code-refactoring（代码重构）

Code Refactoring Skill

🎯 Purpose: Prevent Complexity Before It Becomes a Problem

Problem this solves: "I rarely know when to refactor, typically get caught by accident"

Solution: AI proactively monitors file sizes and complexity patterns, alerting you BEFORE making changes that would worsen the problem.

Works with: JavaScript/TypeScript (React, Node), Python, and general codebases.

🔔 CRITICAL FIRST STEP: Check for File Watcher Alerts

BEFORE doing anything else when this skill is invoked, ALWAYS check for unread alerts from the background file watcher:

node ~/.claude/plugins/marketplaces/custom-skills/code-refactoring/scripts/check-alerts.js

Why this is critical:

• The file watcher runs in the background monitoring code file sizes
• It writes alerts to watcher-alerts.json when files are edited and exceed thresholds
• These alerts tell you which files the user just edited that need refactoring attention
• You must check these alerts FIRST to provide real-time feedback

If alerts exist:

1. Display the alerts using the check-alerts.js script output
2. Ask the user if they want help refactoring the alerted files NOW or LATER
3. Then proceed with the rest of the skill workflow

If no alerts:

• Continue with normal skill workflow (checking file sizes manually, etc.)

User can also manually check alerts with: /check-refactor-alerts

🤖 AUTO-INVOKE CONDITIONS (AI Should Check These)

1. Before Editing Files (PREVENTIVE)

When AI is about to edit a file, check size first:

# Check file size before editing
wc -l [target-file]

# If result:
# 150-200 lines → ⚠️ WARN: Getting large, plan extraction
# 200-300 lines → 🚨 ALERT: Should split before adding more
# 300+ lines   → 🛑 STOP: Must refactor before editing

Example workflow:

User: "Add new feature to UserProfile.tsx"
AI: *Before editing, checks file size*
AI: *Sees 280 lines*
AI: *Auto-invokes code-refactoring skill*
Skill: "🚨 UserProfile.tsx is 280 lines. Before adding features:
       - Extract data to separate file
       - Split modal into separate component
       - Then add new feature safely"

2. During File Creation (PROACTIVE)

While creating new code file, track size:

• At 100 lines → ✅ Good, continue
• At 150 lines → ⚠️ Pause, suggest structure
• At 200 lines → 🚨 Stop, extract now

Example workflow:

AI: *Creating new DashboardService.py*
AI: *Reaches 150 lines*
AI: *Auto-invokes code-refactoring skill*
Skill: "⚠️ File reached 150 lines. Before continuing:
       1. Extract data/config to constants file
       2. Move helper functions to utils module
       3. Continue with cleaner structure"

3. After Reading Files (REACTIVE)

When AI reads a file and discovers size issues:

# After reading, if file >300 lines, suggest refactor
wc -l filename

# If >300 lines, proactively suggest refactor BEFORE user asks for changes

Example workflow:

User: "Take a look at the analytics module"
AI: *Reads analytics.py*
AI: *Sees 450 lines*
AI: *Auto-invokes code-refactoring skill*
Skill: "🛑 analytics.py is 450 lines (150 lines over limit).

       Refactor plan:
       1. Extract constants to analytics_config.py (50 lines)
       2. Create AnalyticsChart class (100 lines)
       3. Create AnalyticsProcessor class (120 lines)
       4. Keep Analytics as orchestrator (180 lines)

       Should I proceed with refactoring before making other changes?"

4. Pattern Detection (AUTOMATIC)

AI should automatically detect these patterns and invoke skill:

Universal Patterns (All Languages)

• ✅ Large data structures (>20 lines) → Suggest separate config/data file
• ✅ Complex conditionals (5+ branches) → Suggest strategy pattern or dispatch table
• ✅ Deeply nested logic (>3 levels) → Suggest extraction
• ✅ Repeated code blocks → Suggest function/method extraction
• ✅ Long parameter lists (>5 params) → Suggest config object or builder pattern

JavaScript/TypeScript/React Patterns

• ✅ 5+ data items hardcoded → Suggest data file extraction
• ✅ 4+ useState/useEffect hooks → Suggest custom hook extraction
• ✅ Modal/Dialog code → Suggest separate component
• ✅ Complex form (10+ fields) → Suggest separate form component
• ✅ Large switch statements → Suggest lookup table or component map

