diff --git a/skills/zopia_ai_skills/API_REFERENCE.md b/skills/zopia_ai_skills/API_REFERENCE.md
new file mode 100644
index 0000000..dc66fc4
--- /dev/null
+++ b/skills/zopia_ai_skills/API_REFERENCE.md
@@ -0,0 +1,261 @@
+# API Reference
+
+Complete request/response schemas for all Zopia API endpoints.
+
+## POST /api/base/create
+
+Create a new project.
+
+**Request**:
+
+```json
+{
+  "baseName": "My Project",
+  "lang": "zh-CN"
+}
+```
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| `baseName` | string | No | Project name (auto-generated if omitted) |
+| `lang` | string | No | UI language: `zh-CN` or `en` (inferred from Accept-Language if omitted) |
+
+**Response** (`201`):
+
+```json
+{
+  "success": true,
+  "data": {
+    "baseId": "base_xxx",
+    "baseName": "My Project"
+  }
+}
+```
+
+---
+
+## POST /api/base/settings
+
+Save project settings. Performs a shallow merge with existing settings, allowing incremental updates.
+
+**Request**:
+
+```json
+{
+  "base_id": "base_xxx",
+  "settings": {
+    "locale": "zh-CN",
+    "aspect_ratio": "16:9",
+    "style": "anime_japanese_korean",
+    "generation_method": "n_grid"
+  }
+}
+```
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| `base_id` | string | Yes | Project ID |
+| `settings` | object | Yes | Settings key-value pairs |
+
+**Response**:
+
+```json
+{
+  "success": true,
+  "settings": {
+    "locale": "zh-CN",
+    "aspect_ratio": "16:9",
+    "style": "anime_japanese_korean"
+  }
+}
+```
+
+---
+
+## GET /api/base/settings
+
+Read current project settings.
+
+**Query Parameters**: `base_id` (required)
+
+**Response**:
+
+```json
+{
+  "success": true,
+  "settings": {
+    "locale": "zh-CN",
+    "aspect_ratio": "16:9",
+    "style": "anime_japanese_korean"
+  }
+}
+```
+
+---
+
+## POST /api/v1/agent/chat
+
+Chat with the project Agent. Returns the full result synchronously.
+
+**Request**:
+
+```json
+{
+  "base_id": "base_xxx",
+  "message": "Please generate a three-act screenplay on a campus theme",
+  "session_id": "session_xxx"
+}
+```
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| `base_id` | string | Yes | Project ID |
+| `message` | string | Yes | Natural language instruction |
+| `session_id` | string | No | Session ID for multi-turn conversation; omit to start new |
+
+**Language rule**: Agent reply language follows `settings.locale`.
+
+**Concurrency rule**: Only one request per `session_id` at a time. Returns `409` if a previous request is still running.
+
+**Prerequisites**: `locale`, `aspect_ratio`, and `style` must be saved in project settings. Returns `400` with `missing_fields` if incomplete.
+
+**Success Response**:
+
+```json
+{
+  "success": true,
+  "session_id": "session_xxx",
+  "base_id": "base_xxx",
+  "base_url": "https://zopia.ai/base/base_xxx",
+  "reply": {
+    "agent": "screenplay_writer",
+    "text": "Agent's text reply"
+  },
+  "actions": [
+    {
+      "tool_name": "write_screenplay",
+      "status": "success",
+      "result": {
+        "content": "...",
+        "ui_display_content": "..."
+      }
+    }
+  ],
+  "workspace": {
+    "files": [
+      {
+        "file_id": "file_xxx",
+        "name": "Screenplay",
+        "docu_type": "text",
+        "record_count": 0
+      }
+    ]
+  }
+}
+```
+
+**Response fields**:
+
+| Field | Description |
+|-------|-------------|
+| `reply.agent` | The Agent role that responded |
+| `reply.text` | Agent's full text reply |
+| `actions` | Tools invoked by the Agent and their results |
+| `workspace.files` | Current file list with record counts per table |
+
+**Error Response (settings incomplete)**:
+
+```json
+{
+  "success": false,
+  "error": "basic_settings_incomplete",
+  "missing_fields": ["locale", "style"],
+  "reference": {
+    "locale": "Dialogue language, e.g.: en, zh-CN, ja",
+    "aspect_ratio": "Aspect ratio: 16:9 | 9:16",
+    "style": "Style ID: anime_japanese_korean | realistic_3d_cg | ..."
+  }
+}
+```
+
+---
+
+## GET /api/base/list
+
+List all projects for the current user, ordered by update time (newest first).
