Coding

Coding Skill — Claude Best Practices

Overview

This skill guides Claude to produce production-grade code across all languages and frameworks. Use when the user asks to write, debug, refactor, review, or explain code.

---

Core Principles

1. Understand Before Writing

• Clarify requirements, language, framework, and target environment first.
• Ask about constraints: Python version, browser support, mobile/desktop, performance needs.
• Identify edge cases before writing a single line.

2. Code Quality Standards

Always produce code that is:

• Correct — handles edge cases, errors, and nulls
• Readable — clear names, small functions, comments where non-obvious
• Maintainable — DRY, modular, follows language conventions
• Tested — include unit tests or at least test snippets
• Secure — no hardcoded secrets, validate inputs, no SQL injection risks

3. Language-Specific Rules

Python

• Use type hints for all function signatures
• Prefer f-strings over .format() or %
• Use pathlib over os.path
• Use dataclasses or pydantic for data models
• Use contextlib for resource management
• Async: always await coroutines, use asyncio.gather() for concurrency
• Error handling: specific exceptions, never bare except:

# Good
async def fetch_data(url: str, timeout: int = 10) -> dict:
    async with httpx.AsyncClient() as client:
        response = await client.get(url, timeout=timeout)
        response.raise_for_status()
        return response.json()

JavaScript / TypeScript

• Prefer TypeScript over JS for any non-trivial project
• Use const/let, never var
• Prefer async/await over .then() chains
• Use optional chaining ?. and nullish coalescing ??
• Type all function params and returns in TS

// Good
async function fetchUser(id: string): Promise<User | null> {
  try {
    const res = await fetch(`/api/users/${id}`);
    if (!res.ok) return null;
    return res.json() as User;
  } catch {
    return null;
  }
}

React

• Functional components only (no class components)
• Custom hooks for reusable logic
• useMemo/useCallback only when profiling shows need
• Keys must be stable and unique (not array index)
• Co-locate state as close as possible to where it's used

SQL

• Always parameterize queries (never string concat)
• Use explicit column names, never SELECT *
• Add indexes for columns used in WHERE/JOIN
• Use CTEs for complex queries

---

Debugging Workflow

1. Reproduce — minimal reproducible example
2. Isolate — binary search the problem
3. Hypothesize — form a theory before changing code
4. Verify — confirm fix doesn't break other things
5. Document — comment why the fix was needed

---

Code Review Checklist

• Does it handle null/undefined/empty inputs?
• Are all errors caught and handled meaningfully?
• Are there any N+1 query problems?
• Is sensitive data logged anywhere?
• Are all external inputs validated/sanitized?
• Are async operations awaited?
• Are there memory leaks (event listeners, subscriptions not cleaned up)?

---

File Generation Rules

• For code > 20 lines → always create a file artifact
• Include a docstring/comment block at the top of every file
• Include if __name__ == "__main__": guards in Python scripts
• Include package.json or requirements.txt when relevant
• Always show how to run the code

---

Refactoring Patterns

Extract Function

When a block of code does one clear thing → extract it.

Replace Magic Numbers

# Bad
if age > 18:
# Good
LEGAL_AGE = 18
if age > LEGAL_AGE:

Early Return

# Bad
def process(data):
    if data:
        if validate(data):
            return transform(data)
# Good
def process(data):
    if not data: return None
    if not validate(data): return None
    return transform(data)

---

Common Patterns

Retry with Backoff

import time, random
from functools import wraps

def retry(max_attempts=3, base_delay=1.0, exceptions=(Exception,)):
    def decorator(func):
        @wraps(func)
        def wrapper(*args, **kwargs):
            for attempt in range(max_attempts):
                try:
                    return func(*args, **kwargs)
                except exceptions as e:
                    if attempt == max_attempts - 1:
                        raise
                    delay = base_delay * (2 ** attempt) + random.uniform(0, 1)
                    time.sleep(delay)
        return wrapper
    return decorator

Singleton

class Singleton:
    _instance = None
    def __new__(cls):
        if cls._instance is None:
            cls._instance = super().__new__(cls)
        return cls._instance

Observer / Event Bus

from collections import defaultdict
from typing import Callable

class EventBus:
    def __init__(self):
        self._listeners: dict[str, list[Callable]] = defaultdict(list)

    def on(self, event: str, callback: Callable):
        self._listeners[event].append(callback)

    def emit(self, event: str, *args, **kwargs):
        for cb in self._listeners[event]:
            cb(*args, **kwargs)

---

Performance Tips

• Profile before optimizing (cProfile, Chrome DevTools)
• Use generators for large sequences
• Batch database operations
• Cache expensive computations (functools.lru_cache)
• Use connection pooling for databases
• Lazy-load heavy modules

---

Security Checklist

• Never hardcode secrets → use env vars or secret managers
• Validate all user input server-side
• Use parameterized SQL queries
• Set appropriate CORS headers
• Use HTTPS everywhere
• Rate-limit API endpoints
• Hash passwords with bcrypt/argon2, never MD5/SHA1

---

Output Format

When generating code files, always:

1. Start with a brief description comment
2. Show imports at the top
3. Define constants/config before logic
4. Put main logic in functions/classes
5. End with usage example or __main__ block