@crewx/sdk 0.8.0 → 0.8.1-rc.1

package/templates/documents/crewx-manual.md

@@ -0,0 +1,2278 @@
+<!-- AUTO-GENERATED: Do not edit manually. Run "npm run build:manual" to regenerate. -->
+
+# Welcome to AI Virtual Company
+
+> Imagine running your own company with AI employees that work 24/7, around the clock.
+
+CrewX is the tool that turns that imagination into reality.
+
+You are the **boss**. Hire AI employees, organize departments, and assign tasks.
+Your employees work in their areas of expertise, collaborate with each other, and report results.
+No coding knowledge required. Just a few clicks on the web interface and your company is up and running.
+
+---
+
+## For Example
+
+> "Write me a report draft" → An AI employee gathers materials and creates a draft for you.
+> "Review this document" → Another AI employee finds errors and suggests corrections.
+> Meanwhile, you grab a coffee and just check the results.
+
+AI employees come from **diverse backgrounds**. You can hire employees from different
+AI companies like Claude (Anthropic), Gemini (Google), and more. It's like headhunting.
+Each employee can have **specialties** configured, so you can assign them to roles that match their
+strengths — one for writing documents, another for fast analysis.
+
+---
+
+## New Here? Start Here
+
+| Guide | One-Line Summary |
+|-------|-----------------|
+| **Getting Started** | From opening shop to assigning your first task in 5 minutes |
+| **Setting Up AI Tools** | Connect AI providers like Claude, Gemini, etc. |
+
+---
+
+## Assigning Tasks to Employees
+
+| Guide | One-Line Summary |
+|-------|-----------------|
+| **Assigning Tasks to Employees** | Pick an employee from the main screen and assign work |
+| **Resuming a Previous Conversation** | Find and continue a previous task assignment |
+| **Checking Activity Logs** | Review what your employees have been doing |
+
+---
+
+## Hiring and Managing Employees
+
+| Guide | One-Line Summary |
+|-------|-----------------|
+| **Hiring a New Employee** | Add a new AI employee to your company |
+| **Editing Employee Info** | Change an employee's role, skills, or department |
+
+---
+
+## Running Your Company
+
+| Guide | One-Line Summary |
+|-------|-----------------|
+| **Checking Operating Costs** | View AI employee activity and costs at a glance |
+| **Company Settings** | Change theme, language, and integration settings |
+| **Managing AI Tools** | Install and keep AI tools up to date |
+
+---
+
+## Real-World Examples — How Sowon Labs Uses It
+
+> Use Sowon Labs' actual AI company structure as a reference to build your own organization.
+
+| Guide | One-Line Summary |
+|-------|-----------------|
+| **Building Your C-Suite** | CEO, CTO, CMO... AI executives transform your company |
+| **Getting a Secretary** | Create an all-purpose JARVIS-like secretary and free yourself from busywork |
+| **Running a Dev Team** | Compare two team structures and build the right team for your project |
+| **Building a Marketing Team** | Hand off blog and SNS content to your AI marketing team |
+| **Hiring Fun Employees** | Fortune telling, game guides... AI helps with hobbies too |
+| **Running Persona Surveys** | Ask AI customers first and validate reactions before launch |
+
+---
+
+## Tour Your Company (Screen-by-Screen Guide)
+
+Curious what each office (screen) looks like?
+
+| Office | Description |
+|--------|------------|
+| **Lobby — Main Screen** | See your employee list and assign tasks right away |
+| **Meeting Room — Chat** | One-on-one workspace for exchanging tasks with employees |
+| **Meeting Archives — Thread List** | Search and reopen past conversations |
+| **HR — Employee Management** | View employees, hire, or remove |
+| **Hiring Form — Employee Registration** | Form for entering a new employee's name, role, and skills |
+| **Admin Office — AI Tools** | Install and update AI tools |
+| **Finance — Cost Dashboard** | Usage and cost charts |
+| **Settings Room — Preferences** | Screen theme, language, and company settings |
+| **Library — Document Management** | Store and browse company documents |
+
+> See **screens/INDEX.md** for the full list of offices (screens).
+
+---
+
+## Need Help?
+
+**Q. What is CrewX?**
+It's a tool for building your own virtual company with AI employees. Hire employees, assign tasks, and review results.
+
+**Q. Can I use it without knowing how to code?**
+Yes, everything can be done from the web interface. No coding knowledge required.
+
+**Q. How many AI employees can I have?**
+There's no limit. Hire as many as you need. You can also divide them into departments to share responsibilities.
+
+**Q. Does it cost money?**
+When AI employees work, usage fees from their provider (Anthropic, Google, etc.) apply. You can check costs in real-time on the operating costs screen.
+
+**Q. Can employees collaborate with each other?**
+Yes, one employee can hand off tasks to another, or you can set up multiple employees to work together.
+
+---
+
+## Are You a Developer?
+
+For technical documentation, configuration file structures, and other development references, see the **Developer Guide**.
+
+---
+
+# Getting Started
+
+> "I want to try vibe-coding but... I'm stuck at installation."
+> Don't worry. CrewX handles the complex setup for you. You can get started in 5 minutes.
+
+---
+
+## This Guide Is For You If...
+
+- You've heard of Claude Code but gave up because installation looked too hard
+- You've tried asking ChatGPT to "write code" but got tired of copy-pasting
+- You want to build your own ideas without a developer
+
+---
+
+## The Full Flow — Just 3 Steps
+
+```
+Step 1. Check AI Tools → Step 2. Log In → Step 3. Choose a Team Preset → Start!
+```
+
+All you have to do is follow the on-screen instructions.
+
+---
+
+## Before You Start: Install CrewX
+
+If you haven't installed CrewX yet, run the following command in your terminal:
+
+```bash
+```
+
+> **Don't have npm?** Install Node.js first — download the LTS version from [nodejs.org](https://nodejs.org) and npm will be installed along with it.
+
+Once installed, check the version:
+
+```bash
+crewx --version
+```
+
+When a new version is released, you can update with:
+
+```bash
+```
+
+> When you launch CrewX, a **setup wizard** will appear automatically.
+> If you've already completed setup, it will go straight to the main screen.
+
+---
+
+## Step 1. Check AI Tools
+
+CrewX **automatically detects** whether AI coding tools are installed on your computer.
+
+| Status | Meaning |
+|--------|---------|
+| **Green** — Ready | Already installed and logged in. Ready to use! |
+| **Yellow** — Just need to log in | Already installed — we'll show you how to log in in the next step |
+| **Gray** — Not installed yet | That's fine for now. You can add it later |
+
+Select the tools you want to use and click **[Next]**.
+
+> **You can get started with just Claude Code!**
+> You can add the rest later when you need them.
+
+
+---
+
+## Step 2. Log In
+
+If any of your selected tools require login, CrewX will walk you through each one step by step.
+
+1. Click the **[Log In]** button shown on screen — the command is automatically copied
+2. Open your terminal and **paste** (Ctrl+V or Cmd+V)
+3. Done! CrewX automatically detects when login is complete
+
+> **"I don't know what a terminal is"**
+> A terminal is just a window where you type commands to your computer. It's not hard — just unfamiliar!
+> All you need to do here is one thing — paste. Do it once and it's automatic from then on.
+
+> If you're already logged into everything, this step is skipped automatically.
+
+
+---
+
+## Step 3. Choose a Team Preset
+
+This is the key part! A **preset** is a pre-configured AI team built by experts.
+Choose a preset that matches your needs and AI team members are set up automatically.
+
+| Preset | Best For |
+|--------|----------|
+| **Dev Team** | Those who want to build websites, apps, etc. |
+| **Marketing Team** | Those who want to create blog and SNS content |
+| **Planning Team** | Those who want to organize proposals and reports |
+| **Solo Founder** | Those who want to start quickly on their own |
+| **Custom** | Those who want to build a team from scratch |
+
+For example, choosing the "Dev Team" preset includes these AI team members:
+
+| AI Team Member | Role | You Can Ask Them To... |
+|----------------|------|----------------------|
+| **PM** | Project Manager | "Plan this feature", "Organize the schedule" |
+| **Dev Lead** | Development Lead | "Build this feature", "Fix this bug" |
+| **Frontend** | Frontend Developer | "Build the login page" |
+| **Backend** | Backend Developer | "Build the API", "Design the database" |
+| Others | QA, Designer, etc. | Composition varies by preset |
+
+> These configurations are templates based on real, actively running development teams.
+> No need to worry about "which agents should I set up and how?"
+
+Choose your preset and click **[Get Started]** — you're all set!
+
+
+---
+
+## All Set! Try Assigning Your First Task
+
+Your newly configured AI team members are now visible on the main screen.
+Just say what you want to do naturally:
+
+- Click the **Dev Lead** card → *"Build me a blog website"*
+- Click the **PM** card → *"Organize an e-commerce business plan"*
+- Click the **Frontend** card → *"Build a portfolio page"*
+
+Clicking a card automatically inserts the mention in the input field.
