tdd-workflow

name	tdd-workflow
description	Test-driven development execution workflow with red-green-refactor cycle, increment decomposition, and pause/continue rules. Use when building features or fixing bugs using TDD. Framework-agnostic with language-specific configs for Python, TypeScript, Go, and Ruby delegation.

TDD Workflow

Core Principle

Never write implementation code before a failing test exists for that behavior.

Execution Flow

1. Setup

LANG = config_read("tech_stack", "unknown")
If LANG == "unknown": detect from project files (package.json → TypeScript, Gemfile → Ruby, go.mod → Go, pyproject.toml → Python)
If LANG == "unknown": AskUserQuestion("What language/framework?")

RUNNER = lookup LANG in [references/language-configs.md]
If LANG in [ruby]: delegate to `rspec-coder` or `minitest-coder` skill for runner details

INCREMENTS = decompose feature using patterns from [references/increments.md]
  - Read requirements/plan
  - Identify pattern (Data Transformation, CRUD, State Machine, Calculation, Integration)
  - Break into ordered increments: degenerate → happy path → variations → edge cases → errors

For each INCREMENT in INCREMENTS:
  TaskCreate(subject: INCREMENT.name, description: INCREMENT.test_description)

AskUserQuestion("Review increments. Adjust ordering or scope?", options: ["Looks good", "Modify"])

2. TDD Loop (per increment)

For each INCREMENT in INCREMENTS:
  TaskUpdate(INCREMENT.task_id, status: "in_progress")

  # RED
  Write failing test to real file (not code block)
  RUN = Bash(RUNNER.test_command)
  If RUN.status == PASS: STOP — investigate unexpected pass
  If RUN.status == FAIL (wrong reason): fix test, rerun
  PAUSE → show test + failure output, wait for user

  # GREEN
  Write minimal implementation code
  RUN = Bash(RUNNER.full_suite_command)
  If RUN.status == FAIL: show output, PAUSE → ask user to fix or hand off
  If RUN.status == PASS: auto-continue (no pause)

  # REFACTOR
  If improvement opportunities exist:
    Refactor implementation and/or test code
    RUN = Bash(RUNNER.full_suite_command)
    If RUN.status == FAIL: revert refactor, PAUSE → discuss
    If RUN.status == PASS: auto-continue (no pause)

  TaskUpdate(INCREMENT.task_id, status: "completed")
  Brief summary of what was implemented
  → immediately begin next increment (no pause between increments)