fix
Simple	Single function, clear root cause	Fix + test
Medium	Multiple files, needs investigation	Diagnose → fix → test
Complex	Intermittent, cross-system, race condition	Deep investigation → targeted fix

2. Error Message Parsing

Extract structured information from error messages and stack traces.

Parsed elements:

Error type and message
Stack trace file and line number
Affected component or function
Request context (URL, user, timestamp)

Quick wins from error messages:

Cannot read property 'X' of undefined → Null check needed
ETIMEDOUT → Network or connection pool issue
UNIQUE constraint failed → Duplicate entry handling missing

3. Systematic Investigation

For complex bugs, trace from symptom to root cause.

Investigation sequence:

Reproduce the bug consistently
Identify the earliest point of failure in the call stack
Trace data flow to find where state diverges
Isolate the fix to minimum change
Verify fix doesn’t break adjacent behavior

4. Regression Prevention

Ensure the fix includes safeguards against recurrence.

Prevention measures:

Unit test covering the bug scenario
Error handling for the failure mode
Monitoring alert for the metric that should have caught this
Code comment explaining non-obvious defensive code

Prerequisites

Error message, stack trace, or bug description
Steps to reproduce (ideally)
Environment where bug occurs (production, staging, local)

Configuration

No configuration required. Operates through codebase analysis and targeted fixes.

Debug skill integration: For production issues, fix delegates investigation to debugger agent.

Best Practices

1. Reproduce before fixing A bug you can’t reproduce is a bug you might not fix. Insist on reproduction steps.

2. Fix the root cause, not the symptom Adding a null check around a symptom hides the real problem. Find where the null originates.

3. One bug, one commit Keep bug fixes isolated. Bundling fixes with features makes bisecting future regressions impossible.

Common Use Cases

Use Case 1: API 500 Error

Scenario: Users report intermittent 500 errors on the payment endpoint.

Investigation workflow:

Check error logs for stack trace
Identify error type (DB timeout? Validation failure? Third-party API?)
Find earliest stack frame in application code
Check for missing null handling, race conditions, or missing error handling
Apply targeted fix + add error logging

Output: Fix with explanation + test case + monitoring recommendation.

Use Case 2: UI State Bug

Scenario: Shopping cart count shows wrong number after adding items.

Investigation workflow:

Identify state update logic
Check if state is mutated vs replaced (common React bug)
Verify event handler calls correct state updater
Check for stale closure or missing dependency in useEffect

Output: State fix + unit test for cart count behavior.

Troubleshooting

Issue: Fix works locally but not in production Solution: Check environment differences (env vars, DB version, feature flags). Production bugs often have environmental root causes.

Issue: Fix breaks other tests Solution: The fix may have exposed a pre-existing assumption. Review related test failures as new bugs, not regressions.

Cook - Full feature implementation
Sequential Thinking - Step-by-step bug investigation
Analytics - Monitor fix effectiveness

/fix — Primary bug fixing command
/ckm:sequential-thinking — Deep investigation workflow
/ckm:cook — Full implementation after fix