PACT Coding Standards

Clean code principles for the Code phase of PACT. This skill provides essential coding guidelines and links to detailed patterns for implementation.

Core Principles

1. Single Responsibility Principle

Each function, class, or module should have exactly one reason to change.

// BAD: Multiple responsibilities
class UserManager {
  createUser(data) { }
  validateEmail(email) { }
  sendWelcomeEmail(user) { }
  generateReport() { }
  updateUserPreferences(userId, prefs) { }
}

// GOOD: Single responsibility each
class UserService {
  createUser(data) { }
  updateUser(userId, data) { }
}

class EmailValidator {
  validate(email) { }
}

class NotificationService {
  sendWelcomeEmail(user) { }
}

class UserReportGenerator {
  generate(criteria) { }
}

2. DRY (Don't Repeat Yourself)

Extract duplicated logic into reusable functions.

// BAD: Duplicated validation logic
function createUser(data) {
  if (!data.email || !data.email.includes('@')) {
    throw new Error('Invalid email');
  }
  // ...
}

function updateUser(id, data) {
  if (data.email && !data.email.includes('@')) {
    throw new Error('Invalid email');
  }
  // ...
}

// GOOD: Extracted validation
function validateEmail(email) {
  if (!email || !email.includes('@')) {
    throw new ValidationError('Invalid email format');
  }
}

function createUser(data) {
  validateEmail(data.email);
  // ...
}

function updateUser(id, data) {
  if (data.email) {
    validateEmail(data.email);
  }
  // ...
}

3. KISS (Keep It Simple, Stupid)

Choose the simplest solution that works.

// BAD: Over-engineered
class ConfigurationFactoryBuilderManager {
  static getInstance() {
    return new ConfigurationFactoryBuilder()
      .withDefaults()
      .withEnvironment()
      .build()
      .getConfiguration();
  }
}

// GOOD: Simple and direct
const config = {
  port: process.env.PORT || 3000,
  dbUrl: process.env.DATABASE_URL,
  debug: process.env.NODE_ENV !== 'production'
};

4. Defensive Programming

Validate inputs, handle edge cases, and fail gracefully.

function processOrder(order) {
  // Guard clauses
  if (!order) {
    throw new ValidationError('Order is required');
  }

  if (!order.items || order.items.length === 0) {
    throw new ValidationError('Order must have at least one item');
  }

  if (order.total < 0) {
    throw new ValidationError('Order total cannot be negative');
  }

  // Safe property access
  const customerEmail = order.customer?.email ?? 'no-email@placeholder.com';

  // Proceed with valid data
  return {
    id: generateOrderId(),
    items: order.items.map(item => ({
      ...item,
      price: Math.max(0, item.price)  // Ensure non-negative
    })),
    total: order.total,
    customerEmail
  };
}

---

Naming Conventions

Functions

// Use verb + noun pattern
function getUser(id) { }           // Retrieval
function createOrder(data) { }     // Creation
function updateProfile(id, data) { } // Mutation
function deleteComment(id) { }     // Deletion
function validateEmail(email) { }  // Validation
function calculateTotal(items) { } // Calculation
function formatDate(date) { }      // Transformation
function isActive(user) { }        // Boolean check
function hasPermission(user, action) { } // Boolean check
function canEdit(user, resource) { }     // Boolean check

Variables

// Descriptive nouns
const userCount = 42;               // Not: n, count, uc
const activeUsers = users.filter(); // Not: arr, filtered
const maxRetryAttempts = 3;         // Not: max, retries

// Boolean variables
const isActive = true;              // is/has/can/should prefix
const hasPermission = false;
const canEdit = user.role === 'admin';
const shouldRetry = attempts < maxRetryAttempts;

// Collections are plural
const users = [];                   // Not: userList, userArray
const orderItems = [];              // Not: items (too generic)

// Maps/objects describe content
const userById = {};                // Not: userMap
const priceByProductId = {};        // Describes key-value relationship

Constants

// SCREAMING_SNAKE_CASE for true constants
const MAX_RETRY_ATTEMPTS = 3;
const DEFAULT_PAGE_SIZE = 20;
const API_BASE_URL = 'https://api.example.com';

// Configuration objects
const CONFIG = Object.freeze({
  database: {
    host: process.env.DB_HOST,
    port: parseInt(process.env.DB_PORT, 10)
  }
});

Classes

// PascalCase, noun-based
class UserRepository { }     // Not: UsersRepo, UserRepo
class OrderService { }       // Not: OrdersService
class EmailValidator { }     // Not: ValidateEmail
class PaymentGateway { }     // Not: PaymentGatewayService

// Interface names (TypeScript)
interface Cacheable { }      // Adjective for capabilities
interface UserRepository { } // Noun for contracts

