---

name: tdd description: Enforces test-driven development (TDD) following Kent Beck's methodology. MANDATORY micro-cycle approach - write ONE test, watch it fail, write minimal implementation, refactor, then NEXT test. NEVER write all tests first or implementation first. Use for ANY code writing task. allowed-tools: Read, Glob, Grep, Bash(bun test:), Bash(node --import tsx --test:), Write, Edit

Test-Driven Development (TDD) Enforcement

CRITICAL RULE: ONE Test at a Time

NEVER write implementation code before writing a test. NEVER write multiple tests before writing implementation.

The cycle is: ONE test → implementation for that test → refactor → REPEAT

When the user asks you to implement ANY code (features, bug fixes, new functions, deciders, projections, handlers), you MUST follow this EXACT sequence:

1. Write ONE test (just one!)
2. Run test → verify it FAILS (Red phase)
3. Write MINIMAL implementation to make THAT test pass (Green phase)
4. Run test → verify it PASSES
5. Refactor if needed while keeping tests passing (Refactor phase)
6. Run test → verify still passing
7. GOTO step 1 for the NEXT test

You are FORBIDDEN from:

• ❌ Writing all tests first, then all implementation
• ❌ Writing complete implementation, then adding tests
• ❌ Writing multiple tests before any implementation
• ❌ Skipping the "run test and watch it fail" step

When This Skill Applies

This skill is MANDATORY for:

• ✅ Implementing new features
• ✅ Creating new functions or modules
• ✅ Writing deciders or projections (event sourcing)
• ✅ Creating HTTP handlers
• ✅ Fixing bugs (write test that reproduces bug first)
• ✅ Refactoring (tests ensure behavior doesn't change)
• ✅ Adding business logic

This skill does NOT apply to:

• ❌ Configuration files (package.json, tsconfig.json)
• ❌ View templates (.ejs files)
• ❌ Documentation files (.md)
• ❌ Build scripts

Project-Specific Test Patterns

Test File Naming Convention

Implementation:  src/usecases/budget/allocate.ts
Test:           src/usecases/budget/allocate.spec.ts

Implementation:  src/products/add.ts
Test:           src/products/add.spec.ts

Pattern: Same directory, same name, .spec.ts extension

Running Tests

# Run all tests in a package
bun test

# Run specific test file
node --import tsx --test 'src/usecases/budget/allocate.spec.ts'

# Run tests matching pattern
node --import tsx --test 'src/**/*.spec.ts'

Event Sourcing Decider Tests

For deciders, test both decide() and evolve() functions:

import { describe, it } from "node:test"
import assert from "node:assert"
import { AllocateBudget, type AllocateBudgetCommand } from "./allocate"

describe("AllocateBudget", () => {
  describe("decide", () => {
    it("should create budget.allocated event when budget does not exist", async () => {
      const decider = AllocateBudget()
      const command: AllocateBudgetCommand = {
        type: "budget.allocate",
        id: "budget_123",
        description: "Groceries",
        amount: 500,
        frequency: "monthly",
      }

      const events = await decider.decide(command, decider.initialState())

      assert.strictEqual(events.length, 1)
      assert.strictEqual(events[0].event.type, "budget.allocated")
      assert.strictEqual(events[0].event.id, "budget_123")
      assert.strictEqual(events[0].event.description, "Groceries")
      assert.strictEqual(events[0].event.amount, 500)
      assert.strictEqual(events[0].event.frequency, "monthly")
    })

    it("should return empty array when budget already exists (idempotency)", async () => {
      const decider = AllocateBudget()
      const command: AllocateBudgetCommand = {
        type: "budget.allocate",
        id: "budget_123",
        description: "Groceries",
        amount: 500,
        frequency: "monthly",
      }

      // First allocation
      const events = await decider.decide(command, decider.initialState())

      // Evolve state
      const state = decider.evolve(decider.initialState(), {
        position: 1n,
        timestamp: new Date(),
        streamId: ["budget_123"],
        type: "budget.allocated",
        event: events[0].event,
      })

      // Second allocation (should be idempotent)
      const secondEvents = await decider.decide(command, state)

      assert.strictEqual(secondEvents.length, 0)
    })
  })

  describe("evolve", () => {
    it("should mark budget as allocated in state", () => {
      const decider = AllocateBudget()
      const initialState = decider.initialState()

      const state = decider.evolve(initialState, {
        position: 1n,
        timestamp: new Date(),
        streamId: ["budget_123"],
        type: "budget.allocated",
        event: {
          type: "budget.allocated",
          id: "budget_123",
          description: "Groceries",
          amount: 500,
          frequency: "monthly",
        },
      })

      assert.strictEqual(state["budget_123"], true)
    })
  })
})

