implementing-code

# TDD Implementation Workflow

**NO PRODUCTION CODE WITHOUT A FAILING TEST FIRST.**

Code before test? Delete it. Start over.

## Mode Selection

**If continuing from a plan:**
- Read the plan file from `~/.claude/plans/`
- Resume from the last completed task
- Follow the established task order

**If new implementation:**
- Create task breakdown with TodoWrite
- Set up progress tracking

---

## For Each Task

### Step 1: TDD Implementation

```javascript
Task({
  subagent_type: "build-mode:task-implementer",
  description: "Implement task with TDD",
  prompt: `Task: [TASK]

Requirements:
[specific requirements]

Follow TDD: RED -> GREEN -> REFACTOR
Complete self-review checklist before reporting.`
})
```

**TDD Cycle:**
1. **RED:** Write failing test → verify fails for right reason
2. **GREEN:** Minimal code to pass → no over-engineering
3. **REFACTOR:** Clean up while green → tests pass after every change

### Step 2: Spec Review

```javascript
Task({
  subagent_type: "build-mode:spec-reviewer",
  description: "Verify spec compliance",
  prompt: `Review implementation for:
- Task: [TASK]
- Files modified: [FILES]

DO NOT trust implementer report. Read actual code.
3-pass review: Literal, Intent, Edge Cases.`
})
```

### Step 3: Quality Review (only if spec passes)

```javascript
Task({
  subagent_type: "build-mode:quality-reviewer",
  description: "Review code quality",
  prompt: `Review code quality for:
- Files: [FILES]

3-pass review: Style, Patterns, Maintainability.
Only report issues with 80%+ confidence.`
})
```

### Step 4: Error Handling

```javascript
Task({
  subagent_type: "build-mode:silent-failure-hunter",
  description: "Check error handling",
  prompt: `Hunt for silent failures in:
- Files: [FILES]

Find: empty catches, silent returns, broad catches.
Priority: P0 must fix, P1 should fix, P2 consider.`
})
```

### Step 5: Visual Test (if UI)

```javascript
Task({
  subagent_type: "build-mode:visual-tester",
  description: "Verify UI visually",
  pro