qredence/python-backend-reviewer

Python Backend Code Reviewer

Expert analysis and refactoring of Python backend code to eliminate duplication, reduce complexity, and enforce best practices.

Overview

This skill helps identify and fix common issues in Python backend code, particularly problems introduced by AI code generation:

• Duplicate code across multiple files
• Recreated utilities instead of imports
• Over-engineered solutions
• High complexity functions and classes
• Anti-patterns and code smells
• Concurrency issues in async code (shared state mutation)

The skill provides automated analysis tools and comprehensive refactoring guidance.

⚠️ Architecture-Aware Prioritization

Static analysis finds issues, but architectural context determines priority.

Before prioritizing fixes, identify:

1. Critical paths: Which code runs on every request?

   • WebSocket/HTTP handlers
   • Main workflow orchestration
   • Shared services/middleware

2. Secondary paths: Less critical code

   • CLI tools
   • Scripts
   • Dev-only utilities
   • One-time migrations

3. Concurrency model: How is state shared?

   • Are handlers concurrent?
   • Are instances shared across requests?
   • Is there mutable singleton state?

Prioritization rule: Correctness in critical paths > Complexity in secondary paths

| Finding | Critical Path | Secondary Path | | --------------------- | ---------------------------- | --------------- | | Shared state mutation | 🔴 Fix immediately | 🟡 Review | | High complexity (>25) | 🟡 Refactor carefully | 🟢 Backlog | | Duplicates | 🟡 Extract if >3 occurrences | 🟢 Nice to have | | God class | 🟡 Migrate to façade | 🟢 Low priority |

Pragmatic Thresholds

For orchestration/workflow code, use realistic thresholds:

| Metric | Strict Threshold | Pragmatic Threshold | Notes | | --------------------- | ---------------- | ------------------- | -------------------------------------------- | | Cyclomatic complexity | 10 | 25 | Orchestrators naturally have decision points | | Function length | 50 lines | 150 lines | Async flows can be longer | | Nesting depth | 4 | 5 | Guard clauses help more than extracting | | God class methods | 20 | N/A | OK if it's a façade that delegates |

Hard limits (always fix):

• No functions > 300 lines
• No nesting > 7 levels
• No shared-state mutation without synchronization guard

Quick Start

1. Run Automated Analysis

Start with automated tools to identify issues:

# Detect duplicate code blocks
uv run python scripts/detect_duplicates.py <path>

# Analyze imports and utility reimplementation
uv run python scripts/analyze_imports.py <path>

# Check code complexity
uv run python scripts/complexity_analyzer.py <path>

# Check for concurrency issues (shared state mutation)
uv run python scripts/concurrency_analyzer.py <path>

2. Review Analysis Results

Each tool outputs:

• Severity levels: Warnings (must fix) vs Info (should review)
• File locations: Exact line numbers for each issue
• Specific recommendations: What to change and why

3. Apply Fixes

Use the reference guides to refactor issues:

• See refactoring_patterns.md for step-by-step fixes
• See python_antipatterns.md for anti-pattern examples
• See best_practices.md for Python conventions

Main Workflows

Review a Python File

When a user asks to review a specific file:

1. Run all analysis tools on the file:

   python scripts/detect_duplicates.py path/to/file.py
   python scripts/analyze_imports.py path/to/file.py
   python scripts/complexity_analyzer.py path/to/file.py
   

2. Analyze results and categorize issues:

   • Critical: Duplicates, high complexity, security issues
   • Important: Utility reimplementation, deep nesting
   • Minor: Style issues, minor inefficiencies

3. Provide specific fixes:

   • Quote exact code locations with line numbers
   • Show before/after examples
   • Explain why the change improves the code

4. Offer to implement fixes if requested

Check Backend for Duplicates

When a user asks to check a project/module for duplicates:

1. Run duplicate detection on the entire directory:

   python scripts/detect_duplicates.py src/
   

2. Group duplicates by severity:

   • High: 10+ lines duplicated, 3+ occurrences
   • Medium: 5-10 lines, 2+ occurrences
   • Low: Helper functions that could be extracted

3. Recommend consolidation strategy:

   • Extract to shared utilities for cross-cutting concerns
   • Create base classes for inherited behavior
   • Use decorators for repeated patterns

Analyze Module Over-Engineering

When code appears over-engineered:

1. Run complexity analysis:

   python scripts/complexity_analyzer.py --max-complexity 10 --max-length 50 path/
   

2. Identify over-engineering patterns:

   • Premature abstractions (base classes with one implementation)
   • Excessive configuration options
   • God classes (20+ methods)
   • Deep inheritance hierarchies

3. Suggest simplifications:

   • Replace abstractions with simple functions
   • Remove unused configuration
   • Split god classes by responsibility
   • Flatten inheritance

4. Reference specific patterns from python_antipatterns.md

Optimize Following Best Practices

When asked to optimize code or ensure best practices:

1. Run all analysis tools to get baseline metrics

2. Check against best practices:

   • DRY principle violations
   • SOLID principle violations
   • Type hint coverage
   • Error handling patterns
   • Async/await consistency

3. Prioritize optimizations:

   • First: Correctness (bugs, security)
   • Second: Maintainability (duplicates, complexity)
   • Third: Performance (N+1 queries, inefficiencies)
   • Fourth: Style (naming, imports)

4. Reference best_practices.md for specific guidelines

Analyze Concurrency Safety

When reviewing async code that handles concurrent requests:

1. Run concurrency analysis:

   uv run python scripts/concurrency_analyzer.py services/ workflows/
   

2. Prioritize by severity:

   • Critical: Fix before production deployment
   • Warning: Review for actual sharing patterns
   • Info: Consider but often acceptable

3. Common fixes for shared state mutation:

   # ❌ Before: Mutating shared instance state
   class Workflow:
       async def run(self, task):
           self.current_task = task  # Race condition!
   