+
+**Response**:
+
+```json
+{
+  "success": true,
+  "data": [
+    {
+      "id": "base_xxx",
+      "name": "My Project",
+      "thumbnails": ["https://..."],
+      "createdAt": "2025-01-01T00:00:00.000Z",
+      "updatedAt": "2025-01-02T00:00:00.000Z"
+    }
+  ]
+}
+```
+
+---
+
+## GET /api/base/{id}
+
+Get project details including session list and settings.
+
+**Response**:
+
+```json
+{
+  "name": "My Project",
+  "user_id": 123,
+  "sessions": [
+    {
+      "id": "session_xxx",
+      "title": "Campus Screenplay",
+      "created_at": "...",
+      "updated_at": "..."
+    }
+  ],
+  "settings": {
+    "locale": "zh-CN",
+    "aspect_ratio": "16:9",
+    "style": "anime_japanese_korean"
+  },
+  "profile": {
+    "username": "user1",
+    "image_url": "https://..."
+  }
+}
+```
+
+---
+
+## GET /api/billing/getBalance
+
+Query credit balance.
+
+**Response**:
+
+```json
+{
+  "accounts": [
+    {
+      "id": 1,
+      "credit_type": "free",
+      "balance": "100.00",
+      "created_at": "...",
+      "expires_at": "..."
+    }
+  ],
+  "summary": {
+    "totalBalance": 100.0,
+    "totalHeld": 0.0,
+    "totalAvailable": 100.0
+  }
+}
+```
diff --git a/skills/zopia_ai_skills/EXAMPLES.md b/skills/zopia_ai_skills/EXAMPLES.md
new file mode 100644
index 0000000..77c798d
--- /dev/null
+++ b/skills/zopia_ai_skills/EXAMPLES.md
@@ -0,0 +1,93 @@
+# Conversation Examples
+
+Practical examples of multi-turn Agent conversations and video generation best practices.
+
+## Multi-turn Conversation: Screenplay -> Characters -> Storyboard
+
+### Round 1: Generate Screenplay
+
+Do not pass `session_id` on the first call. The response returns a `session_id` for subsequent calls.
+
+```bash
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","message":"Please generate a three-act screenplay on a campus youth theme"}'
+```
+
+### Round 2: Generate Character Designs
+
+Use the `session_id` returned from the previous response.
+
+```bash
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","session_id":"session_xxx","message":"Please generate detailed character designs for the main characters, including design illustrations"}'
+```
+
+### Round 3: Generate Storyboard
+
+```bash
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","session_id":"session_xxx","message":"Please generate a storyboard for Act 1, listing all shots with their framing and descriptions"}'
+```
+
+## Video Generation Best Practices
+
+**Video generation is expensive. Always confirm the number of shots with the user before generating. Never generate all shots at once.**
+
+### Recommended flow:
+
+1. After storyboarding completes, check `workspace.files[].record_count` for the total number of shots
+2. Ask the user how many shots to generate and which range
+3. Generate in batches of 3-5 shots
+
+### Batch Generation
+
+```bash
+# Generate first batch (shots 1-3)
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","session_id":"session_xxx","message":"Start generating videos for shot1 through shot3"}'
+
+# Continue with next batch (shots 4-6)
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","session_id":"session_xxx","message":"Continue generating shots 4 through 6"}'
+```
+
+### Regenerate a Specific Shot
+
+```bash
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","session_id":"session_xxx","message":"Please regenerate the video for shot 2, change the camera movement to a slow push-in"}'
+```
+
+## End-to-End Quick Start
+
+```bash
+# 1. Create project
+curl -X POST "https://zopia.ai/api/base/create" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"baseName":"Agent Demo","lang":"zh-CN"}'
+
+# 2. Save settings (locale, aspect_ratio, style are required)
+curl -X POST "https://zopia.ai/api/base/settings" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","settings":{"locale":"zh-CN","aspect_ratio":"16:9","style":"anime_japanese_korean","generation_method":"n_grid"}}'
+
+# 3. Chat with Agent
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","message":"Please generate a three-act campus-themed screenplay"}'
+```
diff --git a/skills/zopia_ai_skills/README.md b/skills/zopia_ai_skills/README.md
new file mode 100644
index 0000000..18f72a3
--- /dev/null
+++ b/skills/zopia_ai_skills/README.md
@@ -0,0 +1,255 @@
+# Zopia AI Skills
+
+> **AI-Driven Video Production Platform** -- From screenplay to finished video, powered by natural language conversations.
+
+[中文文档](./README_CN.md) | English
+
+---
+
+## What is Zopia?