Python Patterns

• ✅ Class >300 lines → Suggest splitting into multiple classes
• ✅ Function >50 lines → Suggest breaking into smaller functions
• ✅ Module >500 lines → Suggest splitting into package
• ✅ Too many methods (>15) → Suggest composition or mixins
• ✅ Complex init (>20 lines) → Suggest factory method or builder

📏 File Size Guidelines (Language-Specific)

JavaScript/TypeScript/React Components

• 100-200 lines: ✅ Perfect for most components
• 200-300 lines: ⚠️ Good for complex components, watch closely
• 300+ lines: 🛑 MUST split into logical sub-components

Python Modules

• 150-250 lines: ✅ Good module size
• 250-400 lines: ⚠️ Consider splitting by responsibility
• 400+ lines: 🛑 Split into multiple modules or package

Python Classes

• 100-200 lines: ✅ Well-focused class
• 200-300 lines: ⚠️ Check single responsibility principle
• 300+ lines: 🛑 Split into multiple classes or use composition

General Files (Any Language)

• 150-250 lines: ✅ Maintainable size
• 250-400 lines: ⚠️ Getting complex
• 400+ lines: 🛑 Refactor needed

🔍 Refactoring Analysis Template

When skill is invoked, provide this analysis:

## Code Refactoring Analysis: [Filename]

**Current State:** [X] lines - [✅ Good / ⚠️ Warning / 🛑 Critical]

**Identified Issues:**
- File size issues (>200 or >300 lines)
- Complex nested logic / too many responsibilities
- Language-specific patterns (hooks, data arrays, classes, etc.)

**Refactoring Recommendations:**
1. Extract [what] to [new-file] (saves [X] lines, [X] mins)
2. Extract [what] to [new-file] (saves [X] lines, [X] mins)

**Result:** Main file → [target] lines | Risk: [Low/Med/High] | Time: [X] mins

**Proceed?** [Yes/No/Later/Review]

For complete template with all fields, see FORMS.md

🚨 Automatic Splitting Triggers

When ANY of these conditions occur, AUTO-INVOKE this skill:

Universal Triggers (All Languages)

1. ✅ File reaches 150 lines → Suggest extraction points
2. ✅ File reaches 200 lines → Alert and plan refactor
3. ✅ File reaches 300 lines → Stop and refactor first
4. ✅ Complex nested logic detected
5. ✅ Repeated code patterns found

JavaScript/TypeScript Triggers

1. ✅ Component reaches 150 lines → Extract next section
2. ✅ 3+ useState hooks → Extract to custom hook
3. ✅ Large data array → Move to data file immediately
4. ✅ Modal/Dialog code → Separate component from start
5. ✅ Complex form → Separate form component immediately

Python Triggers

1. ✅ Class reaches 250 lines → Plan class split
2. ✅ Function reaches 40 lines → Suggest breaking down
3. ✅ Module reaches 350 lines → Plan module split
4. ✅ 10+ methods in class → Consider composition
5. ✅ Complex init → Suggest factory pattern

📋 Quick Decision Matrix

When to extract what (Quick Reference):

JavaScript/TypeScript/React:

• Data file: 5+ items, arrays >20 lines, config objects
• Sub-component: Modal/dialog, complex list items, distinct UI sections
• Custom hook: 4+ useState, complex logic, reusable state

Python:

• Config file: 5+ variables, large data (>20 lines), magic numbers
• New class: Class >300 lines, multiple responsibilities, 10+ methods
• Utility function: Repeated logic, pure functions, generic operations
• New module: File >400 lines, distinct responsibilities, testable independently

For detailed extraction patterns and examples, see REFERENCE.md

💡 Proactive Suggestions

AI should auto-suggest at these milestones:

• New file: Monitor size, alert at 150 lines
• 100 lines: Looking good, will alert at 150
• 150 lines: ⚠️ Check if should extract before continuing
• 200 lines: 🚨 Recommend refactoring now
• Reading large file: Offer refactoring plan before changes

🔧 EXECUTION PHASE (After User Approval)

IMPORTANT: This skill can EXECUTE refactoring, not just suggest it!

