Prod Python

Production Python

Write Python like a senior developer — simple, readable, PEP-compliant. No AI slop. No over-engineering. No unnecessary abstractions.

Core principle: The best code is the simplest code that solves the problem correctly.

When to Use

• Writing any Python code (scripts, APIs, libraries, CLIs)
• Refactoring existing Python code
• Code reviews — flag AI patterns
• Code feels bloated, over-abstracted, or "AI-generated"

AI Slop — Eliminate on Sight

Module-Level Docstrings on Every File

# SLOP
"""User authentication module for handling login and registration."""

from flask import request

# CLEAN — filename says it all
from flask import request

Only add module docstrings for genuinely complex algorithms or non-obvious design decisions. Never add # src/path/file.py path comments.

Over-Documented Obvious Code

# SLOP — 20 lines to say "sum prices"
def calculate_total(items: List[Dict[str, Any]]) -> float:
    """
    Calculate the total price of items.

    Args:
        items: List of item dictionaries containing prices

    Returns:
        float: The total sum of all item prices

    Raises:
        ValueError: If items is empty
        KeyError: If price key missing
    """
    total = 0.0
    for item in items:
        total += item['price']
    return total

# CLEAN
def calculate_total(items: list[dict]) -> float:
    return sum(item['price'] for item in items)

Docstrings: one line max for obvious functions. Skip entirely if the signature tells the story.

Narrating Code with Comments

# SLOP
# Initialize the user list
users = []
# Loop through each record
for record in records:
    # Create user object
    user = User(record)
    # Append to list
    users.append(user)

# CLEAN
users = [User(r) for r in records]

Comments explain WHY, never WHAT. If you need to explain what code does, rewrite the code.

Legacy typing Imports

# SLOP — pre-3.9 style
from typing import List, Dict, Optional, Union, Tuple, Any

def process(data: List[Dict[str, Any]]) -> Optional[Dict[str, Union[str, int]]]:
    ...

# CLEAN — modern builtins + PEP 604
def process(data: list[dict]) -> dict | None:
    ...

Never import from typing for: List, Dict, Tuple, Set, FrozenSet, Type, Optional, Union. Use builtins and | syntax.

Premature Abstraction

# SLOP
class DataProcessor:
    def __init__(self, config):
        self.config = config

    def process(self, data):
        return self._transform(self._validate(data))

    def _validate(self, data):
        return data

    def _transform(self, data):
        return [x * 2 for x in data]

# CLEAN
def process_data(data: list[int]) -> list[int]:
    return [x * 2 for x in data]

Classes are for state. If your class has one method or no meaningful state, it's a function.

Over-Engineered Return Types

# SLOP
class ValidationResult:
    def __init__(self, is_valid: bool, error: str | None = None):
        self.is_valid = is_valid
        self.error = error

def validate(user: dict) -> ValidationResult:
    if not user.get('email'):
        return ValidationResult(False, "Email required")
    return ValidationResult(True)

# CLEAN
def validate(user: dict) -> str | None:
    """Return error message or None if valid."""
    if not user.get('email'):
        return "Email required"
    return None

Wrapper Functions That Add Nothing

# SLOP
def get_user_by_email(email: str) -> dict | None:
    """Retrieve a user record from the database by their email address."""
    query = "SELECT * FROM users WHERE email = :email"
    result = db.execute(query, {"email": email}).fetchone()
    return dict(result) if result else None

# CLEAN
def get_user(email: str) -> dict | None:
    return db.execute(
        "SELECT * FROM users WHERE email = :email", {"email": email}
    ).fetchone()

Noise Logging

# SLOP
logger.info("Starting user processing")
logger.info(f"Found {len(users)} users")
for user in users:
    logger.debug(f"Processing user {user.id}")
    process(user)
logger.info("Finished processing users")

# CLEAN
logger.info("Processing %d users", len(users))
for user in users:
    process(user)
if failures:
    logger.error("Failed %d users: %s", len(failures), failures[:5])

Log events that matter: errors, warnings, key metrics. Not every step.

Defensive Programming Against Impossible States

# SLOP — checking for things that can't happen
def process_order(order: Order) -> None:
    if order is None:
        raise ValueError("Order cannot be None")
    if not isinstance(order, Order):
        raise TypeError("Expected Order instance")
    if not hasattr(order, 'items'):
        raise AttributeError("Order must have items")
    # actual logic starts here...

# CLEAN — trust your own code, validate at boundaries
def process_order(order: Order) -> None:
    for item in order.items:
        item.fulfill()

Validate at system boundaries (user input, external APIs). Trust internal code.