+Type your task and hit **Send** — and vibe-coding begins!
+
+
+> **Experience the magic of the first 30 seconds**
+> Say "Build me a blog" and the PM starts planning while the Dev Lead starts writing code.
+> Real files are created on your computer — this isn't just text output like ChatGPT.
+> It's a completely different experience from doing everything yourself.
+
+---
+
+## Skip — It's Always OK
+
+You can click **Skip** at any step to go straight to the main screen.
+You can always reconfigure later in the settings screen, so feel free to proceed without pressure.
+
+---
+
+## Want to Do More?
+
+- **Assigning Tasks to Agents** — Complete your first deliverable in 10 minutes
+- **Adding an Agent** — Add designers, QA, and more to your team
+- **Setting Up AI Tools** — Switch to your preferred AI tool
+
+
+---
+
+# Setting Up AI Tools
+
+> This guide explains how to connect AI coding tools like Claude, Gemini, etc. to CrewX.
+
+---
+
+## This Guide Is For You If...
+
+- You want to connect additional AI tools
+- You want to check the login status of your AI tools
+
+---
+
+## Supported AI Tools
+
+| Tool | Provider ID | Description |
+|------|------------|-------------|
+| **Claude Code** | `cli/claude` | Anthropic's coding agent |
+| **Gemini CLI** | `cli/gemini` | Google's coding agent |
+| **GitHub Copilot** | `cli/copilot` | GitHub's coding agent |
+| **Codex CLI** | `cli/codex` | OpenAI's coding agent |
+
+---
+
+## Tool-by-Tool Guide
+
+### Claude Code (`cli/claude`)
+
+Anthropic's AI coding agent. As CrewX's default provider, we recommend setting this up first.
+
+- **Install**: Download from [claude.ai/download](https://claude.ai/download)
+- **Auth**: Auto-guided login when running the `claude` command after installation
+
+### Gemini CLI (`cli/gemini`)
+
+Google's AI coding agent.
+
+- **Install**: `npm install -g @anthropic-ai/gemini-cli` or see official docs
+- **Auth**: Google account login
+
+### GitHub Copilot (`cli/copilot`)
+
+GitHub's AI coding agent.
+
+- **Install**: `npm install -g @githubnext/github-copilot-cli` or see official docs
+- **Auth**: GitHub account login
+
+### Codex CLI (`cli/codex`)
+
+OpenAI's coding agent. Supports GPT-4o-based code generation and analysis.
+
+- **Install**: `npm install -g @openai/codex` or see official docs
+- **Auth**: Set OpenAI API key (`OPENAI_API_KEY` environment variable)
+- **Usage examples**:
+  ```bash
+  crewx q "@codex 이 함수를 최적화해줘"
+  crewx x "@codex 테스트 코드 작성해줘" --thread="codex-work"
+  ```
+
+> Use the `crewx doctor` command to check installation and auth status for all tools at once.
+
+---
+
+## Configure from the CLI Management Screen
+
+You can view and manage installed tools from the CLI management screen (`/cli`).
+
+> For more details, see **Installing & Updating CLI Tools**.
+
+
+---
+
+## Recommended Next Guides
+
+- **Getting Started**
+- **Installing & Updating CLI Tools**
+- **Adding an Agent**
+
+
+---
+
+# Assigning Tasks to Agents
+
+> This guide explains how to pick an agent from the main screen and start a conversation.
+
+---
+
+## This Guide Is For You If...
+
+- You're assigning a task to an agent for the first time
+- You don't know what @mentions are
+- You're wondering "which agent should I ask?"
+
+---
+
+## Exploring the Main Screen
+
+When you open the app, the main screen (`/`) appears.
+
+### Agent Grid
+
+
+- **Avatar** (2-letter initials)
+- **Name** and a one-line role description
+
+
+### Suggested Task Cards
+
+Commonly used tasks are pre-configured. Click one and it goes straight into the input field.
+
+---
+
+## Starting a Conversation
+
+### Method 1: Click an Agent Card
+
+1. Click the desired agent card
+2. `@AgentName` is automatically entered in the input field
+3. Type your task and hit **Send** (or Enter)
+
+### Method 2: Click a Suggested Task Card
+
+1. Click a suggested task card
+2. `@Agent task description` is automatically entered
+3. Edit if needed, then hit **Send**
+
+### Method 3: Type Directly
+
+1. Type `@AgentName your task` directly in the input field
+2. Hit **Send**
+
+---
+
+## What You Can Do on the Chat Screen
+
+After sending, a new conversation (thread) is created and you're taken to the chat screen.
+
+| Element | Description |
+|---------|------------|
+| **Message Bubble** | Agent's response (supports markdown) |
+| **Activity Log** | Real-time display of what the agent is doing |
+| **Execution Status** | Shows `Running` / `Complete` in the header |
+| **Follow-up Messages** | Send additional questions/instructions from the input field |
+
+> For more details on activity logs, see **Checking Activity Logs**.
+
+
+---
+
+## Tips
+
+- Each agent has a different role. Choosing the right agent leads to better results.
+- You can exchange multiple messages in a single conversation.
+- An agent can also invoke other agents (Sub-Agent).
+
+---
+
+## Recommended Next Guides
+
+- **Resuming a Conversation**
+- **Checking Activity Logs**
+- **Adding an Agent**
+
+
+---
+
+# Resuming a Conversation
+
+> This guide explains how to find and continue a previous conversation.
+
+---
+
+## This Guide Is For You If...
+
+- You want to find a conversation from yesterday
+- You want to manage multiple conversations
+
+---
+
+## Finding Recent Conversations in the Sidebar
+
+Recent conversations are listed in the left sidebar.
+Click any conversation to resume it right away.
+
+
+---
+
+## Finding Conversations in the Thread List
+
+To see more conversations, go to the **Thread List** screen (`/threads`).
+
+| Feature | Description |
+|---------|------------|
+| **Search** | Search by conversation title or content |
+| **Filter** | Filter by agent or date |
+| **Sort** | Newest first, oldest first |
+
+Click a conversation to go to the chat screen.
+
+
+---
+
+## Continuing a Conversation
+
+Enter a follow-up message in the input field at the bottom of the chat screen, and the conversation continues within the same context.
+
+---
+
+## Recommended Next Guides
+
+- **Assigning Tasks to Agents**
+- **Checking Activity Logs**
+
+
+---
+
+# Checking Activity Logs
+
+> This guide explains how to check what tasks an agent has performed in real time.
+
+---
+
+## This Guide Is For You If...
+
+- You're curious about what the agent is doing
+- You want to review task results in detail
+
+---
+
+## What Are Activity Logs?
+
+Every action an agent performs during a task is recorded in the **activity log**.
+
+| Icon | Meaning |
+|------|---------|
+| 💬 | Agent's internal reasoning/judgment |
+| 🔧 | Command execution (Bash, file read, etc.) |
+| ✅ | Task completed |
+| ❌ | Error occurred |
+| 📄 | File read |
+| ✏️ | File modified |
+| 🔍 | Search |
+| 🤖 | Invoked another agent (Sub-Agent) |
+
+---
+
+## How to View Activity Logs
+
+Activity logs appear below the agent's message bubble on the chat screen.
+
+- **Collapse/Expand**: Click the log header
+- **Show More**: For long logs, click "Show N more" to expand
+
+
+---
+
+## Sub-Agent Activity
+
+When an agent invokes another agent, it appears in the **Sub-Agent Reply** area.
+
+- Execution status of each sub-agent (Running/Complete)
+- Time elapsed
+- Task summary
+
+<!-- Sub-Agent Reply screenshot to be added separately -->
+
+---
+
+## Recommended Next Guides
+
+- **Assigning Tasks to Agents**
+- **Resuming a Conversation**
+
+
+---
+
+# Adding an Agent
+
+> This guide explains how to create and register a new AI team member (agent).
+
+---
+
+## This Guide Is For You If...
+
+- You need an agent with a new role
+- You want to try creating an agent yourself
+
+---
+
+## Navigate to the Add Agent Screen
+
+1. Click **[+ Add Agent]** in the sidebar or on the agent management screen (`/agents`)
+2. Navigate to the add agent screen (`/agents/new`)
+
+
+---
+
+## Enter Basic Information
+
+| Field | Description | Example |
+|-------|------------|---------|
+| **ID** | Unique agent identifier (alphanumeric, cannot be changed) | `frontend_dev` |
+| **Name** | Display name | `Frontend Developer` |
+| **Role** | One-line description | `React/TypeScript UI Development` |
+| **Provider** | AI engine selection | `cli/claude`, `cli/gemini`, etc. |
+| **Model** | AI model to use | `opus`, `sonnet`, etc. |
+
+
+---
+
+## Save
+
+Enter all information and click **[Save]** to create the agent.
