---

name: project-init description: Use when a non-developer user wants to create a new automation tool project. Sets up Python environment, project structure, and CLAUDE.md context file. Guides through environment setup → context gathering → structure initialization in user's preferred language.

Project Init

Overview

This skill initializes new automation tool projects for non-developers. It guides users through Python installation, gathers project context, and sets up a standardized documentation-driven project structure.

Use this skill when:

• User wants to start a new automation project
• User mentions creating a tool for their work
• User asks how to begin with AI-assisted coding

Workflow

Step 1: Check Python Installation

Goal: Ensure Python 3.12+ is installed before proceeding.

Process:

1. Ask user about their operating system (if not already known):

   • "어떤 운영체제를 사용하시나요? (Windows / macOS / Linux)"
   • "Which operating system are you using? (Windows / macOS / Linux)"

2. Check if Python is installed:

   Copypython --version
   # or
   python3 --version
   

3. If Python is NOT installed or version < 3.12:

   • Read the appropriate setup guide based on OS:
     • macOS: assets/SETUP-python-macos.md
     • Windows: assets/SETUP-python-windows.md
     • Linux: assets/SETUP-python-linux.md
   • Present the setup instructions in user's preferred language
   • Guide user through installation step by step
   • Wait for user to complete installation and verify

4. If Python 3.12+ is installed:

   • Confirm version: "Python 3.12.x가 설치되어 있습니다. 다음 단계로 진행하겠습니다."
   • Proceed to Step 2

Important:

• Language constraint explanation: Explain to user in their preferred language why we use Python:
  • Best for data processing and automation
  • Easy to install and use
  • Rich ecosystem for common tasks
  • TypeScript support can be added later if needed (e.g., n8n workflows)

---

Step 2: Gather User Context

Goal: Understand user's background, work domain, and project goals to personalize guidance.

Process:

1. Load the question reference:

   CopyRead: references/user-context-questions.md
   

2. Ask questions naturally (don't interrogate, have a conversation):

   • Preferred communication language
   • Work domain and tool purpose
   • Technical background
   • Confirm project type (Python-based local execution)

3. Example conversation flow:

   CopyAI: 안녕하세요! 어떤 분야에서 일하시나요?
       그리고 이 도구로 무엇을 자동화하고 싶으신가요?
   
   User: 마케팅팀에서 일하고, 매일 리포트 만드는 걸 자동화하고 싶어요.
   
   AI: 좋습니다! 프로그래밍 경험이 있으신가요?
   
   User: 엑셀 함수는 쓸 줄 알지만 코딩은 처음이에요.
   
   AI: 알겠습니다. Python 기반으로 컴퓨터에서 직접 실행되는 도구를
       만들게 됩니다. 먼저 프로젝트 구조를 만들겠습니다...
   

4. Store answers mentally (you'll use them in Step 3)

Reference: references/user-context-questions.md contains detailed question flow and examples.

---

Step 3: Create CLAUDE.md

Goal: Create a project-specific context file that will guide all future AI interactions.

Process:

1. Load the template:

   CopyRead: assets/CLAUDE.md.template
   

2. Fill in placeholders with answers from Step 2:

   • {{PREFERRED_LANGUAGE}} → e.g., "한국어" or "English"
   • {{WORK_DOMAIN}} → e.g., "마케팅"
   • {{TOOL_PURPOSE}} → e.g., "매일 하는 리포트 자동화"
   • {{TECHNICAL_BACKGROUND}} → e.g., "엑셀 함수는 쓸 줄 알지만 코딩은 처음"
   • {{PROJECT_TYPE}} → "Python automation tool"
   • {{DETAILED_TOOL_PURPOSE}} → More detailed explanation if user provided
   • {{KEY_REQUIREMENTS}} → Specific requirements mentioned by user
   • {{CONSTRAINTS}} → Any constraints (time, resources, etc.)

3. Write CLAUDE.md to project root:

   CopyWrite: CLAUDE.md (with filled template)
   

4. Confirm with user:

   • "CLAUDE.md 파일을 생성했습니다. 이 파일은 앞으로 AI가 프로젝트를 이해하는 데 사용됩니다."
   • "Created CLAUDE.md file. This will help AI understand your project context."

---

Step 4: Initialize Project Structure

Goal: Create standardized directory structure for documentation-driven development.

Process:

1. Run the initialization script:

   Copypython scripts/init_structure.py
   

   Or if user wants a specific directory:

   Copypython scripts/init_structure.py /path/to/project
   

2. The script creates:

   Copy├── docs/
   │   ├── knowledge-base/  # Domain knowledge, setup guides, API docs
   │   ├── plans/           # Implementation plans (PLN-XXX-YYYY-MM-DD-name.md)
   │   └── architecture/    # Component descriptions (how things work)
   ├── src/                 # Source code
   ├── tests/               # Test files
   └── .gitignore
   

3. Create .gitignore:

   The init_structure.py script automatically copies the gitignore template.

   Template includes:

   • Python artifacts (pycache, *.pyc, etc.)
   • Virtual environments (venv/, env/, etc.)
   • IDE files (.vscode/, .idea/, etc.)
   • Environment variables (.env, .env.local)
   • Configuration files (config.yaml, config.json, credentials.json) ← Security!
   • Test artifacts (.pytest_cache/, .coverage, etc.)

4. Initialize git repository (if not already initialized):

   Copygit init
   git add .
   git commit -m "Initial project structure"
   

5. Explain the structure to user in their preferred language:

   • docs/ - 모든 지식은 여기에 문서로 저장됩니다
   • docs/plans/ - 구현 계획이 PLN-001, PLN-002 형식으로 저장됩니다
   • src/ - 실제 코드가 들어갑니다
   • tests/ - 테스트 파일이 들어갑니다

---

Step 5: Next Steps Guidance

Goal: Guide user to the next skill (plan) and set expectations.

