Development Workflow Skill

Version: 1.0.0 Category: Workflows Triggers: Implementing features, starting development, planning work

Quick Reference

6-Phase Workflow

Phase 1: User Requirements (READ user_prompt.md)
   ↓
Phase 2: YAML Configuration (Transform to structured config)
   ↓
Phase 3: Pseudocode Review (Design algorithm, get approval)
   ↓
Phase 4: TDD Implementation (Write tests first)
   ↓
Phase 5: Code Implementation (Implement following spec)
   ↓
Phase 6: Bash Execution (Single command entry point)

Phase 1: User Requirements

File: user_prompt.md

AI Actions:

• ✅ READ this file completely
• ✅ IDENTIFY ambiguous requirements
• ✅ ASK clarifying questions
• ✅ WAIT for user response
• ❌ NEVER EDIT this file
• ❌ NEVER DELETE this file

User's Role:

• Creates/updates user_prompt.md
• Reviews AI questions
• Provides guidance
• Approves approach

Phase 2: YAML Configuration

File: config/input/<feature-name>.yaml

Purpose: Transform natural language to structured, machine-readable config

AI Actions:

• Read user_prompt.md
• Generate YAML with all parameters explicit
• Add comments for complex settings
• Ask about ambiguous choices
• Save to config/input/

Structure:

metadata:
  feature: "feature-name"
  created: "2025-01-14"
  status: "draft"

requirements:
  input:
    - type: "csv"
    - path_type: "relative"
  processing:
    - calculate_statistics: true
  output:
    - format: "html"
    - interactive: true
  constraints:
    - max_response_time_sec: 5

Before Generating, Ask:

• What should be default values for X?
• Should Y be configurable or hardcoded?
• How should we handle Z edge case?

Phase 3: Pseudocode Review

File: docs/pseudocode/<feature-name>.md

Purpose: Design algorithm before implementation

Template:

## Module: DataLoader

FUNCTION load_csv(file_path):
  VALIDATE file_path is relative
  CHECK file_size <= 100MB
  TRY:
    data = read_csv(file_path)
    RETURN data
  CATCH error:
    LOG error
    RAISE DataLoadError

Before Generation, Ask:

• Should we use algorithm A or B?
• How should we handle edge case X?
• What's priority: speed or memory efficiency?

User's Role:

• Reviews pseudocode
• Validates logic
• Approves or requests changes
• Signs off before implementation

Phase 4: TDD Implementation

Workflow:

1. RED   → Write failing test
2. GREEN → Write minimal code to pass
3. REFACTOR → Improve code quality
4. REPEAT → For each feature

Key Rules:

• Tests written FIRST
• No code without test
• All tests pass before moving on
• Tests remain green during refactoring

First Subtask: Always "Write tests for [COMPONENT]"

Phase 5: Code Implementation

Follow:

• Pseudocode design exactly
• Keep tests passing at ALL times
• Match surrounding code style
• Add comments for "why" not "what"

Testing:

• Run full test suite frequently
• Fix failures immediately
• Never commit broken tests

Phase 6: Bash Execution

Pattern:

./scripts/run_feature.sh config/input/feature-name.yaml

Requirements:

• Single command entry point
• YAML file as input
• No complex tool chains
• Direct execution path

Quick Command Reference

# 1. Read requirements
vim user_prompt.md

# 2. Generate config
./scripts/generate_config.sh user_prompt.md > config/input/feature.yaml
vim config/input/feature.yaml  # User reviews

# 3. Generate pseudocode
./scripts/generate_pseudocode.sh config/input/feature.yaml > docs/pseudocode/feature.md
# User reviews and approves

# 4-5. TDD and implementation
vim tests/unit/test_feature.py  # Write tests
./tests/run_tests.sh            # Run (should fail)
vim src/modules/feature/        # Implement
./tests/run_tests.sh            # Run (should pass)

# 6. Execute
./scripts/run_feature.sh config/input/feature.yaml

Key Principles

Never Skip Phases

✅ All 6 phases for every feature ❌ Don't jump from requirements to code

TDD is Mandatory

✅ Test → Code → Refactor ❌ Code → Test (or no tests)

Ask Before Proceeding

✅ Clarify ambiguous requirements ✅ Wait for user approval at each phase ❌ Make assumptions

Documentation First

✅ Pseudocode approved before coding ❌ Implement first, design later

Simple Solutions

✅ Minimal code to pass tests ✅ Refactor for clarity, not cleverness ❌ Over-engineer solutions

Full Reference

See: @docs/modules/workflow/DEVELOPMENT_WORKFLOW.md

Use this when starting feature implementation, planning development work, or reviewing workflow compliance.