Code Smell Detector Skill

Purpose

This skill identifies anti-patterns that violate amplihack's development philosophy and provides constructive, specific fixes. It ensures code maintains ruthless simplicity, modular design, and zero-BS implementations.

When to Use This Skill

• Code review: Identify violations before merging
• Refactoring: Find opportunities to simplify and improve code quality
• New module creation: Catch issues early in development
• Philosophy compliance: Ensure code aligns with amplihack principles
• Learning: Understand why patterns are problematic and how to fix them
• Mentoring: Educate team members on philosophy-aligned code patterns

Core Philosophy Reference

Amplihack Development Philosophy focuses on:

• Ruthless Simplicity: Every abstraction must justify its existence
• Modular Design (Bricks & Studs): Self-contained modules with clear connection points
• Zero-BS Implementation: No stubs, no placeholders, only working code
• Single Responsibility: Each module/function has ONE clear job

Code Smells Detected

1. Over-Abstraction

What It Is: Unnecessary layers of abstraction, generic base classes, or interfaces that don't provide clear value.

Why It's Bad: Violates "ruthless simplicity" - adds complexity without proportional benefit. Makes code harder to understand and maintain.

Red Flags:

• Abstract base classes with only one implementation
• Generic helper classes that do very little
• Deep inheritance hierarchies (3+ levels)
• Interfaces for single implementations
• Over-parameterized functions

Example - SMELL:

# BAD: Over-abstracted
class DataProcessor(ABC):
    @abstractmethod
    def process(self, data):
        pass

class SimpleDataProcessor(DataProcessor):
    def process(self, data):
        return data * 2

Example - FIXED:

# GOOD: Direct implementation
def process_data(data):
    """Process data by doubling it."""
    return data * 2

Detection Checklist:

• Abstract classes with only 1-2 concrete implementations
• Generic utility classes that don't encapsulate state
• Type hierarchies deeper than 2 levels
• Mixins solving single problems

Fix Strategy:

1. Identify what the abstraction solves
2. Check if you really need multiple implementations now
3. Delete the abstraction - use direct implementation
4. If multiple implementations needed later, refactor then
5. Principle: Avoid future-proofing

---

2. Complex Inheritance

What It Is: Deep inheritance chains, multiple inheritance, or convoluted class hierarchies that obscure code flow.

Why It's Bad: Makes code hard to follow, creates tight coupling, violates simplicity principle. Who does what becomes unclear.

Red Flags:

• 3+ levels of inheritance (GrandparentClass -> ParentClass -> ChildClass)
• Multiple inheritance from non-interface classes
• Inheritance used for code reuse instead of composition
• Overriding multiple levels of methods
• "Mixin" classes for cross-cutting concerns

Example - SMELL:

# BAD: Complex inheritance
class Entity:
    def save(self): pass
    def load(self): pass

class TimestampedEntity(Entity):
    def add_timestamp(self): pass

class AuditableEntity(TimestampedEntity):
    def audit_log(self): pass

class User(AuditableEntity):
    def authenticate(self): pass

Example - FIXED:

# GOOD: Composition over inheritance
class User:
    def __init__(self, storage, timestamp_service, audit_log):
        self.storage = storage
        self.timestamps = timestamp_service
        self.audit = audit_log

    def save(self):
        self.storage.save(self)
        self.timestamps.record()
        self.audit.log("saved user")

Detection Checklist:

• Inheritance depth > 2 levels
• Multiple inheritance from concrete classes
• Methods overridden at multiple inheritance levels
• Inheritance hierarchy with no code reuse

Fix Strategy:

1. Use composition instead of inheritance
2. Pass services as constructor arguments
3. Each class handles its own responsibility
4. Easier to test, understand, and modify

---

3. Large Functions (>50 Lines)

What It Is: Functions that do too many things and are difficult to understand, test, and modify.

Why It's Bad: Violates single responsibility, makes testing harder, increases bug surface area, reduces code reusability.

Red Flags:

• Functions with >50 lines of code
• Multiple indentation levels (3+ nested if/for)
• Functions with 5+ parameters
• Functions that need scrolling to see all of them
• Complex logic that's hard to name

Example - SMELL:

# BAD: Large function doing multiple things
def process_user_data(user_dict, validate=True, save=True, notify=True, log=True):
    if validate:
        if not user_dict.get('email'):
            raise ValueError("Email required")
        if not '@' in user_dict['email']:
            raise ValueError("Invalid email")

    user = User(
        name=user_dict['name'],
        email=user_dict['email'],
        phone=user_dict['phone']
    )

    if save:
        db.save(user)

    if notify:
        email_service.send(user.email, "Welcome!")

    if log:
        logger.info(f"User {user.name} created")

    # ... 30+ more lines of mixed concerns
    return user

Example - FIXED:

# GOOD: Separated concerns
def validate_user_data(user_dict):
    """Validate user data structure."""
    if not user_dict.get('email'):
        raise ValueError("Email required")
    if '@' not in user_dict['email']:
        raise ValueError("Invalid email")

def create_user(user_dict):
    """Create user object from data."""
    return User(
        name=user_dict['name'],
        email=user_dict['email'],
        phone=user_dict['phone']
    )

def process_user_data(user_dict):
    """Orchestrate user creation workflow."""
    validate_user_data(user_dict)
    user = create_user(user_dict)
    db.save(user)
    email_service.send(user.email, "Welcome!")
    logger.info(f"User {user.name} created")
    return user