   # ✅ After: Request-scoped state
   class Workflow:
       async def run(self, task):
           execution = ExecutionContext(task=task)
           return await self._execute(execution)
   

4. Alternative patterns:

   • Pass state through parameters (preferred)
   • Use contextvars for request-scoped data
   • Use asyncio.Lock for truly shared state
   • Create new instances per request

Analysis Tools

detect_duplicates.py

Finds duplicate code blocks using AST analysis.

Usage:

uv run python scripts/detect_duplicates.py <path>
uv run python scripts/detect_duplicates.py --min-lines 10 <path>

Detects:

• Duplicate functions (identical implementations)
• Duplicate classes
• Repeated code blocks

Options:

• --min-lines N: Minimum lines for a block to be considered (default: 5)

analyze_imports.py

Analyzes import organization and detects recreated utilities.

Usage:

uv run python scripts/analyze_imports.py <path>

Detects:

• Wildcard imports (from module import *)
• Relative imports in non-package contexts
• Functions that look like reimplemented utilities
• Common patterns that should use libraries

Common utilities flagged:

• JSON serialization → use json or orjson
• Retry logic → use tenacity or backoff
• Validation → use pydantic
• HTTP clients → use requests or httpx

complexity_analyzer.py

Measures cyclomatic complexity, function length, and nesting depth.

Usage:

uv run python scripts/complexity_analyzer.py <path>
uv run python scripts/complexity_analyzer.py --max-complexity 10 --max-length 50 <path>

Metrics:

• Cyclomatic complexity: Number of decision points (default threshold: 10)
• Function length: Lines in function (default threshold: 50)
• Nesting depth: Maximum levels of nested control structures (threshold: 4)
• God classes: Classes with 20+ methods

Options:

• --max-complexity N: Cyclomatic complexity threshold (default: 10)
• --max-length N: Function length threshold (default: 50)

concurrency_analyzer.py

Detects concurrency issues in async Python code.

Usage:

uv run python scripts/concurrency_analyzer.py <path>

Detects:

• Shared state mutation in async methods (self.x = y in async def)
• Module-level mutable state (shared across requests)
• Missing synchronization patterns
• Potentially unsafe singleton patterns

Severity levels:

• Critical: Mutation of shared state like client, session, agent, config
• Warning: Any self.attr mutation in async context
• Info: Module-level mutable objects

When to use: Run on services, handlers, and workflow code that handles concurrent requests.

Reference Documentation

python_antipatterns.md

Comprehensive catalog of anti-patterns with examples:

• Code duplication patterns
• Over-engineering examples
• God objects
• Complexity issues
• Import problems
• Error handling mistakes
• Performance anti-patterns

Use when: You identify an issue but need to see the anti-pattern and solution

refactoring_patterns.md

Step-by-step refactoring techniques:

• Extract function/variable
• Consolidate duplicates
• Simplify conditionals
• Break up god classes
• Reduce complexity
• Improve imports

Use when: You know what's wrong and need concrete refactoring steps

best_practices.md

Python backend best practices and principles:

• Core principles (DRY, SOLID)
• Code organization
• Type hints
• Error handling
• Async patterns
• Database practices
• API design
• Security guidelines

Use when: Establishing coding standards or need authoritative guidance

Example Reviews

Example 1: Duplicate Validation Logic

User request: "Review this code for quality issues"

Analysis:

uv run python scripts/detect_duplicates.py api/

Finding: Email validation duplicated in 5 files

Recommendation:

# Extract to utils/validation.py
def validate_email(email: str) -> None:
    if not email or "@" not in email:
        raise ValueError("Invalid email")

# Import everywhere
from utils.validation import validate_email

Example 2: Recreated Retry Logic

User request: "Check if we're recreating utility functions"

Analysis:

uv run python scripts/analyze_imports.py services/

Finding: Custom retry logic in 3 services

Recommendation:

# Replace with tenacity
from tenacity import retry, stop_after_attempt, wait_exponential

@retry(stop=stop_after_attempt(3), wait=wait_exponential())
async def fetch_data(url: str):
    return await client.get(url)

Example 3: Complex Function

User request: "This function is hard to understand"

Analysis:

uv run python scripts/complexity_analyzer.py utils/processor.py

Finding: Complexity 23, nesting depth 6

Recommendation: Extract nested logic into helper functions (see refactoring_patterns.md)

When NOT to Refactor

⚠️ Avoid refactoring when:

• No tests exist and can't be added
• Close to deadline
• Code won't be modified again
• Would break public APIs without migration path

Output Format

When reviewing code, structure feedback as:

1. Summary: Brief overview of findings
2. Critical Issues: Must-fix problems (duplicates, security)
3. Important Issues: Should-fix problems (complexity, utilities)
4. Suggestions: Nice-to-have improvements
5. Code Examples: Specific before/after for each issue
6. Next Steps: Recommended action plan

Always include:

• Exact file paths and line numbers
• Severity level for each issue
• Concrete code examples
• References to patterns/practices when applicable