+
+[Zopia](https://zopia.ai) is an AI-driven video production platform. Through its API, you can create projects, configure visual styles, and then converse with an intelligent Agent that **automatically** handles the full production pipeline:
+
+```
+Screenplay Writing -> Character Design -> Storyboard Illustration -> Video Generation
+```
+
+No manual intervention is needed between stages -- simply describe your creative vision in natural language, and the Agent orchestrates every step.
+
+## Key Features
+
+- **Natural Language Driven** -- Describe your story in plain language; the Agent translates it into a professional production pipeline
+- **Full Production Pipeline** -- Covers screenplay, character design, storyboard keyframes, and video clips end-to-end
+- **Multi-Turn Conversations** -- Iterate and refine through continuous dialogue with session persistence
+- **Multiple Visual Styles** -- 9 built-in styles from Japanese anime to photorealistic, pixel art to Pixar 3D
+- **Multiple Video Models** -- 9 video generation models including Kling O3, Vidu Q3 Pro, Seedance 1.5, and more
+- **Flexible Configuration** -- Aspect ratio, resolution, generation method, and style are all configurable per project
+- **Batch Video Generation** -- Generate video clips in controllable batches of 3-5 shots to manage costs
+- **Credit System** -- Query balance and track usage via API
+
+## Architecture
+
+```
+┌─────────────────────────────────────────────────────────┐
+│                      Zopia Platform                     │
+│                                                         │
+│  ┌─────────┐    ┌──────────────────────────────────┐    │
+│  │  Your   │───>│          Agent (AI Core)          │    │
+│  │  App    │<───│                                   │    │
+│  └─────────┘    │  ┌──────────────┐ ┌────────────┐ │    │
+│    REST API     │  │ Screenplay   │ │ Character  │ │    │
+│                 │  │ Writer       │ │ Designer   │ │    │
+│                 │  └──────────────┘ └────────────┘ │    │
+│                 │  ┌──────────────┐ ┌────────────┐ │    │
+│                 │  │ Storyboard   │ │   Video    │ │    │
+│                 │  │ Artist       │ │ Producer   │ │    │
+│                 │  └──────────────┘ └────────────┘ │    │
+│                 └──────────────────────────────────┘    │
+└─────────────────────────────────────────────────────────┘
+```
+
+The Agent internally dispatches four specialized tools:
+
+| Tool | Role |
+|------|------|
+| `screenplay_writer` | Generates structured screenplays with acts and scenes |
+| `character_designer` | Creates detailed character designs with visual illustrations |
+| `storyboard_artist` | Produces keyframe storyboards with shot-by-shot breakdowns |
+| `video_producer` | Generates video clips from storyboard frames |
+
+## Workflow
+
+```
+1. Create Project
+       │
+       v
+2. Save Settings (locale, aspect_ratio, style)
+       │
+       v
+3. Agent Chat (multi-turn)
+   ├── "Write a screenplay about..."  --> screenplay_writer
+   ├── "Design the characters..."     --> character_designer
+   ├── "Create storyboard for Act 1"  --> storyboard_artist
+   └── "Generate videos for shots 1-3"--> video_producer
+       │
+       v
+4. View Results on https://zopia.ai/base/{project_id}
+```
+
+## Quick Start
+
+### 1. Get Your API Token
+
+Visit [zopia.ai/settings/api-tokens](https://zopia.ai/settings/api-tokens), log in, and click **"Generate New Token"**.
+
+- Format: `zopia-xxxxxxxxxxxx`
+- Validity: 30 days
+
+### 2. Create a Project
+
+```bash
+curl -X POST "https://zopia.ai/api/base/create" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"baseName": "My First Video", "lang": "en"}'
+```
+
+### 3. Configure Settings
+
+```bash
+curl -X POST "https://zopia.ai/api/base/settings" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "base_id": "base_xxx",
+    "settings": {
+      "locale": "en",
+      "aspect_ratio": "16:9",
+      "style": "anime_japanese_korean",
+      "generation_method": "n_grid"
+    }
+  }'
+```
+
+### 4. Chat with the Agent
+
+```bash
+# Round 1: Generate screenplay (no session_id needed)
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id": "base_xxx", "message": "Generate a three-act screenplay about a campus love story"}'
+
+# Round 2: Design characters (use session_id from Round 1 response)
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id": "base_xxx", "session_id": "session_xxx", "message": "Design the main characters with illustrations"}'
+
+# Round 3: Create storyboard
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id": "base_xxx", "session_id": "session_xxx", "message": "Create a storyboard for Act 1 with all shots"}'
+
+# Round 4: Generate videos in batches
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id": "base_xxx", "session_id": "session_xxx", "message": "Generate videos for shots 1 through 3"}'
+```
+
+## API Reference
+
+### Endpoints
+
+| Method | Endpoint | Description |
+|--------|----------|-------------|
+| `POST` | `/api/base/create` | Create a new project |
+| `POST` | `/api/base/settings` | Save project settings |
+| `GET`  | `/api/base/settings?base_id=xxx` | Read current settings |
+| `POST` | `/api/v1/agent/chat` | Chat with the AI Agent |
+| `GET`  | `/api/base/list` | List all projects |
+| `GET`  | `/api/base/{id}` | Get project details |
+| `GET`  | `/api/billing/getBalance` | Query credit balance |
+
+Base URL: `https://zopia.ai`
+
+All requests require the header:
+```
+Authorization: Bearer <TOKEN>
+```
+
+For complete request/response schemas, see [API_REFERENCE.md](./API_REFERENCE.md).