+A new card will appear in the agent grid on the main screen.
+
+> Agent settings are internally saved to `crewx.yaml`, and change history is automatically tracked.
+
+---
+
+## Recommended Next Guides
+
+- **Editing & Deleting Agents**
+- **Assigning Tasks to Agents**
+
+
+---
+
+# Editing & Deleting Agents
+
+> This guide explains how to modify or delete an existing agent's settings.
+
+---
+
+## This Guide Is For You If...
+
+- You want to change an agent's role or model
+- You want to delete an agent you no longer need
+
+---
+
+## Editing an Agent
+
+1. Go to the agent management screen (`/agents`)
+2. Click **[Edit]** on the agent you want to modify
+3. Modify the desired fields (ID cannot be changed)
+4. Click **[Save]**
+
+
+---
+
+## Deleting an Agent
+
+1. Go to the agent management screen (`/agents`)
+2. Click **[Delete]** on the agent you want to remove
+3. Click **[Confirm]** in the confirmation dialog
+
+> Deleted agents cannot be recovered. Please decide carefully.
+
+---
+
+## Change History
+
+Agent setting changes are automatically tracked.
+"View Change History" and "Revert" features are planned for future releases.
+
+---
+
+## Recommended Next Guides
+
+- **Adding an Agent**
+- **Assigning Tasks to Agents**
+
+
+---
+
+# Checking Usage
+
+> This guide explains how to monitor AI usage and costs from the dashboard.
+
+---
+
+## This Guide Is For You If...
+
+- You want to see how much AI you've used
+- You're curious about how much it costs
+
+---
+
+## Usage Dashboard
+
+Navigate to **Usage** in the sidebar or go to `/usage`.
+
+### Available Information
+
+| Item | Description |
+|------|------------|
+| **Daily Usage** | AI call counts and token usage by date |
+| **Monthly Summary** | Monthly total usage and cost trends |
+| **By Provider** | Compare usage across Claude, Gemini, etc. |
+
+
+---
+
+## Recommended Next Guides
+
+- **Configuring Settings**
+
+
+---
+
+# Configuring Settings
+
+> This guide explains how to change app settings like theme, language, and API keys.
+
+---
+
+## This Guide Is For You If...
+
+- You want to switch to dark mode
+- You want to switch between English and Korean
+- You want to configure API keys
+
+---
+
+## Settings Screen
+
+Navigate to **Settings** in the sidebar or go to `/settings`.
+
+### Configurable Items
+
+| Item | Description |
+|------|------------|
+| **Theme** | Switch between light/dark mode |
+| **Language** | Switch between Korean/English |
+| **API Keys** | Manage API keys per AI provider |
+
+
+---
+
+## Recommended Next Guides
+
+- **Getting Started**
+- **Setting Up AI Tools**
+
+
+---
+
+# Installing & Updating CLI Tools
+
+> This guide explains how to install AI coding tools (Claude Code, Gemini CLI, etc.) and keep them up to date.
+
+---
+
+## This Guide Is For You If...
+
+- You want to install a new AI tool
+- You see an update notification and aren't sure what to do
+
+---
+
+## CLI Management Screen
+
+Navigate to **CLI Management** in the sidebar or go to `/cli`.
+
+You can see the current status of all installed AI tools at a glance:
+- Installation status
+- Current version
+- Latest version
+- Login status
+
+
+---
+
+## Installing a New Tool
+
+1. Check for uninstalled tools on the CLI management screen
+2. Click the **[Install]** button
+3. Monitor installation progress
+4. Follow the login instructions after completion
+
+---
+
+## Updating
+
+1. Tools with available updates show a notification
+2. Click the **[Update]** button
+3. Confirm the update is complete
+
+---
+
+## Recommended Next Guides
+
+- **Setting Up AI Tools**
+- **Getting Started**
+
+
+---
+
+# Building Your C-Suite — AI Executives Transform Your Company
+
+> Set up C-level AI executives like CEO, CTO, CMO to establish strategic, technical, and marketing direction.
+
+For example, when evaluating a new business venture, you can ask the CTO about technical feasibility, the CPO about market viability, and the CLO about legal risks. Instead of one person worrying about everything, you get simultaneous reviews from multiple AI executives' perspectives.
+
+---
+
+## This Guide Is For You If...
+
+- You're juggling strategy, technology, and marketing all by yourself
+- You need "advisors to help see the big picture"
+- You want to use AI agents as **decision-making partners**, not just task executors
+
+---
+
+## How Sowon Labs Uses It
+
+Sowon Labs actually operates with an AI executive team. The human CEO makes final decisions while AI executives provide expert advice in their respective domains.
+
+### Organization Chart at a Glance
+
+```
+       사장님 (사람)
+          │
+    ┌─────┼─────┬─────┬─────┬─────┐
+    │     │     │     │     │     │
+   CEO   CTO   CPO   CMO   CSO   CLO
+  (비전) (기술) (제품) (마케팅) (전략) (법무)
+    │     │
+    │     ├── 개발자 (Claude)
+    │     ├── 개발자 (Codex)
+    │     └── 개발자 (Gemini)
+    │
+    └── 자비스 (비서)
+```
+
+### Each Executive's Role in One Line
+
+| Executive | What They Do | In a Word |
+|-----------|-------------|-----------|
+| **CEO** | Remembers vision and principles, resolves conflicts | "Here's what the boss intended" |
+| **CTO** | Tech strategy, delegates to developers | "The executive who delegates, not codes" |
+| **CPO** | Product roadmap, feature prioritization | "The person who decides what to build first" |
+| **CMO** | SNS marketing, GA4 analysis | "I'll handle the marketing" |
+| **CSO** | Business strategy, revenue model | "The profitable direction is this way" |
+| **CLO** | Contract review, legal risk | "Let me check if it's legally sound" |
+
+---
+
+## Follow Along: Creating a CEO
+
+### Step 1: Go to the Add Agent Screen
+
+Click **[Agent Management]** → **[+ Add Agent]** in the sidebar.
+
+### Step 2: Enter CEO Information
+
+| Field | Example Input |
+|-------|--------------|
+| **ID** | `ceo` |
+| **Name** | `CEO (Chief Executive Officer)` |
+| **Role** | `Vision guardian, direction setter, conflict resolver` |
+| **Provider** | `cli/claude` |
+| **Model** | `opus` (recommended — the smartest model) |
+
+### Step 3: System Prompt — "Making the CEO Remember Your Vision"
+
+Write your company's vision and principles in the system prompt. The CEO will use it as the basis for all decisions.
+
+**Prompt example:**
+
+```
+당신은 우리 회사의 CEO입니다. 반말로 대화하세요.
+
+핵심 역할:
+1. 회사 비전 기억 — "왜 이 방향인지" 설명
+2. 갈등 조정 — CTO vs CPO 의견 충돌 시 방향 제시
+3. 정렬 점검 — "이게 우리 목표와 맞아?" 질문
+
+우리 회사의 비전:
+(여기에 회사 비전을 적으세요)
+
+대화 스타일:
+- 결론부터 말하고 이유 설명
+- 모르는 것은 "모르겠어, 확인 필요해"라고 솔직히
+```
+
+Click **Save** and your CEO agent is created.
+
+---
+
+## Creating a CTO — Delegating Tech Strategy
+
+A CTO is an **executive who doesn't code directly but delegates to developer agents**.
+
+### Tips for Setting Up a "Delegating CTO"
+
+Add these principles to your prompt:
+
+```
+당신은 CTO입니다. 반말로 대화하세요.
+
+핵심 원칙:
+- 직접 코딩하지 않음. 개발 작업은 개발자 에이전트에게 위임
+- 기술 전략과 아키텍처 결정에 집중
+- 코드 리뷰와 품질 검토 수행
+
+위임 대상:
+- @개발자_claude: 복잡한 로직, 신규 기능
+- @개발자_codex: 리팩토링, 테스트
+- @개발자_gemini: 기술 리서치, 비교 분석
+```
+
+> For developer agent setup, see **Running a Dev Team**.
+
+---
+
+## Executive Usage Tips
+
+### Hold Meetings Between Executives (Using Threads)
+
+By asking multiple executives in the same **thread (chat room)** in sequence, you can gather opinions like a real meeting.
+
+**Example scenario:**
+
+1. To CEO: "Give me the top 3 priorities for this year"
+2. To CTO: "Pick which of the CEO's priorities are technically feasible"
+3. To CPO: "Which of the CTO's picks resonates most with users?"
+
+This lets you review a single topic from multiple perspectives.
+
+### Business Decisions to CEO, Tech Decisions to CTO
+
+Each executive has a different specialty. Choose the right executive for your question:
+
+| For questions like... | Ask this executive |
+|----------------------|-------------------|
+| "Are we heading in the right direction?" | CEO |
+| "Should we use this technology?" | CTO |
+| "What should we build first?" | CPO |
+| "How should we promote this?" | CMO |
+| "What's the revenue model?" | CSO |
+| "Is this legally sound?" | CLO |
+
+---
+
+## Remote Executive Consulting — Managing Multiple Projects at Once
+
+At Sowon Labs, the CTO and CPO from headquarters participate in other projects as **remote consultants**. Like a retained attorney — their main job is at HQ, but they provide consulting to other projects when needed.