Projection Tests

For projections, test the catchup() method and query methods:

import { describe, it } from "node:test"
import assert from "node:assert"
import { eventstore } from "@groceries/event-sourcing"
import { InMemoryProjection } from "./in-memory-projection"
import type { BudgetEvent } from "./event"

describe("BudgetProjection", () => {
  it("should return empty array when no budgets exist", async () => {
    const store = new eventstore.InMemory<BudgetEvent>()
    const projection = InMemoryProjection(store)

    await projection.catchup()
    const budgets = await projection.all()

    assert.strictEqual(budgets.length, 0)
  })

  it("should return allocated budget after catchup", async () => {
    const store = new eventstore.InMemory<BudgetEvent>()
    await store.save([{
      streamId: ["budget_123"],
      type: "budget.allocated",
      event: {
        type: "budget.allocated",
        id: "budget_123",
        description: "Groceries",
        amount: 500,
        frequency: "monthly",
      },
    }])

    const projection = InMemoryProjection(store)
    await projection.catchup()
    const budgets = await projection.all()

    assert.strictEqual(budgets.length, 1)
    assert.strictEqual(budgets[0].id, "budget_123")
    assert.strictEqual(budgets[0].description, "Groceries")
    assert.strictEqual(budgets[0].amount, 500)
    assert.strictEqual(budgets[0].frequency, "monthly")
  })

  it("should return budget by id", async () => {
    const store = new eventstore.InMemory<BudgetEvent>()
    await store.save([{
      streamId: ["budget_123"],
      type: "budget.allocated",
      event: {
        type: "budget.allocated",
        id: "budget_123",
        description: "Groceries",
        amount: 500,
        frequency: "monthly",
      },
    }])

    const projection = InMemoryProjection(store)
    await projection.catchup()
    const budget = await projection.byId("budget_123")

    assert.strictEqual(budget.id, "budget_123")
    assert.strictEqual(budget.description, "Groceries")
  })

  it("should throw error when budget not found", async () => {
    const store = new eventstore.InMemory<BudgetEvent>()
    const projection = InMemoryProjection(store)
    await projection.catchup()

    await assert.rejects(
      async () => await projection.byId("budget_nonexistent"),
      { name: "BudgetNotFoundError" }
    )
  })
})

TDD Workflow (Kent Beck Style)

CRITICAL: Work on ONE test at a time. NEVER write multiple tests before implementation.

The Micro-Cycle (Repeat This For Each Test)

Step 1: Write ONE Test (Just One!)

Write a SINGLE test case. Not multiple tests. Just one.

// CORRECT: Write ONE test
it("should create budget.allocated event when budget does not exist", async () => {
  // test code here
})

// WRONG: Don't write multiple tests at once
it("should create event...", async () => { })
it("should be idempotent...", async () => { })  // ❌ STOP! One at a time!
it("should reject negative...", async () => { })  // ❌ NO!

Step 2: Run the Test (Watch It Fail - RED)

Run ONLY this test. It MUST fail. This proves the test is actually testing something.

node --import tsx --test 'src/usecases/budget/allocate.spec.ts'

Expected output: ❌ Test fails (RED)

If it passes without implementation, the test is wrong or testing the wrong thing.

Step 3: Write Minimal Implementation (Make It Pass - GREEN)

Write the SMALLEST amount of code to make THIS ONE test pass. Don't think about future tests.

node --import tsx --test 'src/usecases/budget/allocate.spec.ts'

Expected output: ✅ Test passes (GREEN)

Step 4: Refactor (Keep It GREEN)

Only now, improve the code:

• Remove duplication
• Improve names
• Simplify logic

After EACH change, run tests to ensure they still pass.

Step 5: Repeat the Cycle

Now go back to Step 1 and write THE NEXT test. One test at a time.

Planning Tests (Mental Note Only)

You may mentally list what needs testing:

Mental checklist (don't write these tests yet):
- create event when doesn't exist ← Start here (write this test first)
- idempotent behavior ← Write this test AFTER first passes
- reject negative amounts ← Write this test AFTER second passes
- reject invalid frequency ← Write this test AFTER third passes

But write code for only ONE at a time.

Anti-Patterns (What NOT to Do)

❌ WRONG: Writing Implementation First

// DON'T DO THIS - No test written first!
export function AllocateBudget(): Decider<...> {
  return {
    decide: async (command, state) => {
      // Implementation without tests
    }
  }
}

✅ RIGHT: Writing Test First

// Step 1: Write the test FIRST
describe("AllocateBudget", () => {
  it("should create budget.allocated event", async () => {
    const decider = AllocateBudget()
    const events = await decider.decide(command, state)
    assert.strictEqual(events.length, 1)
  })
})

// Step 2: Run test (it fails - good!)
// Step 3: Write minimal implementation to pass

❌ WRONG: Skipping Test Verification

Don't assume tests pass. Always run them:

# ALWAYS run tests to verify
bun test

❌ WRONG: Writing All Tests First, Then Implementation

❌ write allocate.spec.ts (ALL tests at once)
   - test 1 ❌
   - test 2 ❌
   - test 3 ❌
   - test 4 ❌
❌ write allocate.ts (complete implementation)
   - all tests pass ✅

Why this is wrong: You lose the benefit of incremental design. Tests should guide your implementation one step at a time.