---

Error Handling

Fail Fast, Recover Gracefully

async function createUser(data) {
  // Validate early
  if (!data.email) {
    throw new ValidationError('Email is required');
  }

  if (!isValidEmail(data.email)) {
    throw new ValidationError('Invalid email format');
  }

  // Check business rules
  const existing = await userRepo.findByEmail(data.email);
  if (existing) {
    throw new ConflictError('Email already registered');
  }

  // Proceed with operation
  try {
    const user = await userRepo.save(data);
    await emailService.sendWelcome(user.email);
    return user;
  } catch (error) {
    // Handle specific errors
    if (error instanceof DatabaseError) {
      logger.error('Database error creating user', { error, data });
      throw new ServiceError('Unable to create user, please try again');
    }
    throw error; // Re-throw unexpected errors
  }
}

Custom Error Classes

// Base application error
class AppError extends Error {
  constructor(message, code, statusCode = 500) {
    super(message);
    this.name = this.constructor.name;
    this.code = code;
    this.statusCode = statusCode;
    Error.captureStackTrace(this, this.constructor);
  }
}

// Specific error types
class ValidationError extends AppError {
  constructor(message, details = []) {
    super(message, 'VALIDATION_ERROR', 400);
    this.details = details;
  }
}

class NotFoundError extends AppError {
  constructor(resource, id) {
    super(`${resource} with id ${id} not found`, 'NOT_FOUND', 404);
    this.resource = resource;
    this.resourceId = id;
  }
}

class ConflictError extends AppError {
  constructor(message) {
    super(message, 'CONFLICT', 409);
  }
}

class UnauthorizedError extends AppError {
  constructor(message = 'Authentication required') {
    super(message, 'UNAUTHORIZED', 401);
  }
}

For comprehensive error patterns: See references/error-handling-patterns.md

---

Logging

Log Levels

Level	When to Use	Example
ERROR	Failures requiring attention	Database connection failed
WARN	Potentially harmful situations	Rate limit approaching
INFO	Normal operational events	User created, order placed
DEBUG	Detailed debugging info	Function parameters, intermediate values

Structured Logging

const logger = require('./logger');

// BAD: Unstructured logging
console.log('User created: ' + user.id);
console.log('Error: ' + error.message);

// GOOD: Structured logging
logger.info('User created', {
  userId: user.id,
  email: user.email,
  source: 'signup'
});

logger.error('Failed to process payment', {
  orderId: order.id,
  amount: order.total,
  error: error.message,
  errorCode: error.code,
  stack: error.stack
});

// Request logging middleware
app.use((req, res, next) => {
  const requestId = req.headers['x-request-id'] || uuidv4();
  req.requestId = requestId;

  logger.info('Request received', {
    requestId,
    method: req.method,
    path: req.path,
    userAgent: req.headers['user-agent'],
    ip: req.ip
  });

  res.on('finish', () => {
    logger.info('Request completed', {
      requestId,
      statusCode: res.statusCode,
      duration: Date.now() - req.startTime
    });
  });

  next();
});

---

Code Organization

File Size Guidelines

• Maximum file size: 500 lines
• Maximum function size: 50 lines
• Maximum line length: 100 characters

Module Structure

// 1. Imports (external first, then internal)
const express = require('express');
const { validate } = require('class-validator');

const { UserService } = require('../services/UserService');
const { logger } = require('../utils/logger');

// 2. Constants and configuration
const MAX_PAGE_SIZE = 100;
const DEFAULT_PAGE_SIZE = 20;

// 3. Main exports (class, function, or router)
class UserController {
  constructor(userService) {
    this.userService = userService;
  }

  async getUsers(req, res, next) {
    // Implementation
  }
}

// 4. Helper functions (private to module)
function validatePagination(page, limit) {
  // Implementation
}

// 5. Export statement
module.exports = { UserController };

---

Interpreting design-doc optionals

When a design doc spec'd an OPTIONAL cross-link sentence, closing paragraph, or sister-section reference, the CODE author MUST evaluate the optional content's logical relationship to the preceding paragraph before deciding how to render it.

• Strengthening clause — the optional content reinforces, qualifies, or extends the directive in the preceding paragraph (e.g., a cross-link that tells the reader where the next degree of detail lives). FOLD it inline into the preceding paragraph for tighter prose flow.
• Discursive observation — the optional content introduces a new framing, raises a separate point, or pivots to an adjacent concern. KEEP it as a standalone paragraph.

The judgment lives at write time: read the preceding paragraph and ask "does this optional content add to that paragraph's directive, or does it start a new one?" Strengthening = fold; discursive = keep.