+
+### "Executives at HQ, Projects Everywhere" Pattern
+
+```
+소원랩스 본사                  crewx OSS 프로젝트
+┌──────────┐                 ┌──────────────┐
+│  CTO ────│────원격 자문───→│  CTO (Remote) │
+│  CPO ────│────원격 자문───→│  CPO (Remote) │
+│  ...     │                 │  개발팀       │
+└──────────┘                 └──────────────┘
+```
+
+This way, you don't need to create executives for every project — manage them in one place while providing expertise across multiple projects.
+
+> For Remote agent setup, see the "Remote Consulting" section in **Running a Dev Team**.
+
+---
+
+## Recommended Next Guides
+
+- **Getting a Secretary** — Place an all-purpose secretary under your executives
+- **Running a Dev Team** — Build a dev team under your CTO
+- **Adding an Agent** — Detailed guide for the add agent screen
+
+> **Not working out?** Don't worry. AI agents can be deleted or have their roles modified at any time. Replace any executive you're not happy with immediately.
+
+---
+
+> **5-Minute Quickstart** — To get started right away, go to **Getting Started**.
+
+
+---
+
+# Getting a Secretary — Create an All-Purpose Secretary Like JARVIS
+
+> Delegate research, scheduling, email, and document organizing to an AI secretary so you can focus on core work.
+
+---
+
+## This Guide Is For You If...
+
+- You're losing time to small tasks
+- You need an "all-purpose helper that does everything"
+- You want a secretary like Iron Man's JARVIS that handles things when you just say the word
+
+---
+
+## What Sowon Labs' JARVIS Does
+
+Sowon Labs has an AI secretary called **JARVIS**. Serving as the CEO's personal assistant, it handles tasks like:
+
+| Task | Real Example |
+|------|-------------|
+| **Research** | "Research this market size" → Compiles and reports findings |
+| **Document Organizing** | "Organize today's meeting notes" → Creates meeting minutes |
+| **Schedule Management** | Google Calendar integration for checking and adding events |
+| **Email** | Gmail integration for checking and organizing emails |
+| **Notion Management** | Notion integration for document management |
+
+### The Secretary's Secret: Connected Skills
+
+The reason JARVIS is so versatile is because **skills** are connected to it. Adding skills to an agent gives it new capabilities.
+
+Skills that can be connected to Sowon Labs' JARVIS via MCP server configuration:
+
+| Skill | What It Does |
+|-------|-------------|
+| Gmail | Check, send, and organize emails |
+| Google Calendar | Check and add events |
+| Google Drive | File management |
+| Notion | Notion document management |
+| Web Browser | Internet search, information gathering |
+| PDF Conversion | Convert documents to PDF |
+
+---
+
+## Follow Along: Create Your Own Secretary
+
+### Step 1: Add an Agent
+
+Click **[Agent Management]** → **[+ Add Agent]** in the sidebar.
+
+### Step 2: Write the Secretary Role Prompt
+
+| Field | Example Input |
+|-------|--------------|
+| **ID** | `my_secretary` |
+| **Name** | `Secretary` (or any name you prefer) |
+| **Role** | `Personal secretary — research, document organizing, schedule management` |
+| **Provider** | `cli/claude` |
+| **Model** | `sonnet` (recommended — fast and accurate) |
+
+**System prompt example:**
+
+```
+당신은 나의 AI 비서입니다. 반말로 편하게 대화하세요.
+
+핵심 역할:
+1. 정보 수집 및 리서치 (시장 조사, 경쟁사 분석)
+2. 문서 관리 (회의록, 노트 정리)
+3. 일정 관리
+4. 이메일 관리
+
+응답 스타일:
+- 간결하고 명확하게
+- 실행 전에 계획을 먼저 요약
+- 완료하면 결과를 깔끔하게 보고
+```
+
+### Step 3: Connect Skills — Expanding Your Secretary's Capabilities
+
+Add the skills you need in the **Skills** tab of the agent settings. The more skills you add, the more your secretary can do.
+
+> Skill configuration may vary by provider (AI tool). For more details, see **Setting Up AI Tools**.
+
+---
+
+## Secretary Usage Scenarios
+
+### "Organize my to-do list for today"
+
+Ask your secretary to organize your daily tasks each morning — it checks your schedule and email to create today's to-do list.
+
+**Try saying:**
+> "Check my schedule for today and let me know if there are any urgent emails"
+
+### "Research this topic for me"
+
+When you need research, hand it off to your secretary. It searches the internet and compiles the findings.
+
+**Try saying:**
+> "Research the AI agent market size as of 2025. Include the top 3 players and market outlook"
+
+### "Organize the meeting notes and report back"
+
+Paste meeting content and it organizes it neatly.
+
+**Try saying:**
+> "Organize these meeting notes. Break them into decisions, action items, and next meeting agenda"
+
+---
+
+## Secretary Management Tips
+
+### Tone Setting: Formal vs Casual
+
+You can set your secretary's speaking style in the prompt:
+
+- **Formal**: `Respond politely in formal language` → "I've confirmed. Today's schedule is..."
+- **Casual**: `Chat casually and informally` → "Got it. Today's schedule is..."
+
+Sowon Labs' JARVIS uses a casual style.
+
+### Giving Your Secretary Memory
+
+To let your secretary remember previous conversations, connect the **memory skill**. It will then recall previous research findings and decisions, referencing them in future conversations.
+
+---
+
+## Recommended Next Guides
+
+- **Building Your C-Suite** — Place executives above your secretary
+- **Assigning Tasks to Agents** — How to give tasks to your secretary
+- **Adding an Agent** — Detailed guide for the add agent screen
+
+
+---
+
+# Running a Dev Team — Choose the Right Team Structure for Your Project
+
+> Compare Sowon Labs' two real-world patterns and build a team that fits your project's scale.
+
+---
+
+## This Guide Is For You If...
+
+- You want to deploy AI developers on a project
+- You want to "assign the same task to multiple AIs and compare results"
+- You're wondering how to structure and size your dev team
+
+---
+
+## Comparing Two Patterns
+
+Sowon Labs operates AI dev teams in two different ways.
+
+### Pattern A — Dedicated Assignment per Project (Sowon Labs HQ)
+
+```
+프로젝트A         프로젝트B         프로젝트C
+├── Claude 개발자  ├── Claude 개발자  ├── Claude 개발자
+├── Codex 개발자   ├── Gemini 개발자  └── Codex 개발자
+└── Gemini 개발자  └── Codex 개발자
+```
+
+- **2–3 dedicated developers** per project
+- **Simple and quick** to get started
+- Best for small projects
+
+### Pattern B — Full Team Setup (crewx OSS Project)
+
+```
+┌── CTO (원격 자문)
+│
+├── Dev Lead (팀장)
+│   ├── Claude 개발자
+│   ├── Gemini 개발자
+│   ├── Codex 개발자
+│   └── Copilot 개발자
+│
+├── QA 팀장
+│   └── 테스터
+│
+├── Release Manager
+│
+└── 리서치 에이전트
+    ├── Gemini CLI Analyzer
+    └── Mastra Analyzer
+```
+
+- **Structured team of 10+ members**
+- **Division of labor** from development → QA → release
+- Best for larger projects
+
+### Which Pattern Should You Choose?
+
+| Situation | Recommended Pattern |
+|-----------|-------------------|
+| Personal project, quick start | Pattern A (2–3 dedicated) |
+| Multiple projects simultaneously | Pattern A (dedicated per project) |
+| Large team, quality matters | Pattern B (full team) |
+| Need release/deploy processes | Pattern B (full team) |
+
+---
+
+## Pattern A — Dedicated Assignment per Project
+
+### Structure: 3 Members on the Same Project
+
+At Sowon Labs HQ, 3 developers report to the CTO, each using a different AI:
+
+| Developer | AI | Responsibility | Strengths |
+|-----------|-----|---------------|-----------|
+| Claude Developer | Claude (Anthropic) | Complex logic, new features | Architecture, design |
+| Codex Developer | Codex (OpenAI) | Refactoring, testing | Thoroughness, accuracy |
+| Gemini Developer | Gemini (Google) | Tech research, comparative analysis | Search, bulk analysis |
+
+### Follow Along: Build a 3-Person Dev Team
+
+**Creating the Claude Developer:**
+
+| Field | Example Input |
+|-------|--------------|
+| **ID** | `my_dev_claude` |
+| **Name** | `Developer (Claude)` |
+| **Role** | `New feature development, complex logic implementation` |
+| **Provider** | `cli/claude` |
+| **Model** | `sonnet` |
+
+Create the Gemini developer (`cli/gemini`) and Codex developer (`cli/codex`) the same way.