Process:

1. Explain the overall workflow:

   Copy프로젝트 초기화 완료! ✅
   
   다음 단계:
   1. [plan] - 무엇을 만들지 구체적으로 계획하기
   2. [implement] - 계획을 코드로 구현하기
   3. [debug] - 문제가 생기면 해결하기
   
   "plan 스킬을 사용하고 싶습니다" 라고 말씀해주세요.
   

2. Remind about CLAUDE.md:

   • "CLAUDE.md 파일은 프로젝트 루트에 있습니다"
   • "앞으로 모든 AI 대화에서 이 파일을 참고합니다"
   • "필요하면 언제든 수정할 수 있습니다"

3. Set expectations about Documentation = Source of Truth:

   • 모든 지식은 문서(docs/)에 저장됩니다
   • 코드는 블랙박스 구현입니다
   • 변경사항은 문서에 먼저 또는 동시에 반영됩니다

---

Key Principles

Documentation-Driven Development

All knowledge lives in docs, code is black box implementation.

• docs/knowledge-base/ - Domain knowledge, setup guides, API references
• docs/plans/ - Implementation plans with Plan IDs (PLN-001, PLN-002, ...)
• docs/architecture/ - How components work

Why?

• Non-developers understand documentation better than code
• AI can read docs to understand context without parsing code
• Changes are tracked in human-readable format

Plan ID System

• Format: PLN-001-YYYY-MM-DD-name.md
• Auto-incremented (PLN-001, PLN-002, ...)
• Used in git commit messages for traceability
• Example: PLN-001-2025-10-26-daily-report-automation.md

Language Constraint (Python)

Month 1: Python only (TypeScript deferred based on demand)

Reasons explained to user:

• Best for data processing and automation (primary use case)
• Easy to install and use
• Rich ecosystem for common automation tasks
• TypeScript support can be added as an extension if needed (e.g., n8n workflows)

Communication Language

All communication with user must be in their preferred language (stored in CLAUDE.md).

• If user starts in Korean → Korean
• If user starts in English → English
• If unclear → Ask: "어떤 언어로 대화하는 것이 편하세요? (한국어 / English)"

---

Troubleshooting

Python installation fails

macOS:

• Homebrew permission issues → Don't use sudo, run brew doctor
• Command not found → Restart terminal, check PATH

Windows:

• "python not recognized" → Check "Add Python to PATH" was selected
• Permission denied → Run as Administrator or change execution policy

Linux:

• python3.12 not found → Add deadsnakes PPA (Ubuntu/Debian)
• venv module missing → Install python3.12-venv package

Solution: Refer user to the appropriate SETUP guide and walk through step-by-step.

User has existing project structure

Don't overwrite!

1. Ask user if they want to:

   • Create in new directory
   • Merge with existing structure (careful!)
   • Skip structure creation (only create CLAUDE.md)

2. If merging:

   • Check for conflicts
   • Ask permission before creating each directory
   • Don't overwrite existing files

User wants TypeScript instead of Python

Explain the constraint:

• Current version supports Python only
• TypeScript can be added later as an extension
• For n8n workflows, TypeScript can be integrated when needed

If user insists:

• Acknowledge the limitation
• Offer to proceed with Python and note TypeScript as future extension
• Or defer to program facilitator for special cases

---

Resources Used

assets/

• CLAUDE.md.template - Project context template
• SETUP-python-macos.md - macOS Python 3.12 installation guide
• SETUP-python-windows.md - Windows Python 3.12 installation guide
• SETUP-python-linux.md - Linux Python 3.12 installation guide
• gitignore-python.template - Python project .gitignore template

scripts/

• init_structure.py - Cross-platform project structure initialization (Python)

references/

• user-context-questions.md - Detailed user context gathering guide with examples

---

Example Session

CopyUser: 새 프로젝트 시작하고 싶어요

AI: [Loads project-init skill]

AI: 좋습니다! 먼저 Python이 설치되어 있는지 확인하겠습니다.
    어떤 운영체제를 사용하시나요? (Windows / macOS / Linux)

User: macOS요

AI: [Checks Python version]
    Python 3.12.2가 설치되어 있습니다. ✅

    이제 프로젝트에 대해 몇 가지 여쭤볼게요.
    어떤 분야에서 일하시나요? 그리고 이 도구로 무엇을 자동화하고 싶으신가요?

User: 마케팅팀에서 일하고, 매일 여러 플랫폼에서 데이터 모아서
      리포트 만드는데 이걸 자동화하고 싶어요

AI: 좋은 프로젝트네요! 프로그래밍 경험이 있으신가요?

User: 엑셀 함수는 쓸 줄 아는데 코딩은 처음이에요

AI: 알겠습니다. Python 기반으로 진행됩니다.
    - 데이터 처리와 자동화에 최적화되어 있어요
    - 설치와 사용이 쉽습니다
    - 나중에 필요하면 TypeScript도 추가할 수 있습니다

    [Creates CLAUDE.md with context]
    [Runs init_structure.py]
    [Creates .gitignore]
    [Initializes git]

    프로젝트 초기화 완료! ✅

    다음 단계:
    1. [plan] - 무엇을 만들지 구체적으로 계획하기
    2. [implement] - 계획을 코드로 구현하기
    3. [debug] - 문제가 생기면 해결하기

    "plan 스킬을 사용하고 싶습니다" 라고 말씀해주세요.