Detection signature

Design-doc phrasings like "CODE may include or drop", "optional closing cross-link", or "optional discursive paragraph" mark the entry condition. Apply the strengthening-vs-discursive judgment before composing the section; do not default to either form mechanically.

Worked example — fold-in cross-link

In pact-plugin/skills/pact-testing-strategies/SKILL.md, the section "Sibling-file convention for parametrized noise-budget regression" closes its canonical-mitigation paragraph with an inline cross-reference to the sister "Author-blindness in HANDOFF arithmetic" section rather than a standalone closing paragraph. The cross-link functions as a strengthening clause on the elevated-priority directive that precedes it — telling the reader where the pair-with-cross-stream-verifier rule lives — so it folds into the same paragraph instead of opening a new one. The fold preserves the cross-reference content while keeping the canonical mitigation a single tight directive.

---

Code Quality Checklist

Before completing CODE phase:

Structure

• Functions under 50 lines
• Files under 500 lines
• Single responsibility per function/class
• No deeply nested code (max 3 levels)

Naming

• Descriptive variable names
• Consistent naming conventions
• No abbreviations (except common ones: id, url, api)
• Boolean variables have is/has/can prefix

Error Handling

• All async operations have error handling
• Errors include relevant context
• No silent failures
• User-facing errors are friendly

Documentation

• Complex logic has comments
• Public APIs have JSDoc/docstrings
• No commented-out code
• No TODO comments without tickets

Quality

• No magic numbers (use named constants)
• No duplicate code
• Consistent code style
• Logging at appropriate levels

---

Scripts

Import Hygiene Check (--files mode)

Behavioral tests cannot see an unused import — a dead import os passes every test green. The import-hygiene check closes that blind spot mechanically. Run it over EXACTLY the .py files you modified before running the test suite:

bash <skill-directory>/scripts/lint-check.sh --files '/abs/path/to/modified_a.py' '/abs/path/to/modified_b.py'

One separately quoted absolute path per file, never one quoted list — a filename is untrusted shell input.

(When your dispatch names a plugin root, the script lives at <plugin-root>/skills/pact-coding-standards/scripts/lint-check.sh.)

A bare lint-check.sh with no arguments is refused. Do not present it as a lint command.

Lint Check Script (legacy directory mode)

Whole-tree project-type detection is reachable only with a directory argument:

# From within the skill directory:
chmod +x scripts/lint-check.sh
./scripts/lint-check.sh /path/to/project

Verdict-line contract — the LAST stdout line is always exactly one of:

Verdict	Meaning	Exit code
IMPORT-HYGIENE: PASS	No findings in the checked files	0
IMPORT-HYGIENE: FINDINGS (n)	n findings printed above the verdict	1
IMPORT-HYGIENE: SKIPPED (<reason>)	Check could not run (no arguments given, arguments given but none a checkable .py file, or no usable checker)	0

Record the verdict line verbatim in your HANDOFF produced field — every run ends in a verdict, and a SKIPPED must be visible, never silent.

What to do with findings:

• Fix every finding in files you modified: delete the unused import, or fix the undefined name.
• Never lint-fix files you did not touch — the check's scope is your diff.
• A deliberate side-effect import or re-export keeps a reasoned suppression at the import site: # noqa: F401 # re-export: <why>. Never a bare # noqa without a reason.
• If the script is missing or errors, state that in your HANDOFF instead of skipping silently.

How it checks (internal — the coder-facing contract is just the command and the verdict): an execution-probed ladder tries ruff, then pyflakes, then flake8, each restricted to the unused-import/undefined-name classes only (--select F401,F821) so you are never asked to fix style noise; when no linter is installed it falls back to the stdlib AST checker scripts/check_unused_imports.py (unused imports only, try/except-scoped imports treated as advisory). A checker crash degrades to SKIPPED — the check fails open and never blocks you on its own breakage.

Teach examples for agent-facing CLIs

When you add or change an agent-facing CLI, teach it with examples that paste-execute verbatim: build every example as python3 "<script-path>" <verb> ... — bare python3, never a pinned interpreter, and always quote the script path. In --help, the epilog lists all of the CLI's examples; a usage error appends exactly one example — the offending subcommand's own — separated from the error line by a blank line. Keep each CLI's pre-existing usage-error exit code; teaching examples never changes an exit contract.

---

Detailed References

For comprehensive coding guidance:

• Clean Code Principles: references/clean-code-principles.md

  • SOLID principles in depth
  • Code smells and refactoring
  • Function design guidelines
  • Comment best practices

• Error Handling Patterns: references/error-handling-patterns.md

  • Error handling by language
  • Global error handlers
  • Retry strategies
  • Circuit breaker patterns