Implement Feature with TDD

Overview

Implement features following strict Test-Driven Development workflow. Uses bmad-commands for deterministic file and test operations, ensures acceptance criteria are met through automated verification.

TDD Cycle:

1. RED: Write failing tests
2. GREEN: Implement code to pass tests
3. REFACTOR: Improve code while keeping tests green

Prerequisites

• Task specification exists at workspace/tasks/{task_id}.md
• bmad-commands skill available at .claude/skills/bmad-commands/
• Development environment configured
• Test framework installed (Jest or Pytest)

Workflow

Step 0: Load Task Specification

Action: Use read_file command to load task spec.

python .claude/skills/bmad-commands/scripts/read_file.py \
  --path workspace/tasks/{task_id}.md \
  --output json

Parse Response:

• Extract outputs.content for task specification
• Verify success == true
• Parse task spec sections:
  • Objective
  • Acceptance Criteria
  • Technical Specifications
  • Required Files

Required Sections in Task Spec:

## Objective
[What to build]

## Acceptance Criteria
1. [Testable requirement 1]
2. [Testable requirement 2]

## Technical Specifications
- API endpoints
- Data models
- Dependencies
- File locations

## Required Files
- src/path/to/implementation.ts
- tests/path/to/tests.test.ts

If task spec not found or invalid:

• Return error with message
• Suggest creating task spec first
• Halt implementation

Step 0.5: Check for External Packages and Fetch Documentation

CRITICAL STEP: Before writing any code, identify external packages and fetch current documentation.

Why This Step is Essential:

• Training data may be outdated
• Package APIs change between versions
• Prevents using deprecated or removed methods
• Ensures compatibility with installed version

Action 1: Identify External Packages

Scan task specification and requirements for package mentions:

• npm packages (JavaScript/TypeScript)
• PyPI packages (Python)
• Go modules
• Maven/Gradle dependencies (Java)
• Crates (Rust)
• NuGet packages (.NET)

Action 2: Check Installed Versions

# JavaScript/TypeScript
cat package.json | grep "package-name"
cat package-lock.json | grep -A 5 "package-name"

# Python
cat requirements.txt | grep package-name
pip show package-name

# Go
cat go.mod | grep package-name

# Rust
cat Cargo.toml | grep package-name

Action 3: Fetch Official Documentation

Use WebFetch tool to get current documentation:

JavaScript/TypeScript packages:

WebFetch(
  url: "https://www.npmjs.com/package/{package-name}",
  prompt: "Extract: installation, main API methods, usage examples, version compatibility"
)

# Or check official docs
WebFetch(
  url: "{official-docs-url}",
  prompt: "Extract: API reference for version {version}, method signatures, parameters, return types"
)

Python packages:

WebFetch(
  url: "https://pypi.org/project/{package-name}/",
  prompt: "Extract: installation, API reference, examples for version {version}"
)

WebFetch(
  url: "https://{package}.readthedocs.io/en/stable/",
  prompt: "Extract: API documentation, class methods, function signatures"
)

Go modules:

WebFetch(
  url: "https://pkg.go.dev/{module-path}",
  prompt: "Extract: package documentation, function signatures, examples"
)

Action 4: Verify API Compatibility

Compare fetched documentation with task requirements:

• Confirm method names exist
• Verify parameter types match
• Check return types
• Note any deprecated methods
• Identify breaking changes from older versions

Action 5: Document Findings

Add comments in code referencing documentation:

// Using axios@1.6.0
// Docs: https://axios-http.com/docs/api_intro
// Method: axios.get(url, config) returns Promise<AxiosResponse>
import axios from 'axios';

If Package is Unfamiliar:

• Spend 5-10 minutes reviewing documentation
• Understand core concepts and patterns
• Review example code
• Note common pitfalls
• Prefer official examples over training data

Common Documentation Sources:

If Documentation Not Found:

• Check package repository (GitHub, GitLab)
• Look for README.md
• Check CHANGELOG for version-specific changes
• Search for migration guides
• Escalate to user if critical package lacks documentation

Step 1: Analyze Requirements

Action: Break down acceptance criteria into test cases.

For each acceptance criterion:

1. Identify what needs to be tested
2. Determine test type (unit/integration/e2e)
3. Plan test structure
4. Identify edge cases

Example:

Acceptance Criterion: "User can log in with valid credentials"

Test Cases:
- [Unit] Should return user object when credentials valid
- [Unit] Should return null when email not found
- [Unit] Should return null when password incorrect
- [Integration] Should create session when login successful
- [Integration] Should not create session when login fails

Reference references/tdd-workflow.md for detailed TDD patterns and best practices.

Step 2: RED Phase - Write Failing Tests

Action: Write tests that cover all acceptance criteria.

Test Structure:

// tests/auth/login.test.ts
describe('AuthService.login', () => {
  // Happy path
  it('should return user when credentials valid', async () => {
    // Arrange
    // Act
    // Assert
  });

  // Error cases
  it('should return null when email not found', async () => {
    // ...
  });

  it('should return null when password incorrect', async () => {
    // ...
  });

  // Edge cases
  it('should handle empty email', async () => {
    // ...
  });
});

Run Tests:

python .claude/skills/bmad-commands/scripts/run_tests.py \
  --path . \
  --framework auto \
  --output json

Verify RED Phase:

• Tests should fail (NOT pass)
• Failures should be for the right reason (not syntax errors)
• Parse response: outputs.passed == false

If tests pass in RED phase:

• Tests are not valid (already passing code exists)
• Refine tests to be more specific
• Verify testing the right functionality

Step 3: GREEN Phase - Implement Code

Action: Write minimum code to make tests pass.