Modern Python (3.13+)

Type Syntax

# Builtins as generics (3.9+)
names: list[str] = []
config: dict[str, int] = {}
pair: tuple[str, int] = ("a", 1)
ids: set[int] = set()

# Union with | (3.10+)
def find(id: int) -> User | None: ...
def parse(value: str | bytes) -> dict: ...

# type statement for aliases (3.12+)
type JSON = dict[str, 'JSON'] | list['JSON'] | str | int | float | bool | None
type Handler = Callable[[Request], Response]

# TypeVar with defaults (3.13+)
from typing import TypeVar
type T = TypeVar('T', default=int)

# Generic classes — modern syntax (3.12+)
class Stack[T]:
    def __init__(self) -> None:
        self._items: list[T] = []

    def push(self, item: T) -> None:
        self._items.append(item)

    def pop(self) -> T:
        return self._items.pop()

When to Type-Hint

# SKIP type hints — signature is obvious
def double(x):
    return x * 2

def greet(name):
    return f"Hello, {name}"

# ADD type hints — return type isn't obvious
def fetch_user(user_id: int) -> User | None: ...
def parse_config(path: str) -> dict[str, str]: ...
def connect(url: str, *, timeout: float = 30) -> Connection: ...

Type-hint when the types clarify. Skip when they're noise.

Structural Pattern Matching (3.10+)

# Use match for complex dispatch — not if/elif chains
match command:
    case {"action": "create", "name": str(name)}:
        create_resource(name)
    case {"action": "delete", "id": int(id_)}:
        delete_resource(id_)
    case {"action": str(action)}:
        raise ValueError(f"Unknown action: {action}")

# For HTTP status handling
match response.status_code:
    case 200 | 201:
        return response.json()
    case 404:
        return None
    case 429:
        raise RateLimitError(response.headers.get("Retry-After"))
    case status if status >= 500:
        raise ServerError(f"Server error: {status}")

Exception Groups (3.11+)

# Collect multiple errors, raise together
async def validate_all(items: list[dict]) -> list[dict]:
    errors = []
    valid = []
    for i, item in enumerate(items):
        try:
            valid.append(validate(item))
        except ValueError as e:
            errors.append(e)

    if errors:
        raise ExceptionGroup("Validation failed", errors)
    return valid

F-Strings — Always

# Use f-strings, never % or .format()
name = f"{first} {last}"
path = f"/api/v{version}/users/{user_id}"
msg = f"Processed {count:,} items in {elapsed:.2f}s"

Walrus Operator

# Assign and test in one expression
if (match := pattern.search(text)) is not None:
    process(match.group(1))

if (user := get_user(email)) is not None:
    send_welcome(user)

# In while loops
while (chunk := f.read(8192)):
    process(chunk)

# In comprehensions
results = [
    cleaned
    for raw in data
    if (cleaned := clean(raw)) is not None
]

Production Patterns

Error Handling

# Specific exceptions with context
def load_config(path: str) -> dict:
    try:
        with open(path) as f:
            return json.load(f)
    except FileNotFoundError:
        raise ValueError(f"Config not found: {path}") from None
    except json.JSONDecodeError as e:
        raise ValueError(f"Invalid JSON in {path}: {e}") from e

# Custom exceptions — only when callers need to catch them
class PaymentError(Exception):
    def __init__(self, amount: float, reason: str):
        self.amount = amount
        self.reason = reason
        super().__init__(f"Payment of {amount} failed: {reason}")

Never catch bare Exception unless re-raising. Never silently swallow errors.

Data Classes over Dicts

# When you need structured data with known fields
from dataclasses import dataclass

@dataclass(frozen=True, slots=True)
class User:
    name: str
    email: str
    active: bool = True

# For config / settings
@dataclass(frozen=True, slots=True)
class DBConfig:
    host: str
    port: int = 5432
    pool_size: int = 10

Use slots=True for memory efficiency. Use frozen=True when immutability makes sense.

Async Done Right

import asyncio
import httpx

async def fetch_users(urls: list[str]) -> list[dict]:
    async with httpx.AsyncClient() as client:
        tasks = [client.get(url) for url in urls]
        responses = await asyncio.gather(*tasks, return_exceptions=True)

    results = []
    for resp in responses:
        if isinstance(resp, Exception):
            logger.error("Fetch failed: %s", resp)
            continue
        results.append(resp.json())
    return results

Context Managers for Resources

from contextlib import contextmanager

@contextmanager
def db_transaction(conn):
    try:
        yield conn
        conn.commit()
    except Exception:
        conn.rollback()
        raise

# Usage
with db_transaction(conn) as tx:
    tx.execute("INSERT INTO users ...")

Comprehensions — Know When to Stop

# YES — simple transforms and filters
active = [u for u in users if u.active]
by_id = {u.id: u for u in users}
emails = {u.email for u in users}

# NO — too complex, use a loop
# SLOP
results = [
    transform(item)
    for group in data
    for item in group.items
    if item.valid and item.category in allowed
    and not item.archived
]

# CLEAN
results = []
for group in data:
    for item in group.items:
        if item.valid and item.category in allowed and not item.archived:
            results.append(transform(item))

If a comprehension needs more than one if or is hard to read in 2 seconds, use a loop.