+
+**Prompt tip**: Put the same project description in all 3 prompts. Only the AI differs — the context should be identical!
+
+### Usage: Developing from Multiple Perspectives
+
+Just as diverse human team members produce better results, the same applies to AI.
+
+Normally, **let Claude handle the development.** It understands instructions well and grasps intent accurately. But sometimes there are bugs that Claude just can't fix no matter what. Strangely, **Codex sometimes solves it** when you hand it over. That's because each coding agent product approaches problems differently.
+
+This is the real reason for having multiple AIs:
+
+- **Claude**: Primary developer. Understands complex requirements well, from design to implementation
+- **Codex**: Relief pitcher. Takes a different approach to break through when Claude is stuck
+- **Gemini**: Research specialist. Strong at tech comparisons and exploring alternatives
+
+If one can't do it, hand it off to another agent. **The same problem gets solved differently by different products.**
+
+### Dedicated Developers per Project
+
+Sowon Labs assigns dedicated developers to each project:
+
+- **InMyRace Project** → InMyRace dedicated developer
+- **Quickstart Package** → Quickstart dedicated developer
+- **Launcher Project** → Launcher dev team (Claude, Gemini, Codex)
+
+Each dedicated developer has their **working directory** set to the respective project folder, allowing them to focus on that project's code.
+
+---
+
+## Pattern B — Full Team Setup (Dev, QA, Release)
+
+The crewx open-source project runs a more structured team. Over 10 agents actually divide roles and work together.
+
+### 10-Member Team Structure
+
+| Role | Count | What They Do |
+|------|-------|-------------|
+| **Dev Lead** (Team Lead) | 1 | Issue assignment, agent coordination |
+| **Developers** | 4 | Code development (Claude, Gemini, Codex, Copilot) |
+| **QA Lead** | 1 | Test strategy, release approval |
+| **Tester** | 1 | Test execution, bug reports |
+| **Release Manager** | 1 | Branch management, deployment |
+| **Research** | 2 | Competitor analysis, tech research |
+
+### Multi-Provider Strategy
+
+The core strategy of crewx OSS is **"same prompt, same docs, different AI"**.
+
+All 4 developers share the same project documents and development rules. The only difference is the **AI engine**:
+
+| Developer | AI Engine | Characteristics |
+|-----------|----------|----------------|
+| Claude Developer | Anthropic Claude | Strong at design, complex logic |
+| Gemini Developer | Google Gemini | Strong at research, bulk analysis |
+| Codex Developer | OpenAI Codex | Strong at thorough implementation, refactoring |
+| Copilot Developer | GitHub Copilot | Natural conversation, cost-effective |
+
+### Separate QA Team
+
+Code created by developers is **verified by different agents**. Just like how real organizations separate dev and QA teams.
+
+- **QA Lead**: Sets test strategy and decides whether to release
+- **Tester**: Runs actual tests and files bug reports when issues are found
+
+### Release Manager
+
+When code is complete, the release manager handles the entire deployment process:
+- Creating RC (release candidate) branches
+- Coordinating with the QA team for test verification
+- Final merge and deployment
+
+### Dev Lead
+
+The Dev Lead doesn't write code directly. Instead:
+- Assigns GitHub issues to developers
+- Coordinates work between developers
+- Manages project status and reporting
+
+### Follow Along: Setting Up a 10-Person Full Team (Step by Step)
+
+1. **Start with 4 developers** (Claude, Gemini, Codex, Copilot)
+2. **Add Dev Lead** — Team lead role, no direct coding
+3. **Add QA Lead** — Test strategy planning
+4. **Add Tester** — Dedicated test execution
+5. **Add Release Manager** — Deployment process management
+
+> You don't need to create everyone at once! Start with 2–3 developers and add QA and release manager as needed.
+
+---
+
+## Going Further
+
+### Remote Consulting — Bringing in HQ Executives
+
+This feature lets you **invoke an agent from another project in your project**. Like an external consultant or advisor.
+
+A **Remote agent** is a feature that lets you invoke an agent from another project in your project. Like an external consultant or advisor.
+
+#### Sowon Labs Case Study
+
+The crewx OSS project registers Sowon Labs HQ's CTO and CPO as **Remote agents** for tech and product consulting:
+
+- Need an architecture decision → Remote query to HQ CTO
+- Need product direction confirmation → Remote query to HQ CPO
+
+When adding an agent, set the provider to **Remote** type and specify the path to the project where the original agent resides. You can leverage all of the original agent's settings (prompt, skills, documents, etc.) as-is.
+
+### Specialized Research Agents
+
+When you need to reference other open-source projects, assigning your own developers mixes perspectives. So Sowon Labs created **separate agents dedicated to analyzing specific projects**. They don't know your code, but focus entirely on the analysis target:
+
+| Agent | Mission |
+|-------|---------|
+| Gemini CLI Analyzer | Analyzing Google's Gemini CLI source code |
+| Mastra Analyzer | Analyzing the Mastra framework |
+
+These agents are **placed in external project folders**, analyzing that code and providing insights applicable to your project.
+
+Two key points when creating a research agent:
+
+1. **Set the working directory to the analysis target**: Point the agent's working directory to the project folder you want to analyze
+2. **Be specific about the mission in the prompt**: "Analyze the OO pattern in this project and find ways to apply it to our project"
+
+**Prompt example:**
+
+```
+당신은 [분석 대상 프로젝트] 전문 분석가입니다.
+
+미션: [분석 대상]의 [특정 기능]을 분석해서
+우리 프로젝트에 적용할 수 있는 패턴을 찾아주세요.
+
+분석 포인트:
+- 아키텍처 패턴
+- 재사용 가능한 코드
+- 우리에게 적용할 때 주의할 점
+```
+
+---
+
+### Sharing Project Documents
+
+In crewx OSS, the **documents** system ensures all developers follow the same rules:
+
+| Document | Content |
+|----------|---------|
+| Branch Protection Rules | "No direct commits to main branch" |
+| Development Process | Workflow, code review procedures |
+| Release Workflow | RC version rules, deployment procedures |
+| Report Structure | Test report format |
+
+This ensures that whether you have 4 developers or 10, everyone works by the same rules. Even when you add a new developer agent, just sharing the documents makes them follow team rules immediately.
+
+---
+
+## Recommended Next Guides
+
+- **Building Your C-Suite** — Place a CTO above your dev team
+- **Building a Marketing Team** — Build a marketing team alongside your dev team!
+- **Adding an Agent** — Detailed guide for the add agent screen
+
+
+---
+
+# Building a Marketing Team — Hand Off Blog and SNS to Your AI Team
+
+> From content planning to blog publishing and SNS posting, your AI marketing team handles it all.
+
+---
+
+## This Guide Is For You If...
+
+- You need to write blog posts but don't have time
+- You want to do SNS marketing systematically
+- Hiring a marketing professional feels too expensive
+
+---
+
+## Sowon Labs Marketing Team Structure
+
+Sowon Labs runs a 3-person AI marketing team:
+
+```
+CMO (마케팅 총괄)
+├── 블로그 매니저 (콘텐츠 제작)
+└── 템플릿 매니저 (소개 페이지 관리)
+```
+
+### CMO — Marketing Director
+
+Sowon Labs' CMO handles:
+- **SNS Marketing**: Writing promotional posts for X (Twitter), Threads
+- **GA4 Analysis**: Reviewing Google Analytics data to measure marketing effectiveness
+- **Content Strategy**: Deciding what topics to write about and which channels to use
+
+### Blog Manager — Content Creation & Publishing
+
+The Blog Manager actually writes and publishes articles:
+- Technical blog posts
+- Tutorials, use case articles
+- SEO-optimized content
+
+### Template Manager — Landing Page Management
+
+Manages the website's about pages and landing pages.
+
+---
+
+## Follow Along: Setting Up a Marketing Team
+
+### Step 1: Create a CMO — Marketing Strategist
+
+| Field | Example Input |
+|-------|--------------|
+| **ID** | `my_cmo` |
+| **Name** | `CMO (Marketing Director)` |
+| **Role** | `SNS marketing, content strategy, data analysis` |
+| **Provider** | `cli/claude` (or your preferred AI) |
+| **Model** | `sonnet` |
+
+**System prompt example:**
+
+```
+당신은 마케팅 총괄(CMO)입니다. 반말로 대화하세요.