Focus on:

• Making tests pass (not elegance)
• Implementing simplest solution
• Following TDD cycle strictly

Implementation Pattern:

// src/auth/login.ts
export class AuthService {
  async login(email: string, password: string): Promise<User | null> {
    // Step 1: Find user
    const user = await this.userRepository.findByEmail(email);
    if (!user) return null;

    // Step 2: Verify password
    const valid = await this.passwordService.verify(password, user.passwordHash);
    if (!valid) return null;

    // Step 3: Return user
    return user;
  }
}

Run Tests:

python .claude/skills/bmad-commands/scripts/run_tests.py \
  --path . \
  --framework auto \
  --output json

Verify GREEN Phase:

• Parse response: outputs.passed == true
• Check outputs.coverage_percent >= 80
• Verify outputs.failed_tests == 0

If tests still failing:

• Review failure messages
• Fix implementation
• Re-run tests
• Repeat until GREEN

Step 4: REFACTOR Phase

Action: Improve code quality while keeping tests green.

Refactoring Targets:

• Remove duplication
• Improve naming
• Extract functions/methods
• Simplify conditionals
• Add type safety
• Improve error handling

After Each Refactor:

python .claude/skills/bmad-commands/scripts/run_tests.py \
  --path . \
  --framework auto \
  --output json

Verify tests stay green:

• outputs.passed == true after each refactor
• If tests break, revert refactor
• Only commit refactors that keep tests green

Reference references/refactoring-patterns.md for common refactoring patterns and techniques.

Step 5: Verify Acceptance Criteria

Action: Check that all acceptance criteria from task spec are met.

For each acceptance criterion:

1. Identify corresponding tests
2. Verify tests pass
3. Verify behavior matches requirement
4. Check edge cases covered

Automated Verification:

# Run full test suite
python .claude/skills/bmad-commands/scripts/run_tests.py \
  --path . \
  --framework auto \
  --output json

Check:

• outputs.passed == true
• outputs.coverage_percent >= 80
• outputs.total_tests >= expected_count
• All acceptance criteria have corresponding passing tests

Manual Verification:

• Review code against technical specifications
• Verify API contracts match
• Check data models are correct
• Ensure error handling is complete

Step 6: Final Validation

Action: Run comprehensive checks before completion.

Checks:

1. All tests passing
2. Coverage >= 80%
3. No syntax errors
4. No linting errors
5. All files created as specified
6. Code follows project standards

Run Tests One Final Time:

python .claude/skills/bmad-commands/scripts/run_tests.py \
  --path . \
  --framework auto \
  --output json

Acceptance Criteria Verification:

• ✅ tests_passing: outputs.passed == true
• ✅ coverage_threshold: outputs.coverage_percent >= 80
• ✅ no_syntax_errors: No syntax errors in output
• ✅ task_spec_loaded: Task spec was successfully loaded in Step 0
• ✅ all_requirements_met: All acceptance criteria verified in Step 5

Output

Return structured output with telemetry:

{
  "implementation_complete": true,
  "test_coverage_percent": 87,
  "files_modified": [
    "src/auth/login.ts",
    "tests/auth/login.test.ts"
  ],
  "tests_passed": true,
  "telemetry": {
    "skill": "implement",
    "task_id": "task-auth-002",
    "duration_ms": 45000,
    "files_modified_count": 2,
    "tests_total": 12,
    "tests_passed": 12,
    "tests_failed": 0,
    "coverage_percent": 87
  }
}

Error Handling

If any step fails:

1. Task Spec Not Found:

   • Error: "Task specification not found at workspace/tasks/{task_id}.md"
   • Action: Create task spec first

2. Tests Failing:

   • Error: "Tests failing after implementation"
   • Action: Review test failures, fix code, re-run

3. Coverage Below Threshold:

   • Error: "Test coverage {actual}% below threshold 80%"
   • Action: Add more tests to increase coverage

4. Syntax Errors:

   • Error: "Syntax errors detected in implementation"
   • Action: Fix syntax errors, re-run tests

References

Detailed documentation in references/:

• tdd-workflow.md: Complete TDD workflow with patterns
• refactoring-patterns.md: Common refactoring techniques
• testing-best-practices.md: Testing patterns and anti-patterns
• acceptance-criteria-guide.md: Writing good acceptance criteria

Using This Skill

From James subagent:

@james *implement task-auth-002

Directly:

Use .claude/skills/development/implement-v2/SKILL.md with input {task_id: "task-auth-002"}

With Orchestrator: Orchestrator calls this skill automatically during delivery workflow.

Philosophy

This skill embodies BMAD's 3-layer architecture:

• Uses Commands: bmad-commands for read_file, run_tests
• Provides Composition: Sequences commands into TDD workflow
• Enables Orchestration: Called by James subagent with routing

By using commands, this skill is:

• Observable: Telemetry at every step
• Testable: Commands have known contracts
• Composable: Can be used by other workflows
• Reliable: Deterministic command behavior