Pathlib over os.path

from pathlib import Path

config_dir = Path.home() / ".config" / "myapp"
config_dir.mkdir(parents=True, exist_ok=True)

for py_file in Path("src").rglob("*.py"):
    process(py_file)

content = Path("data.json").read_text()
Path("output.txt").write_text(result)

Enums for Fixed Choices

from enum import StrEnum

class Status(StrEnum):
    PENDING = "pending"
    ACTIVE = "active"
    SUSPENDED = "suspended"

# Use StrEnum so values serialize naturally
user.status = Status.ACTIVE
assert user.status == "active"  # works

Import Organization (PEP 8)

# 1. Standard library
import json
import logging
from pathlib import Path

# 2. Third-party
import httpx
from fastapi import FastAPI

# 3. Local
from .models import User
from .config import settings

No blank lines within groups. One blank line between groups. Alphabetical within each group. Absolute imports preferred. Use isort for automation.

Project Layout

project/
    src/
        project_name/
            __init__.py
            main.py
            models.py
            config.py
    tests/
        test_main.py
        test_models.py
        conftest.py
    pyproject.toml

Use pyproject.toml for all config. No setup.py, no setup.cfg, no requirements.txt for libraries (use pyproject.toml dependencies). requirements.txt is acceptable for applications that pin exact versions.

Quick Reference

| Situation | Do | Don't | |---|---|---| | Types | list[str], dict | None | List[str], Optional[Dict] | | Module docstrings | Skip | """Module for X.""" everywhere | | Function docstrings | One line or skip | Multi-paragraph for simple code | | Comments | Explain WHY | Narrate WHAT | | Functions | One clear purpose | _helper, _util suffixes | | Classes | When state is needed | Single-method wrappers | | Errors | Specific + context | Bare except Exception | | Logging | Errors + key events | Every step | | Strings | f-strings | .format() or % | | Paths | pathlib.Path | os.path.join | | Data | @dataclass(slots=True) | Plain dicts for structured data | | Config | pyproject.toml | setup.py + setup.cfg | | Constants | StrEnum / IntEnum | Magic strings/numbers | | Validation | At boundaries only | Defensive checks everywhere |

Red Flags — Simplify Immediately

• from typing import List, Dict, Optional, Union — use builtins + |
• from __future__ import annotations — unnecessary on 3.13+
• Module docstring on every file
• # src/path/file.py comment at top
• Function with _helper, _util, _internal suffix
• Class with one method → make it a function
• Docstring longer than the function body
• More comments than code
• 3+ levels of indentation → refactor
• isinstance checks in internal code
• if x is not None: return x / else: return None → just return x
• Wrapper functions that add nothing
• **kwargs when you know the exact parameters
• Empty __init__.py with docstrings
• try: ... except Exception: pass

Real-World Impact

Before (AI slop — 45 lines):

from typing import List, Dict, Any, Optional

class UserValidator:
    """Validator for user data."""

    def __init__(self) -> None:
        self.required_fields: List[str] = ['email', 'name']

    def validate(self, user: Dict[str, Any]) -> Dict[str, Any]:
        """
        Validate user data.

        Args:
            user: User dictionary to validate

        Returns:
            Validation result with is_valid and errors
        """
        errors: List[str] = []
        for field in self.required_fields:
            if not user.get(field):
                errors.append(f"{field} is required")

        return {
            'is_valid': len(errors) == 0,
            'errors': errors
        }

validator = UserValidator()
result = validator.validate(user_data)
if not result['is_valid']:
    handle_errors(result['errors'])

After (clean — 8 lines):

def validate_user(user: dict) -> list[str]:
    errors = []
    if not user.get('email'):
        errors.append("Email required")
    if not user.get('name'):
        errors.append("Name required")
    return errors

if errors := validate_user(user_data):
    handle_errors(errors)

Result: 82% less code. More readable. Same functionality. No imports needed.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: ankit-aglawe
• Source: ankit-aglawe/python-coding-agent-skill
• License: MIT

Install and usage instructions live in the source repository linked above.