Tdd Workflow

TDD Workflow

Detailed test-driven development procedure with framework-specific commands.

Relationship to tdd-guide agent: The /tdd command dispatches the tdd-guide agent, which enforces the loop. This skill provides the detailed procedure and framework-specific commands that the agent references.

When to Use

• Referenced by tdd-guide agent during enforcement
• Directly when user needs framework-specific test commands
• Any new feature, bug fix, or refactoring that needs test coverage

Procedure

Phase 1: RED (Write Failing Test)

1. Identify the behavior to implement
2. Write the test that describes expected behavior
3. Run the test: it MUST fail
4. If it passes, the test is wrong or the feature already exists

# Framework detection
jest --testPathPattern={test_file} 2>/dev/null || \
pytest {test_file} 2>/dev/null || \
go test -run {test_name} ./... 2>/dev/null || \
mvn test -Dtest={test_class} 2>/dev/null

Phase 2: GREEN (Minimal Implementation)

1. Write the minimum code to make the test pass
2. No extra features, no premature optimization
3. Run the test: it MUST pass
4. If it fails, fix the implementation (not the test)

Phase 3: IMPROVE (Refactor)

1. Clean up implementation while keeping tests green
2. Extract functions, rename variables, remove duplication
3. Run full test suite after each change
4. Verify no regressions

Phase 4: COVERAGE

1. Check coverage meets 80% minimum:

``bash jest --coverage 2>/dev/null || \ pytest --cov={module} 2>/dev/null || \ go test -coverprofile=cover.out ./... 2>/dev/null ``

1. If below 80%, identify uncovered paths and add tests
2. Focus on branch coverage, not just line coverage

Output Format

TDD Cycle: {feature name}
RED:   Test written - {test file}:{test name} - FAILS as expected
GREEN: Implementation - {impl file} - Test PASSES
IMPROVE: Refactored {what changed}
Coverage: {percent}% ({above/below} 80% target)

Rules

• NEVER write implementation before a failing test
• NEVER modify tests to make them pass (fix the implementation)
• Each cycle should be small: one behavior per cycle
• Run the full suite after GREEN phase to catch regressions
• If coverage is below 80%, do not mark task as complete
• Log test failures clearly with file, line, and error message

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: mshadmanrahman
• Source: mshadmanrahman/pm-pilot
• License: MIT
• Homepage: https://github.com/mshadmanrahman/pm-pilot#readme

Install and usage instructions live in the source repository linked above.