Detection Checklist:

• Function body >50 lines
• 3+ levels of nesting
• Multiple unrelated operations
• Hard to name succinctly
• 5+ parameters

Fix Strategy:

1. Extract helper functions for each concern
2. Give each function a clear, single purpose
3. Compose small functions into larger workflows
4. Each function should fit on one screen
5. Easy to name = usually doing one thing

---

4. Tight Coupling

What It Is: Modules/classes directly depend on concrete implementations instead of abstractions, making them hard to test and modify.

Why It's Bad: Changes in one module break others. Hard to test in isolation. Violates modularity principle.

Red Flags:

• Direct instantiation of classes inside functions (db = Database())
• Deep attribute access (obj.service.repository.data)
• Hardcoded class names in conditionals
• Module imports everything from another module
• Circular dependencies between modules

Example - SMELL:

# BAD: Tight coupling
class UserService:
    def create_user(self, name, email):
        db = Database()  # Hardcoded dependency
        user = db.save_user(name, email)

        email_service = EmailService()  # Hardcoded dependency
        email_service.send(email, "Welcome!")

        return user

    def get_user(self, user_id):
        db = Database()
        return db.find_user(user_id)

Example - FIXED:

# GOOD: Loose coupling via dependency injection
class UserService:
    def __init__(self, db, email_service):
        self.db = db
        self.email = email_service

    def create_user(self, name, email):
        user = self.db.save_user(name, email)
        self.email.send(email, "Welcome!")
        return user

    def get_user(self, user_id):
        return self.db.find_user(user_id)

# Usage:
user_service = UserService(db=PostgresDB(), email_service=SMTPService())

Detection Checklist:

• Class instantiation inside methods (Service())
• Deep attribute chaining (3+ dots)
• Hardcoded class references
• Circular imports or dependencies
• Module can't be tested without other modules

Fix Strategy:

1. Accept dependencies as constructor parameters
2. Use dependency injection
3. Create test doubles (mocks) easily
4. Swap implementations without changing code
5. Each module is independently testable

---

5. Missing __all__ Exports (Python)

What It Is: Python modules that don't explicitly define their public interface via __all__.

Why It's Bad: Unclear what's public vs internal. Users import private implementation details. Violates the "stud" concept - unclear connection points.

Red Flags:

• No __all__ in __init__.py
• Modules expose internal functions/classes
• Users uncertain what to import
• Private names (_function) still accessible
• Documentation doesn't match exports

Example - SMELL:

# BAD: No __all__ - unclear public interface
# module/__init__.py
from .core import process_data, _internal_helper
from .utils import validate_input, LOG_LEVEL

# What should users import? All of it? Only some?

Example - FIXED:

# GOOD: Clear public interface via __all__
# module/__init__.py
from .core import process_data
from .utils import validate_input

__all__ = ['process_data', 'validate_input']

# Users know exactly what's public and what to use

Detection Checklist:

• Missing __all__ in __init__.py
• Internal functions (prefixed with _) exposed
• Unclear what's "public API"
• All imports at module level

Fix Strategy:

1. Add __all__ to every __init__.py
2. List ONLY the public functions/classes
3. Prefix internal implementation with _
4. Update documentation to match __all__
5. Clear = users know exactly what to use

---

Analysis Process

Step 1: Scan Code Structure

1. Review file organization and module boundaries
2. Identify inheritance hierarchies
3. Scan for large functions (count lines)
4. Note __all__ presence/absence
5. Check for tight coupling patterns

Step 2: Analyze Each Smell

For each potential issue:

1. Confirm it violates philosophy
2. Measure severity (critical/major/minor)
3. Find specific line numbers
4. Note impact on system

Step 3: Generate Fixes

For each smell found:

1. Provide clear explanation of WHY it's bad
2. Show BEFORE code
3. Show AFTER code with detailed comments
4. Explain philosophy principle violated
5. Give concrete refactoring steps

Step 4: Create Report

1. List all smells found
2. Prioritize by severity/impact
3. Include specific examples
4. Provide actionable fixes
5. Reference philosophy docs

---

Detection Rules

Rule 1: Abstract Base Classes

Check: class X(ABC) with exactly 1 concrete implementation

# BAD pattern detection
- Count implementations of abstract class
- If count <= 2 and not used as interface: FLAG

Fix: Remove abstraction, use direct implementation

Rule 2: Inheritance Depth

Check: Class hierarchy depth

# BAD pattern detection
- Follow inheritance chain: class -> parent -> grandparent...
- If depth > 2: FLAG

Fix: Use composition instead

Rule 3: Function Line Count

Check: All function bodies

# BAD pattern detection
- Count lines in function (excluding docstring)
- If > 50 lines: FLAG
- If > 3 nesting levels: FLAG