+
+핵심 역할:
+1. SNS 마케팅 — X(트위터), Threads 콘텐츠 제작
+2. 콘텐츠 전략 — 블로그 주제 기획
+3. 브랜드 톤 관리
+
+콘텐츠 톤:
+- 기술적이지만 친근하게
+- 개발자 친화적
+- 해시태그: #CrewX #AI #Automation
+
+금지:
+- 과장된 표현
+- 근거 없는 주장
+```
+
+### Step 2: Create a Blog Manager — Writing Expert
+
+| Field | Example Input |
+|-------|--------------|
+| **ID** | `blog_writer` |
+| **Name** | `Blog Manager` |
+| **Role** | `Blog content creation, technical writing` |
+| **Provider** | `cli/claude` |
+| **Model** | `sonnet` |
+
+**Prompt tip**: Adding "When writing blog posts, show the outline first and write the full article after confirmation" ensures the boss reviews the direction before the article is completed.
+
+### Step 3: Connect to Other Projects (Optional)
+
+If your blog is in a separate project like Sowon Labs, you can connect it via **Remote agent**. When you say "Write a blog post" from HQ, the blog project's specialized agent runs.
+
+---
+
+## Marketing Team Usage Scenarios
+
+### "Write 3 blog posts this week"
+
+Ask the CMO to plan this week's content, then have the Blog Manager write the actual articles.
+
+**Try saying:**
+> To CMO: "Recommend 3 topics for this week's blog. Focus on our product's key features"
+>
+> To Blog Manager: "Write a blog post draft on the first topic the CMO recommended"
+
+### "Create promotional copy for X (Twitter)"
+
+Leave SNS posting to the CMO.
+
+**Try saying:**
+> "We're posting a new feature release on X. Create 3 promotional messages under 280 characters"
+
+### "Analyze GA4 data and report"
+
+Connect web browser and GA4 skills to the CMO, and it can analyze Google Analytics data and report back.
+
+**Try saying:**
+> "Compile this month's blog visitor trends and top 5 popular posts"
+
+---
+
+## Tone Guideline Tips
+
+### Embedding Brand Tone in Prompts
+
+Adding a brand tone guide to your marketing agent's prompt ensures all content comes out in a consistent tone.
+
+**Example:**
+
+```
+우리 브랜드 톤:
+- 전문적이지만 딱딱하지 않게
+- "~해보세요" 보다는 "~해봐요" (반말은 아니지만 친근하게)
+- 기술 용어는 쉬운 말로 풀어서
+- 이모지 적당히 사용 OK
+```
+
+### Automatic Hashtag Inclusion
+
+Add frequently used hashtags to the prompt so you don't need to specify them every time:
+
+```
+SNS 포스팅 시 항상 포함할 해시태그:
+#CrewX #AIAgent #생산성
+```
+
+---
+
+## Recommended Next Guides
+
+- **Building Your C-Suite** — Place a CEO above the CMO for overall direction
+- **Running a Dev Team** — Build a dev team alongside your marketing team
+- **Adding an Agent** — Detailed guide for the add agent screen
+
+
+---
+
+# Hiring Fun Employees — Not Just Work? AI Helps with Hobbies Too
+
+> AI employees don't just work. Fortune telling, game guides, workout routines, recipe search, book discussion partners — create your own AI helpers for hobbies and daily life.
+
+---
+
+## This Guide Is For You If...
+
+- You want to "try something fun with AI"
+- You're tired of only making work-focused agents
+- You want to explore what's possible with AI agents
+
+---
+
+## Sowon Labs' "Fun Department"
+
+Sowon Labs doesn't just have work-focused agents. There are agents for hobbies and fun too!
+
+### Fortune Teller — AI Reads Your Fortune?
+
+Yes, it really reads your fortune. Give it your birth date and time, and it analyzes your Four Pillars of Destiny (Saju), including major fortune cycles (10-year periods) and annual fortunes.
+
+**Real conversation example:**
+> User: "I'm a man born on May 15, 1990 at 2 PM"
+>
+> Fortune Teller: "Your day master is Byeonghwa (丙火)! You're the bright, energetic type like the sun. This year..."
+
+Of course, this is **for entertainment only**. There's no scientific basis, so don't take it too seriously!
+
+### Lineage Classic Helper — 10-Year Veteran Gamer AI
+
+Do you play Lineage Classic? There's an AI that answers like a **10-year veteran** about item specs, monster info, hunting ground recommendations, and more.
+
+- **Powerbook Search**: Instant lookup of official data (items, skills, monsters)
+- **Community Search**: Real-world strategies from Inven, DC Inside, etc.
+
+**Real conversation example:**
+> User: "Can a knight survive without Combat Blessing?"
+>
+> Helper: "A knight without CB is like bread without filling lol. It's fine early on but past level 50..."
+
+### Lineage Powerbook Helper — Official Data Search Specialist
+
+There's also a lightweight helper that specializes in searching Powerbook data. Useful for quickly checking specs.
+
+---
+
+## Follow Along: Creating a Fortune Teller
+
+### Step 1: Add an Agent
+
+| Field | Example Input |
+|-------|--------------|
+| **ID** | `saju_reader` |
+| **Name** | `Fortune Teller` |
+| **Role** | `Four Pillars analysis, fortune reading` |
+| **Provider** | `cli/claude` |
+| **Model** | `sonnet` |
+
+### Step 2: Fortune Expert Prompt
+
+**System prompt example:**
+
+```
+당신은 사주팔자 전문가입니다. 반말로 친근하게 대화하세요.
+
+역할:
+- 생년월일시를 받아 사주팔자 분석
+- 대운(10년 주기) 및 세운(연도별) 분석
+- 오행 균형 해석, 성격·적성·운세 조언
+
+대화 스타일:
+- 친근하고 재미있게
+- 어려운 용어는 쉽게 풀어서
+- 구체적인 조언과 함께
+
+주의사항:
+- 반드시 "재미와 참고용"임을 명시
+- "맹신하지 마세요" 항상 언급
+- 긍정적이고 건설적인 조언 중심
+```
+
+### Step 3: Connect a Fortune-Telling Skill
+
+For accurate fortune calculations, connect a **Saju skill**. The skill handles the traditional calendar calculations, allowing the AI to interpret based on accurate fortune data.
+
+---
+
+## Follow Along: Creating a Game Helper
+
+### Nailing the "10-Year Veteran" Character
+
+The key to a game helper is **character (persona)** setup. Instead of stiff AI language, make it talk like a real gamer for maximum fun.
+
+**Prompt example:**
+
+```
+넌 리니지 클래식 10년차 고인물이야. 반말로 편하게 대화해.
+
+답변 스타일:
+- 게이머처럼: "카배 없는 기사는 앙코없는 찐빵"
+- 약어 자연스럽게: 카배, 리덕, 블어, 앱블
+- 실전 팁: "솔플할땐 이거, 파티플땐 저거"
+
+질문받으면 검색부터!
+- 스펙 질문 → 파워북에서 공식 데이터 확인
+- 사냥터/세팅 질문 → 커뮤니티 검색
+- 둘 다 없으면 "파워북/인벤 뒤져봤는데 없네 ㅠ"
+```
+
+### Connecting Search Skills for Powerbook/Community
+
+Connect a **search skill** to your game helper so it can directly look up information from Powerbook (official data) and communities (Inven, DC Inside, etc.).
+
+---
+
+## Using Multiple AIs as Employees — 4 Providers
+
+CrewX lets you use multiple AIs as team members. Choose a provider when adding an agent.
+
+### Supported Providers
+
+| Provider | AI | Characteristics |
+|----------|-----|----------------|
+| `cli/claude` | Anthropic Claude | Strong at design, complex logic |
+| `cli/gemini` | Google Gemini | Strong at research, bulk analysis |
+| `cli/codex` | OpenAI Codex | Strong at thorough implementation, refactoring |
+| `cli/copilot` | GitHub Copilot | Natural conversation, cost-effective |
+
+You can also use `api/ollama` (local AI) or `remote/*` (remote agents).
+
+---
+
+## Your Own Hobby AI Ideas
+
+Beyond fortune telling and games, you can create all kinds of hobby AIs. All you need is an idea!
+
+### Recipe Helper
+
+```
+역할: 요리 레시피 추천, 재료 대체 제안
+프롬프트 핵심: "냉장고에 있는 재료를 말하면 만들 수 있는 요리 추천"
+연결 스킬: 웹 검색 (레시피 사이트 검색)
+```
+
+### Workout Routine Coach
+
+```
+역할: 운동 루틴 설계, 자세 교정 조언
+프롬프트 핵심: "사용자의 체력 수준에 맞는 운동 추천"
+연결 스킬: 메모리 (운동 기록 저장)
+```
+
+### Book Discussion Partner
+
+```
+역할: 책 내용 토론, 질문 주고받기
+프롬프트 핵심: "소크라테스식 질문으로 생각을 이끌어내기"
+연결 스킬: 메모리 (읽은 책 기록)
+```
+
+Let your imagination run wild. There's no rule that AI employees can only do work!
+
+---
+
+## Recommended Next Guides
+
+- **Getting a Secretary** — Create an all-purpose secretary too
+- **Adding an Agent** — Detailed guide for the add agent screen
+- **Assigning Tasks to Agents** — Talk to the agents you've created
+
+---
+
+> **5-Minute Quickstart** — To get started right away, go to **Getting Started**.
