home / skills / martinffx / claude-code-atelier / testing

testing skill

/plugins/atelier-spec/skills/testing

This skill helps you design and execute stub-driven tests across layers, ensuring clear boundaries and rapid feedback from core to edge.

npx playbooks add skill martinffx/claude-code-atelier --skill testing

Review the files below or copy the command above to add this skill to your agents.

Files (6)
SKILL.md
6.7 KB
---
name: testing
description: Stub-Driven TDD and layer boundary testing. Use when writing tests, deciding what to test, or testing at component boundaries.
user-invocable: false
---

# Testing Skill

Stub-Driven Test-Driven Development and layer boundary testing for functional core and effectful edge architecture.

## Core Principle: Stub-Driven TDD

Test-Driven Development workflow for the functional core / effectful edge pattern:

```
1. Stub   → Create minimal interface/function signatures
2. Test   → Write tests against stubs
3. Implement → Make tests pass with real implementation
4. Refactor  → Improve code while keeping tests green
```

**Key insight:** Write interface signatures first, test against those, then implement—not the other way around.

See [references/stub-driven-tdd.md] for complete workflow examples.

## Layer Boundary Testing

Test at the boundaries between functional core and effectful edge, not internal implementation.

```
Test here ──────▼──────────────────▼────── Test here
          Effectful Edge    │    Functional Core
              (stub)        │       (unit test)
```

### Where to Test Each Layer

| Layer | Test Type | What to Stub | What to Assert |
|-------|-----------|--------------|-----------------|
| **Entity** | Unit | Nothing (pure) | Validation, rules, transforms |
| **Service** | Unit | Repositories | Orchestration logic, error handling |
| **Router** | Integration | Service | Status codes, response format |
| **Repository** | Integration | DB connection | CRUD operations, queries |
| **Consumer** | Integration | Service | Event parsing, service calls |

See [references/boundaries.md] for detailed testing patterns by layer.

## Functional Core Testing

### Entity Tests (Pure Functions)

Focus: Validation, business rules, data transformations

```typescript
describe('Order entity', () => {
  describe('validation', () => {
    it('rejects empty items', () => {
      const order = new Order('1', 'C1', [], 'pending', 0);
      expect(order.validate().ok).toBe(false);
    });
  });

  describe('business rules', () => {
    it('prevents cancelling shipped order', () => {
      const order = new Order('1', 'C1', [], 'shipped', 0);
      expect(order.canCancel()).toBe(false);
    });
  });

  describe('transformations', () => {
    it('converts request to entity with calculated total', () => {
      const order = Order.fromRequest({
        customerId: 'C1',
        items: [
          { productId: 'P1', quantity: 2, price: 10 },
          { productId: 'P2', quantity: 1, price: 15 }
        ]
      });
      expect(order.total).toBe(35);
    });
  });
});
```

### Service Tests (Stubbed Dependencies)

Focus: Orchestration logic with stubbed repositories

```typescript
describe('OrderService.createOrder', () => {
  let service: OrderService;
  let mockRepo: OrderRepository;

  beforeEach(() => {
    mockRepo = {
      save: vi.fn().mockResolvedValue({ id: '123' }),
      findById: vi.fn()
    };
    service = new OrderService(mockRepo);
  });

  it('creates order with valid data', async () => {
    const result = await service.createOrder({
      customerId: 'C1',
      items: [{ productId: 'P1', quantity: 2 }]
    });

    expect(result.ok).toBe(true);
    expect(mockRepo.save).toHaveBeenCalled();
  });

  it('does not save when validation fails', async () => {
    const result = await service.createOrder({
      customerId: 'C1',
      items: [] // Invalid
    });

    expect(result.ok).toBe(false);
    expect(mockRepo.save).not.toHaveBeenCalled();
  });
});
```

See [references/core-testing.md] for comprehensive Entity and Service examples.