Fix: Extract helper functions

Rule 4: Dependency Instantiation

Check: Class instantiation inside methods/functions

# BAD pattern detection
- Search for "= ServiceName()" inside methods
- If found: FLAG

Fix: Pass as constructor argument

Rule 5: Missing all

Check: Python modules

# BAD pattern detection
- Look for __all__ definition
- If missing: FLAG
- If __all__ incomplete: FLAG

Fix: Define explicit __all__

---

Common Code Smells & Quick Fixes

Smell: "Utility Class" Holder

# BAD
class StringUtils:
    @staticmethod
    def clean(s):
        return s.strip().lower()

Fix: Use direct function

# GOOD
def clean_string(s):
    return s.strip().lower()

---

Smell: "Manager" Class

# BAD
class UserManager:
    def create(self): pass
    def update(self): pass
    def delete(self): pass
    def validate(self): pass
    def email(self): pass

Fix: Split into focused services

# GOOD
class UserService:
    def __init__(self, db, email):
        self.db = db
        self.email = email

    def create(self): pass
    def update(self): pass
    def delete(self): pass

def validate_user(user): pass

---

Smell: God Function

# BAD - 200 line function doing everything
def process_order(order_data, validate, save, notify, etc...):
    # 200 lines mixing validation, transformation, DB, email, logging

Fix: Compose small functions

# GOOD
def process_order(order_data):
    validate_order(order_data)
    order = create_order(order_data)
    save_order(order)
    notify_customer(order)
    log_creation(order)

---

Smell: Brittle Inheritance

# BAD
class Base:
    def work(self): pass
class Middle(Base):
    def work(self):
        return super().work()
class Derived(Middle):
    def work(self):
        return super().work()  # Which work()?

Fix: Use clear, testable composition

# GOOD
class Worker:
    def __init__(self, validator, transformer):
        self.validator = validator
        self.transformer = transformer

    def work(self, data):
        self.validator.check(data)
        return self.transformer.apply(data)

---

Smell: Hidden Dependencies

# BAD
def fetch_data(user_id):
    db = Database()  # Where's this coming from?
    return db.query(f"SELECT * FROM users WHERE id={user_id}")

Fix: Inject dependencies explicitly

# GOOD
def fetch_data(user_id, db):
    return db.query(f"SELECT * FROM users WHERE id={user_id}")

# Or in a class:
class UserRepository:
    def __init__(self, db):
        self.db = db

    def fetch(self, user_id):
        return self.db.query(f"SELECT * FROM users WHERE id={user_id}")

---

Usage Examples

Example 1: Review New Module

User: Review this new authentication module for code smells.

Claude:
1. Scans all Python files in module
2. Checks for each smell type
3. Finds:
   - Abstract base class with 1 implementation
   - Large 120-line authenticate() function
   - Missing __all__ in __init__.py
4. Provides specific fixes with before/after code
5. Explains philosophy violations

Example 2: Identify Tight Coupling

User: Find tight coupling in this user service.

Claude:
1. Traces all dependencies
2. Finds hardcoded Database() instantiation
3. Finds direct EmailService() creation
4. Shows dependency injection fix
5. Includes test example showing why it matters

Example 3: Simplify Inheritance

User: This class hierarchy is too complex.

Claude:
1. Maps inheritance tree (finds 4 levels)
2. Shows each level doing what
3. Suggests composition approach
4. Provides before/after refactoring
5. Explains how it aligns with brick philosophy

---

Analysis Checklist

Philosophy Compliance

• No unnecessary abstractions
• Single responsibility per class/function
• Clear public interface (__all__)
• Dependencies injected, not hidden
• Inheritance depth <= 2 levels
• Functions < 50 lines
• No dead code or stubs

Code Quality

• Each function has one clear job
• Easy to understand at a glance
• Easy to test in isolation
• Easy to modify without breaking others
• Clear naming reflects responsibility

Modularity

• Modules are independently testable
• Clear connection points ("studs")
• Loose coupling between modules
• Explicit dependencies

---

Success Criteria for Review

A code review using this skill should:

• Identify all violations of philosophy
• Provide specific line numbers
• Show before/after examples
• Explain WHY each is a problem
• Suggest concrete fixes
• Include test strategies
• Reference philosophy docs
• Prioritize by severity
• Be constructive and educational
• Help writer improve future code

---

Integration with Code Quality Tools

When to Use This Skill:

• During code review (before merge)
• In pull request comments
• Before creating new modules
• When refactoring legacy code
• To educate team members
• In design review meetings

Works Well With:

• Code review process
• Module spec generation
• Refactoring workflows
• Architecture discussions
• Mentoring and learning

---

Resources

• Philosophy: .claude/context/PHILOSOPHY.md
• Patterns: .claude/context/PATTERNS.md
• Brick Philosophy: See "Modular Architecture for AI" in PHILOSOPHY.md
• Zero-BS: See "Zero-BS Implementations" in PHILOSOPHY.md

---

Remember

This skill helps maintain code quality by:

1. Catching issues before they become technical debt
2. Educating developers on philosophy
3. Keeping code simple and maintainable
4. Preventing tightly-coupled systems
5. Making code easier to understand and modify

Use it constructively - the goal is learning and improvement, not criticism.