❌ WRONG: Writing All Code, Then All Tests

❌ write allocate.ts (complete implementation)
❌ write projection.ts (complete implementation)
❌ write allocate.spec.ts (all tests)
❌ write projection.spec.ts (all tests)

Why this is wrong: Tests written after implementation are just checking existing behavior, not driving design.

✅ RIGHT: Micro-Cycle - ONE Test, Minimal Code, Repeat

Cycle 1:
✅ write allocate.spec.ts (ONLY test 1: "should create event")
✅ run test → ❌ FAILS (good!)
✅ write allocate.ts (minimal code to pass test 1)
✅ run test → ✅ PASSES
✅ refactor if needed
✅ run test → ✅ still passes

Cycle 2:
✅ write allocate.spec.ts (add ONLY test 2: "should be idempotent")
✅ run test → ❌ FAILS (good!)
✅ update allocate.ts (minimal code to pass test 2)
✅ run test → ✅ PASSES
✅ refactor if needed
✅ run test → ✅ still passes

Cycle 3:
✅ write allocate.spec.ts (add ONLY test 3: "should reject negative amounts")
✅ run test → ❌ FAILS (good!)
✅ update allocate.ts (minimal code to pass test 3)
✅ run test → ✅ PASSES

... repeat for each remaining test

Why this is right: Each test drives exactly the code needed. Implementation emerges from requirements.

Checklist Before Completing Any Task

Before marking a coding task as complete, verify you followed the micro-cycle for EACH test:

For EACH test case, did you:

• Write exactly ONE test (not multiple)
• Run the test and verify it FAILED (Red)
• Write MINIMAL implementation to pass THAT test
• Run the test and verify it PASSED (Green)
• Refactor (if needed)
• Verify tests still pass after refactoring
• Then move to NEXT test (repeat above)

Overall checklist:

• Test file created with .spec.ts extension
• Followed Red-Green-Refactor for every single test
• No existing tests broken
• Edge cases covered (empty arrays, null values, errors)
• Idempotency tested (for deciders)
• All tests run and pass

Red flags (indicates you did NOT follow TDD):

• 🚩 You wrote more than one test before implementation
• 🚩 You wrote implementation before any tests
• 🚩 You didn't run tests to see them fail
• 🚩 You wrote "complete" implementation in one go

Test Assertion Library

Use Node.js built-in assert module:

import assert from "node:assert"

// Equality checks
assert.strictEqual(actual, expected)
assert.deepStrictEqual(actualObject, expectedObject)
assert.notStrictEqual(actual, notExpected)

// Boolean checks
assert.ok(value)
assert.strictEqual(value, true)

// Array/object checks
assert.strictEqual(array.length, 3)
assert.deepStrictEqual(obj, { id: "123", name: "Alice" })

// Error checks
await assert.rejects(
  async () => await functionThatThrows(),
  { name: "ExpectedErrorName" }
)

assert.throws(
  () => functionThatThrows(),
  /expected error message/
)

Test Organization Structure

import { describe, it } from "node:test"
import assert from "node:assert"

describe("Feature/Module Name", () => {
  describe("specific function or method", () => {
    it("should [expected behavior] when [condition]", () => {
      // Arrange: Set up test data
      const input = { ... }

      // Act: Call the function
      const result = functionUnderTest(input)

      // Assert: Verify the result
      assert.strictEqual(result, expectedValue)
    })
  })
})

Examples from Codebase

Reference existing test files as examples:

• apps/app/src/usecases/product/add.spec.ts - Decider tests
• apps/app/src/usecases/product/in-memory-projection.spec.ts - Projection tests
• apps/app/src/usecases/product/change-name.spec.ts - Event evolution tests

Summary: The TDD Mantra

ONE Test → Red → Green → Refactor → REPEAT

1. ONE Test: Write exactly ONE test case
2. Red: Run it, watch it fail
3. Green: Write minimal code to make THAT test pass
4. Refactor: Clean up while keeping tests green
5. REPEAT: Go back to step 1 for the NEXT test

The Golden Rules:

• Write ONE test at a time (never batch tests)
• Always watch tests fail before writing implementation
• Write the simplest code that passes the current test
• Refactor only after tests pass
• Tests are specifications, not documentation

What TDD is NOT:

• ❌ Writing all tests, then all implementation
• ❌ Writing implementation, then retrofitting tests
• ❌ Writing multiple tests before any implementation

What TDD IS:

• ✅ Tiny cycles: one test → code → refactor → next test
• ✅ Tests drive design incrementally
• ✅ Each test adds ONE new behavior