detect-code-smells（检测代码气味）

Detect Code Smells

Detect common code smells and anti-patterns in code, providing immediate feedback on quality issues.

Code Smell Categories

1. Complexity Smells

• Long Method: Functions/methods > 50 lines
• Long Parameter List: > 4 parameters
• Complex Conditionals: Deeply nested if/else, complex boolean expressions
• High Cyclomatic Complexity: > 10 branches
• Deep Nesting: > 4 levels of indentation

2. Duplication Smells

• Duplicate Code: Repeated code blocks
• Similar Functions: Functions with nearly identical logic
• Magic Numbers: Hardcoded numbers without explanation
• String Duplication: Repeated string literals

3. Naming Smells

• Unclear Names: Single letter variables (except loop counters)
• Hungarian Notation: Unnecessary type prefixes
• Inconsistent Naming: Mixed camelCase/snake_case
• Abbreviated Names: Unclear abbreviations (mgr, ctx, tmp)
• Misleading Names: Name doesn't match behavior

4. Object-Oriented Smells

• God Class: Class > 500 lines or too many responsibilities
• Data Class: Class with only getters/setters
• Feature Envy: Method uses more of another class than its own
• Inappropriate Intimacy: Classes too dependent on internal details
• Lazy Class: Class doing too little to justify existence

5. Functional Smells

• Side Effects: Function modifies external state unexpectedly
• Non-Pure Functions: Functions with hidden dependencies
• Mutability Issues: Unexpected mutation of objects
• Callback Hell: Deeply nested callbacks

6. Architecture Smells

• Circular Dependencies: Module A depends on B, B depends on A
• Missing Abstraction: Concrete implementations without interfaces
• Tight Coupling: Hard dependencies on specific implementations
• Leaky Abstraction: Implementation details exposed through interface

Detection Process

1. Parse file - Analyze syntax tree and structure
2. Identify patterns - Look for known code smell patterns
3. Calculate metrics - Measure complexity, length, duplication
4. Assess severity - Determine impact of each smell
5. Generate report - Provide actionable feedback

Analysis Techniques

• Abstract Syntax Tree (AST) parsing
• Pattern matching against known smells
• Metric calculation (LOC, complexity, coupling)
• Comparison with language-specific best practices
• Context-aware analysis (test files have different standards)

Return Value

Return object:

{
  "file": "path/to/file.js",
  "language": "javascript",
  "overallScore": 7.5,
  "smells": [
    {
      "type": "Long Method",
      "severity": "warning",
      "location": {
        "line": 42,
        "endLine": 95,
        "function": "processUserData"
      },
      "description": "Function 'processUserData' is 53 lines long",
      "suggestion": "Extract smaller functions for validation, transformation, and persistence",
      "impact": "Harder to understand, test, and maintain",
      "effort": "medium"
    },
    {
      "type": "Complex Conditional",
      "severity": "warning",
      "location": {
        "line": 67,
        "column": 8
      },
      "description": "Nested conditional with 5 levels of nesting",
      "suggestion": "Extract conditions into well-named boolean variables or separate functions",
      "impact": "Difficult to understand logic flow",
      "effort": "small"
    }
  ],
  "metrics": {
    "linesOfCode": 234,
    "averageComplexity": 4.2,
    "maxComplexity": 12,
    "duplicationPercentage": 8.5
  },
  "recommendations": [
    "Extract 'processUserData' into smaller single-purpose functions",
    "Replace complex conditional at line 67 with early returns",
    "Consider extracting repeated validation logic into a helper function"
  ]
}

Examples

Example 1: Detecting Long Method

Context: User opens user-service.js containing a 150-line function

Autonomous Invocation: Skill automatically analyzes the file and reports:

Code Smell Detected: Long Method

Function 'handleUserRegistration' (lines 23-172) is 150 lines long.

Impact: This function is doing too many things, making it:
- Hard to understand at a glance
- Difficult to test thoroughly
- Prone to bugs when modified
- Challenging to reuse parts of its logic

Suggestion: Extract into smaller functions:
- validateRegistrationData() - lines 23-45
- checkEmailAvailability() - lines 46-68
- createUserAccount() - lines 69-98
- sendWelcomeEmail() - lines 99-130
- logRegistrationEvent() - lines 131-172

This follows the Single Responsibility Principle and makes each piece independently testable.

Example 2: Detecting Complex Conditional

Context: User working on file with deeply nested if/else

Autonomous Invocation: Skill detects and suggests:

Code Smell Detected: Complex Conditional

Lines 45-72 contain a deeply nested conditional (5 levels deep) with multiple boolean conditions.

Current pattern:
if (user) {
  if (user.isActive) {
    if (user.hasPermission('admin')) {
      if (validateToken(user.token)) {
        // ... more nesting
      }
    }
  }
}

Refactoring suggestion using guard clauses:
if (!user) return null;
if (!user.isActive) return null;
if (!user.hasPermission('admin')) return null;
if (!validateToken(user.token)) return null;

// Now do the work without nesting

Benefits: Reduces cognitive load, clearer failure conditions, easier to test edge cases.

Example 3: Detecting Code Duplication

Context: User reviewing file with repeated code blocks

Autonomous Invocation: Skill identifies DRY violation:

Code Smell Detected: Duplicate Code

Found 3 nearly identical code blocks:
- Lines 34-42 (in fetchUserById)
- Lines 78-86 (in fetchUserByEmail)
- Lines 112-120 (in fetchUserByUsername)

All three blocks implement the same error handling and response formatting logic.

Suggestion: Extract common logic into a helper function:

function handleUserFetchResponse(result, errorMessage) {
  if (!result) {
    throw new Error(errorMessage);
  }
  return formatUserResponse(result);
}

Then simplify each call site:
const user = await db.query(...);
return handleUserFetchResponse(user, 'User not found');

Impact: Reduces maintenance burden, ensures consistent behavior, fewer bugs.

Error Handling

• If file cannot be parsed: Return error with details about syntax issue
• If language not supported: Suggest manual review or generic analysis
• If file is too large (> 10k lines): Warn about God Object and suggest splitting
• If file is test file: Apply different standards (longer functions acceptable)
• If file is generated: Skip analysis or warn user

Context Awareness

Test Files

• Allow longer functions (tests often have setup/teardown)
• Allow more duplication (explicit tests > DRY)
• Different naming conventions acceptable

Configuration Files

• JSON/YAML: Check for structural issues
• Don't apply code smell rules meant for logic

Legacy Code

• Flag issues but acknowledge inherited constraints
• Prioritize critical issues over stylistic ones
• Suggest incremental improvement approach

Generated Code

• Identify if file is auto-generated
• Suggest improving generator rather than generated code
• Skip some smell checks

Integration with Development Workflow

• Non-intrusive: Provides info but doesn't block work
• Actionable: Specific suggestions with examples
• Educational: Explains why something is a smell
• Prioritized: Critical issues highlighted over minor ones
• Context-sensitive: Understands different file types

Related Skills

• suggest-performance-fix: Focuses on performance issues
• security-pattern-check: Focuses on security concerns

Notes

This skill acts as a senior developer looking over your shoulder, catching issues that might be missed during fast-paced development. It doesn't replace human code review but augments it by catching obvious issues early.