## Effectful Edge Testing

### Router, Repository, Consumer Integration Tests

Focus: Real HTTP/database/events with stubbed core

```typescript
// Router: real HTTP, stub service
describe('POST /orders', () => {
  it('returns 201 for valid request', async () => {
    const mockService = {
      createOrder: vi.fn().mockResolvedValue(Ok({ id: '123' }))
    };
    const app = createApp(mockService);

    const response = await request(app)
      .post('/orders')
      .send({ customerId: 'C1', items: [{ productId: 'P1', quantity: 2 }] });

    expect(response.status).toBe(201);
  });
});

// Repository: real test database
describe('OrderRepository.save', () => {
  it('persists order to database', async () => {
    const repo = new OrderRepository(testDb);
    const saved = await repo.save({
      id: '123',
      customer_id: 'C1',
      items: '[]',
      status: 'pending',
      total: 0
    });

    const found = await testDb.orders.findOne({ id: '123' });
    expect(found).toBeDefined();
  });
});

// Consumer: real events, stub service
describe('OrderConsumer', () => {
  it('handles OrderPlaced event', async () => {
    const mockService = {
      processOrder: vi.fn().mockResolvedValue(Ok({}))
    };
    const consumer = new OrderConsumer(mockService);

    await consumer.handle({
      type: 'OrderPlaced',
      data: { orderId: '123' }
    });

    expect(mockService.processOrder).toHaveBeenCalledWith('123');
  });
});
```

See [references/edge-testing.md] for Router, Repository, Consumer, Producer, and Client patterns.

## Test Coverage Guidelines

Aim for strategic coverage, not 100%:

**High Coverage (Critical):**
- Entity validation and business rules
- Service orchestration logic
- Critical user journeys (integration tests)
- Data transformations with logic

**Medium Coverage (Important):**
- Error handling paths
- Edge cases in business logic
- API contract validation

**Low Coverage (Optional):**
- Simple getters/setters
- Framework boilerplate
- Trivial mappings
- Internal utilities

## What NOT to Test

Avoid testing implementation details, framework behavior, and trivial code:

- Don't test private methods (test through public API)
- Don't test simple getters/setters (no logic = no test value)
- Don't test framework behavior (Express, database driver already tested)
- Don't test third-party library behavior (lodash, validation libraries)
- Don't test trivial mappings without logic

See [references/anti-patterns.md] for anti-patterns with examples and fixes.

## Testing → Implementation Flow

Follow this dependency order:

```
1. Entity tests    (pure functions, fast)
2. Service tests   (stubbed dependencies, fast)
3. Integration tests (real IO, slower)
```

This enables TDD: write tests first at lower layers, then implement, then build upward.

## Quick Reference

**For Entity Testing:** See [references/core-testing.md]
**For Service Testing:** See [references/core-testing.md]
**For Router/Repo/Consumer:** See [references/edge-testing.md]
**For Workflow Examples:** See [references/stub-driven-tdd.md]
**For What NOT to Do:** See [references/anti-patterns.md]

Overview

This skill teaches stub-driven TDD and layer-boundary testing for a functional core / effectful edge architecture. It focuses on writing minimal interfaces first, testing against stubs, and testing at boundaries between pure business logic and effectful systems. The approach helps keep unit tests fast and integration tests focused on contracts and IO.

How this skill works

Start by creating minimal stubs or interface signatures, write tests against those stubs, implement the real behavior, then refactor while keeping tests green. Test pure entities with unit tests and no stubbing, test services with stubbed repositories or dependencies, and run integration tests for routers, repositories, and consumers using real IO while stubbing the core. Coverage targets are strategic: high for critical logic, medium for error paths, and low for trivial code.

When to use it

  • When practicing TDD and you want a clear workflow to follow
  • When designing or testing a functional core with side-effecting edges
  • When deciding what to test at each architectural layer
  • When writing integration tests for HTTP, DB, or event consumers
  • When aiming for fast unit tests and small, focused integration suites

Best practices

  • Write interface signatures and stubs before implementation (stub-driven TDD)
  • Test at boundaries: unit-test pure core, integration-test effectful edge
  • Stub external dependencies in service tests and stub the core in edge tests
  • Prioritize tests for validation, orchestration, and critical user journeys
  • Avoid testing implementation details, framework behavior, or trivial getters

Example use cases

  • Entity tests that validate business rules and data transformations without IO
  • Service tests that stub repositories to verify orchestration and error handling
  • Router integration tests using real HTTP routes with a stubbed service
  • Repository tests against a test database to assert persistence and queries
  • Consumer tests that feed real events to an event handler while stubbing downstream services

FAQ

Why stub before implementing?

Stubbing interfaces first lets you define clear contracts, drive design with tests, and keep feedback fast while building up the implementation.

What layers should be unit tested vs integrated?

Unit-test pure entities and service orchestration with stubbed dependencies. Integration-test routers, repositories, and consumers with real IO but a stubbed core.