+
+
+---
+
+# Running Persona Surveys — Ask AI Customers Before You Launch
+
+> New features, documents, ideas... Validate reactions with AI customer personas before showing them to real customers.
+
+---
+
+## This Guide Is For You If...
+
+- You're curious about reactions to a new feature or service
+- You wrote a document and want to check "can a regular person understand this?"
+- You want to quickly gather opinions on pricing or marketing copy
+- You want to gauge expected reactions before real user interviews
+
+---
+
+## How Sowon Labs Uses It
+
+At Sowon Labs, every time a manual is written, **a survey is sent to 5 AI customer personas**.
+
+### Real Case: Manual Comprehension Validation
+
+When a new manual is written, it's **pre-validated by surveying 5 AI customer personas**. When testing 10 use case manuals, **all 5 personas responded positively that they "could follow along immediately"**, and **3 FAQ improvements** were identified such as "How much does it cost?" and "Is data security okay?" — and the documents were strengthened accordingly. First-round feedback in 30 minutes, with no human interviews.
+
+The persona composition used for validation:
+
+| Persona | Profile | Validation Point |
+|---------|---------|-----------------|
+| 20s Designer | Interested in UI/UX, unfamiliar with CLI | "Are the screen descriptions intuitive?" |
+| 30s Marketer | Content planning experience, tech beginner | "Can I use this for marketing work right away?" |
+| 20s Creator | Active on SNS, eager to try new tools | "Does it look fun and easy?" |
+| 40s Executive | Business decision-maker, values efficiency | "Does it feel worth the investment?" |
+| 30s Planner | Project management experience, analytical | "Can I apply this to real work immediately?" |
+
+**Results:** Each persona honestly shares "parts that are hard to understand," "things they want to know more about," and "overall impressions." Incorporating this feedback to improve documents makes them much easier to understand when real customers see them.
+
+---
+
+## Follow Along: Running a Persona Survey
+
+### Step 1: Create a Survey Manager Agent
+
+Click **[Agent Management]** → **[+ Add Agent]** in the sidebar.
+
+| Field | Example Input |
+|-------|--------------|
+| **ID** | `survey_manager` |
+| **Name** | `Survey Manager` |
+| **Role** | `Customer persona survey management` |
+| **Provider** | `cli/copilot` (recommended) |
+| **Model** | (default) |
+
+**System prompt example:**
+
+```
+당신은 고객 반응 검증 전문가입니다.
+
+핵심 역할:
+1. 고객 페르소나를 만들어 다양한 관점에서 피드백을 수집합니다
+2. 페르소나 스킬을 활용해 각 고객 유형의 솔직한 반응을 시뮬레이션합니다
+3. 수집한 피드백을 요약하고 개선점을 제안합니다
+
+응답 스타일:
+- 각 페르소나의 반응을 구분해서 정리
+- 공통 의견과 특이 의견을 분리
+- 구체적인 개선 제안 포함
+```
+
+**Skill connection:** Be sure to add the persona skill.
+
+### Step 2: Create Customer Personas
+
+Start a conversation with the survey manager agent and ask it to create personas.
+
+**Try saying:**
+> "Create 5 customer personas by our service's main target customer types:
+> 1. 20s Designer — UI/UX interest, tech beginner
+> 2. 30s Marketer — Content planner, extensive practical experience
+> 3. 20s Creator — Active on SNS, eager to try new tools
+> 4. 40s Executive — Decision-maker, ROI-focused
+> 5. 30s Planner — Project management, analytical"
+
+The agent uses the persona skill to create profiles for each person. Profiles include values, speaking style, and behavior patterns so they can react like real people.
+
+### Step 3: Run the Survey
+
+Once personas are ready, show them what you want to validate and ask for reactions.
+
+**Try saying:**
+> "Show the following manual to each of the 5 personas and have them answer these questions:
+> 1. Could you follow along after reading this document? (1–5 points)
+> 2. Which parts are hard to understand?
+> 3. What else would you like to know?
+> 4. What's your overall impression?
+>
+> (paste manual content here)"
+
+### Step 4: Compile Feedback and Apply
+
+The agent compiles all 5 personas' reactions. Results look like this:
+
+| Persona | Score | Key Feedback |
+|---------|-------|-------------|
+| 20s Designer | 4/5 | "Screenshots would be nice" |
+| 30s Marketer | 3/5 | "I don't know what 'provider' means" |
+| 20s Creator | 5/5 | "Looks fun! I'll try it right away" |
+| 40s Executive | 3/5 | "Wish the cost-benefit was clearer" |
+| 30s Planner | 4/5 | "The step-by-step explanations are well-organized" |
+
+**Common improvement:** Change the technical term "provider" → "AI source company"
+
+By incorporating this feedback to revise documents, you can improve quality before showing them to real customers.
+
+---
+
+## Usage Scenarios
+
+### New Feature Reaction Check
+
+When you've developed a new feature, check customer reactions before launch.
+
+**Try saying:**
+> "We built an 'auto report generation' feature. Show this feature description to the personas and have them choose 'want to use / wouldn't use / not sure.' Ask for reasons too."
+
+### Document Difficulty Validation
+
+Check whether a document is at an appropriate level for the target audience.
+
+**Try saying:**
+> "Check with the personas whether this guide can be followed by a non-coder. Ask them to point out difficult words or sections that need more explanation."
+
+### Pricing Opinion Gathering
+
+Survey different customer types before deciding on pricing.
+
+**Try saying:**
+> "We have three pricing options: $15/mo, $25/mo, $40/mo. Ask each persona 'which price feels right and why.'"
+
+### Marketing Copy Comparison
+
+Get votes from personas on which copy is most appealing.
+
+**Try saying:**
+> "Here are 3 homepage headline candidates. Have each persona pick the most appealing one and explain why."
+
+---
+
+## Why Copilot Is a Good Fit
+
+For persona surveys, we recommend the **Copilot (`cli/copilot`)** provider.
+
+| Reason | Description |
+|--------|------------|
+| **Natural Conversation** | Strong at natural dialogue and roleplay, bringing each person's speaking style and personality to life |
+| **Cost-Effective** | Given the nature of surveys requiring multiple persona runs, the cost burden is low |
+| **Versatility** | Flexibly handles surveys on any topic — marketing, planning, development, etc. |
+
+> You can certainly run surveys with other providers (Claude, Gemini, etc.). But for persona simulation where "diverse perspective roleplay" matters, Copilot's natural conversation skills and cost efficiency make it a great match.
+
+---
+
+## Recommended Next Guides
+
+- **Building a Marketing Team** — Build a team that applies survey results to marketing
+- **Building Your C-Suite** — Connect persona validation to strategic decision-making
+- **Adding an Agent** — Detailed guide for the add agent screen
+
+---
+
+> **5-Minute Quickstart** — To get started right away, go to **Getting Started**.
+
+
+---
+
+# New Chat (Main Screen)
+
+> Spec ID: **THDUI000** | Path: `/`
+
+---
+
+## Screen Overview
+
+This is the first screen you see when opening the app. View your registered agent team at a glance and start conversations right away.
+
+---
+
+## Screen Layout
+
+| Area | Description |
+|------|------------|
+| **Sidebar** | Recent conversation list, tool menu (CLI Management, Usage, Settings) |
+| **Agent Grid** | Registered agent card list + add card |
+| **Suggested Task Cards** | Quick-access shortcuts for common tasks |
+| **Input Field** | Message input in `@agent task` format |
+
+
+---
+
+## Key Actions
+
+| Action | Result |
+|--------|--------|
+| Click agent card | `@AgentName ` auto-inserted in input field |
+| Click suggested card | `@Agent task description` auto-entered |
+| Send message | New conversation created → Navigate to chat |
+
+---
+
+## Related Use Cases
+
+- **Assigning Tasks to Agents**
+
+
+---
+
+# Chat
+
+> Spec ID: **THDUI002** | Path: `/threads/:threadId`
+
+---
+
+## Screen Overview
+
+The core screen for exchanging messages with agents. View messages, activity logs, and Sub-Agent execution status all in one place.
+
+---
+
+## Screen Layout
+
+| Area | Description |
+|------|------------|
+| **Header** | Conversation title, participant avatars, execution status badge |
+| **Message List** | User/agent message bubbles (markdown supported) |
+| **Activity Log** | Agent's work history (tool calls, file modifications, etc.) |
+| **Sub-Agent Reply** | Sub-agent execution status and results |
+| **Context Bar** | Conversation context usage visualization (shown when entering warm zone) |
+| **Input Field** | Send follow-up messages |
+
+
+---
+
+## Key Actions
+
+| Action | Result |
+|--------|--------|
+| Send message | Send additional instructions to agent |
+| Collapse/expand activity log | View detailed work history |
+| Click Sub-Agent | View sub-agent detailed logs |
+
+---
+
+## Related Use Cases
+
+- **Assigning Tasks to Agents**
+- **Checking Activity Logs**
+- **Resuming a Conversation**
+
+
+---
+
+# Thread List
+
+> Spec ID: **THDUI001** | Path: `/threads`
+
+---
+
+## Screen Overview
+
+A screen for browsing, searching, and filtering all conversations (threads).