Step 1: Present Plan and Get User Approval

After analyzing file size and creating refactoring plan, always ask for approval:

## Refactoring Plan Ready

**File:** [filename] ([X] lines)
**Target:** Reduce to <[target] lines

### Proposed Changes:
1. Extract [data/component/class] to [new-file] (saves [X] lines)
2. Extract [component/function] to [new-file] (saves [X] lines)
3. Extract [hook/utility] to [new-file] (saves [X] lines)

**Result:** Main file reduced to ~[X] lines
**Estimated time:** [X] minutes
**Risk level:** Low (all changes are git-committed incrementally)

### Options:

**✅ Yes** - Execute refactoring now (recommended)
  - I'll split the file step-by-step
  - Each step will be tested and committed
  - You can rollback if needed
  - Then I'll continue with your original request

**❌ No** - Skip refactoring
  - I'll continue with your edit
  - Warning: File will grow larger
  - May become harder to maintain

**⏰ Later** - Save plan for later
  - I'll save the plan to `refactoring-plan-[filename].md`
  - You can execute it manually later
  - I'll continue with your edit now

**📋 Review** - Show detailed step-by-step execution plan
  - I'll explain exactly what will happen
  - Then you can choose Yes/No/Later

**Your choice?**

Step 2: Execute Refactoring (Only if User Says "Yes")

When user approves, execute refactoring incrementally:

1. Create backup commit (safety first)
2. Verify starting state (lint, type-check, test must pass)
3. Execute step-by-step (extract → verify → commit for each change)
4. Final verification (full test suite)
5. Rollback automatically if any step fails

Key execution principles:

• ✅ Each step is atomic and committed separately
• ✅ Each step is verified before proceeding
• ✅ Automatic rollback on any failure
• ✅ User can cancel at any time

For detailed execution procedures, safety measures, rollback procedures, and examples, see REFERENCE.md → "EXECUTION PHASE - Detailed Procedures"

Step 3: Continue With Original Request

After execution (success or skip):

Refactoring: ✅ Complete (or ⏭️ Skipped)
Original request: "[user's request]"

Now proceeding with your request...

🔄 Integration with Other Skills

Works with:

• ui-ux-audit - Check file sizes during UI audits
• technical-writing - Document refactored structure
• qa-testing - Ensure tests cover refactored code

Triggers:

• Before major edits to large files
• During file creation
• After reading files >200 lines
• When detecting size/complexity patterns

🔍 CODEBASE AUDIT MODE (For Existing Technical Debt)

Use this mode when user asks to:

• "Audit the codebase" or "find large files"
• Check for "technical debt" or "existing problems"
• "Scan for refactoring opportunities"
• Review legacy/inherited codebase
• Plan major refactoring initiative

Audit Workflow (Brief):

1. Scan for oversized files by language/type
2. Categorize by severity (Critical >300, High 200-300, Medium 150-200, Good <150)
3. Prioritize using formula: (Size/100) + (Change Frequency × 2) + (Business Impact × 3)
4. Create refactoring roadmap with time estimates
5. Execute incrementally (one file per sprint)

Quick scan commands:

# Find oversized files by type
find src -name "*.tsx" -exec wc -l {} \; | awk '$1 > 200' | sort -rn

# Use helper script
bash scripts/check-size.sh src/ '*.tsx'

For detailed audit procedures, report templates, prioritization matrices, and batch analysis commands, see REFERENCE.md → "CODEBASE AUDIT MODE - Detailed Procedures"

📚 Language-Specific Details

For detailed refactoring patterns specific to each language, see:

• REFERENCE.md - Comprehensive language-specific guides
• FORMS.md - Refactoring templates and checklists

To check file size quickly:

# Use the helper script
bash ~/.claude/plugins/marketplaces/custom-skills/code-refactoring/scripts/check-size.sh [filename]

✅ Success Criteria

Refactoring is successful when:

1. ✅ Main file within target size for language
2. ✅ All extracted files well-organized
3. ✅ No broken imports or references
4. ✅ All tests pass
5. ✅ Functionality unchanged
6. ✅ Code more maintainable and readable
7. ✅ Clear separation of concerns
8. ✅ Documentation updated

Remember: Prevention > Cure

This skill prevents "caught by accident" scenarios by:

• Checking BEFORE editing (preventive)
• Alerting DURING creation (proactive)
• Suggesting AFTER reading (reactive)
• Enforcing patterns automatically

The goal: Never again discover an oversized, complex file by accident.

Multi-language support: Works across JavaScript/TypeScript, Python, and general codebases with language-specific patterns.