+
+---
+
+## Screen Layout
+
+| Area | Description |
+|------|------------|
+| **Search Bar** | Search by conversation title/content |
+| **Filters** | Filter by agent, by date |
+| **Conversation Card List** | Conversation title, participating agents, last activity time |
+
+
+---
+
+## Related Use Cases
+
+- **Resuming a Conversation**
+
+
+---
+
+# Agent Management
+
+> Spec ID: **AGTUI001** | Path: `/agents`
+
+---
+
+## Screen Overview
+
+A screen for viewing registered agents and adding/editing/deleting them.
+
+---
+
+## Screen Layout
+
+| Area | Description |
+|------|------------|
+| **Header** | Agent count display + [Add Agent] button |
+| **Agent Card List** | Avatar, name (@ID), role, provider, model, [Edit]/[Delete] |
+
+
+---
+
+## Related Use Cases
+
+- **Adding an Agent**
+- **Editing & Deleting Agents**
+
+
+---
+
+# Add/Edit Agent
+
+> Spec ID: **AGTUI002** | Path: `/agents/new`, `/agents/:id/edit`
+
+---
+
+## Screen Overview
+
+A full-screen form for creating a new agent or modifying an existing agent's settings.
+
+---
+
+## Input Fields
+
+| Field | Required | Description |
+|-------|----------|------------|
+| **ID** | O | Unique identifier (alphanumeric, editable only when adding) |
+| **Name** | O | Display name |
+| **Role** | O | One-line role description |
+| **Provider** | O | AI engine selection (`cli/claude`, `cli/gemini`, etc.) |
+| **Model** | - | AI model to use |
+
+
+---
+
+## Related Use Cases
+
+- **Adding an Agent**
+- **Editing & Deleting Agents**
+
+
+---
+
+# CLI Management
+
+> Spec ID: **CLIVUI001** | Path: `/cli`
+
+---
+
+## Screen Overview
+
+A screen for managing the installation, updates, and status of AI coding tools (Claude Code, Gemini CLI, etc.).
+
+---
+
+## Screen Layout
+
+| Area | Description |
+|------|------------|
+| **Tool Card List** | Installation status, version, and login status for each AI tool |
+| **Action Buttons** | [Install], [Update], [Check Status] |
+
+
+---
+
+## Related Use Cases
+
+- **Installing & Updating CLI Tools**
+- **Setting Up AI Tools**
+
+
+---
+
+# Usage Dashboard
+
+> Spec ID: **USGUI001** | Path: `/usage`
+
+---
+
+## Screen Overview
+
+A dashboard screen for monitoring AI usage and costs.
+
+---
+
+## Screen Layout
+
+| Area | Description |
+|------|------------|
+| **Daily Chart** | Usage trend graph by date |
+| **Monthly Summary** | Monthly total usage and costs |
+| **Provider Tabs** | Filter by Claude, Gemini, etc. |
+
+
+---
+
+## Related Use Cases
+
+- **Checking Usage**
+
+
+---
+
+# Settings
+
+> Spec ID: **STGUI001** | Path: `/settings`
+
+---
+
+## Screen Overview
+
+A screen for configuring the app's theme, language, API keys, and more.
+
+---
+
+## Settings Items
+
+| Item | Description |
+|------|------------|
+| **Theme** | Light/dark mode |
+| **Language** | Korean/English |
+| **API Keys** | Manage API keys per provider |
+
+
+---
+
+## Related Use Cases
+
+- **Configuring Settings**
+
+
+---
+
+# Document Management
+
+> Spec ID: **DOCUI001** | Path: `/docs`
+
+---
+
+## Screen Overview
+
+A screen for browsing project markdown documents in a tree structure and viewing their contents.
+
+---
+
+## Screen Layout
+
+| Area | Description |
+|------|------------|
+| **Document Tree** | Folder/file tree structure |
+| **Document Viewer** | Markdown rendering of selected document |
+| **Table of Contents (TOC)** | Heading-based navigation within documents |
+
+
+<!-- Document management screenshot to be added separately -->
+
+
+---
+
+# Screen-by-Screen Guide
+
+> Describes the components and features of each screen. Referenced by spec ID.
+
+---
+
+## Main Layout
+
+| Document | Screen ID | Path | Status |
+|----------|-----------|------|--------|
+| **Main Layout** | NAVUI001 | All | Implemented |
+
+---
+
+## Conversations (Thread)
+
+| Document | Screen ID | Path | Status |
+|----------|-----------|------|--------|
+| **New Chat** | THDUI000 | `/` | Implemented |
+| **Thread List** | THDUI001 | `/threads` | Implemented |
+| **Chat** | THDUI002 | `/threads/:threadId` | Implemented |
+
+---
+
+## Agents
+
+| Document | Screen ID | Path | Status |
+|----------|-----------|------|--------|
+| **Agent Management** | AGTUI001 | `/agents` | Implemented |
+| **Add/Edit Agent** | AGTUI002 | `/agents/new`, `/agents/:id/edit` | Implemented |
+
+---
+
+## Tool Management
+
+| Document | Screen ID | Path | Status |
+|----------|-----------|------|--------|
+| **CLI Management** | CLIVUI001 | `/cli` | Implemented |
+| **Usage Dashboard** | USGUI001 | `/usage` | Implemented |
+
+---
+
+## Settings
+
+| Document | Screen ID | Path | Status |
+|----------|-----------|------|--------|
+| **Settings** | STGUI001 | `/settings` | Implemented |
+
+---
+
+## Document Management
+
+| Document | Screen ID | Path | Status |
+|----------|-----------|------|--------|
+| **Document Management** | DOCUI001 | `/docs` | Implemented |
+
+---
+
+## Unimplemented Screens (Planned)
+
+| Screen ID | Screen Name | Path | Notes |
+|-----------|------------|------|-------|
+| ONBUI001 | Onboarding Wizard | `/onboarding` | Design complete, not implemented |
+| PJTUI000~002 | Project Management | `/projects` | Planned for v1 |
+| DOCUI002 | Document Creation | `/docs/new` | Planned for v1 |
+| USGUI002 | Usage by Agent | `/usage` (tab) | Planned for v1 |
+| SRCHUI001 | Unified Search | `/search` | Implemented (docs not written) |
+| OVLUI001 | Overlay Chat | Electron | Planned for v1 |
+
+
+---
+
+# CrewX Developer Documentation
+
+Technical documentation for both those who **configure** agents via crewx.yaml and developers who **integrate** CrewX SDK into their services.
+
+---
+
+## Power Users — crewx.yaml Configuration
+
+A guide for those who work directly with crewx.yaml for agent configuration, prompt tuning, skill setup, and more.
+
+### Prompt Templates
+
+crewx.yaml prompts are not plain text. A built-in [Handlebars](https://handlebarsjs.com/) template engine allows you to inject dynamic data at agent execution time.
+
+```yaml
+prompt: |
+  너는 {{agent.name}}이야.
+
+  {{#if (eq mode "execute")}}
+  파일 수정이 가능합니다.
+  {{else}}
+  읽기 전용 모드입니다.
+  {{/if}}
+```
+
+In addition to variables and conditionals, you can also inject **command execution results** (memory, design doc structure, git status, etc.) into prompts.
+
+→ **Prompt Writing Guide** — Template syntax, exec helper, layout system, debugging
+
+### Guide List
+
+| Document | Description |
+|----------|------------|
+| **Prompt Writing Guide** | Prompt structure basics, template syntax, exec helper, layouts, debugging |
+| ↳ **Exec Helper Guide** | `{{{exec}}}` security model, anti-patterns, audit logs, troubleshooting |
+| **Remote Agent Testing Guide** | HTTP/file:// remote agent verification — unit tests, manual tests, security checklist |
+| **crewx.yaml Reference** | Full configuration file options reference |
+| Skill Usage Guide | Skill loading/execution/configuration (coming soon) |
+
+---
+
+## SDK Developers — Service Integration
+
+A reference for developers integrating agents running via crewx.yaml into their own services using the CrewX SDK (`@crewx/sdk`).
+
+| Document | Description |
+|----------|------------|
+| REST API | Thread, message, agent API endpoints (coming soon) |
+| MCP Endpoints | JSON-RPC 2.0-based MCP server API (coming soon) |
+| **CLI Command Reference** | Full command reference for `query`, `execute`, `agent`, etc. |
+| Skill Development Guide | Custom skill authoring (coming soon) |
+
+---
+
+> For detailed specifications, see the **Design Document**.