+
+## Configuration Options
+
+### Visual Styles
+
+| Style ID | Description |
+|----------|-------------|
+| `anime_japanese_korean` | Japanese anime style |
+| `realistic_3d_cg` | High-detail 3D CG realistic |
+| `pixar_3d_cartoon` | Pixar 3D cartoon |
+| `photorealistic_real_human` | Photorealistic real human |
+| `3D_CG_Animation` | 3D CG animation / Chinese style |
+| `anime_chibi` | Chibi / cute style |
+| `anime_shinkai` | Makoto Shinkai style |
+| `anime_ghibli` | Studio Ghibli style |
+| `stylized_pixel` | Pixel art |
+
+### Video Models
+
+| Model | ID | Recommended Methods |
+|-------|----|---------------------|
+| Kling O3 *(recommended)* | `generate_video_by_kling_o3` | `n_grid`, `multi_ref`, `multi_ref_v2` |
+| Vidu Q3 Pro *(recommended)* | `generate_video_by_viduq3_pro` | `n_grid` |
+| Kling v3.0 | `generate_video_by_kling_v3_0` | `n_grid` |
+| Seedance 1.5 | `generate_video_by_seedance_15` | `start_frame` |
+| Vidu Q2 Pro | `generate_video_by_viduq2_pro` | `start_frame` |
+| Hailuo 02 | `generate_video_by_hailuo_02` | `start_frame` |
+| Kling v2.6 Pro | `generate_video_by_kling_v26_pro` | `start_frame` |
+| Wan 2.6 I2V | `generate_video_by_wan26_i2v` | `start_frame` |
+| Wan 2.6 Flash | `generate_video_by_wan26_i2v_flash` | `start_frame` |
+
+### Generation Methods
+
+| Method | Description |
+|--------|-------------|
+| `n_grid` *(recommended)* | Multi-frame grid -- AI auto-plans keyframe sequence with good visual continuity |
+| `multi_ref` | Multi-reference v1 -- uses multiple reference images (Kling O3 only) |
+| `multi_ref_v2` | Multi-reference v2 -- improved version (Kling O3 only) |
+| `start_frame` | Start-frame driven -- uses keyframe as the first frame for precise control |
+
+### Resolution Options
+
+| Setting | Options | Default |
+|---------|---------|---------|
+| Keyframe (`image_size`) | `1K`, `2K`, `4K` | `2K` |
+| Video (`video_resolution`) | `480p`, `720p`, `1080p` | `480p` |
+| Aspect Ratio (`aspect_ratio`) | `16:9`, `9:16` | -- |
+
+## Multi-Turn Session Management
+
+- **First call**: Omit `session_id`. The response returns a generated `session_id`.
+- **Subsequent calls**: Pass the `session_id` to continue the same conversation.
+- **New conversation**: Omit `session_id` again to start fresh.
+- **Concurrency**: Only one request per `session_id` at a time. Returns `409` if busy.
+
+## Best Practices
+
+1. **Always configure settings first** -- `locale`, `aspect_ratio`, and `style` are required before chatting with the Agent
+2. **Confirm shot count before generating videos** -- Video generation is resource-intensive; check `workspace.files[].record_count` after storyboarding
+3. **Generate videos in batches** -- Use batches of 3-5 shots to manage costs and allow review between batches
+4. **Use `n_grid` generation method** -- Provides the best visual continuity across shots
+5. **Wait for responses** -- Image and video generation can be slow; always wait for completion before the next request
+
+## Error Codes
+
+| Code | Meaning |
+|------|---------|
+| `400` | Bad parameters or missing required settings |
+| `401` | Unauthenticated or token expired |
+| `402` | Insufficient credits |
+| `403` | No permission (not project owner) |
+| `404` | Resource not found |
+| `409` | Session busy -- another request is still running |
+
+## Documentation
+
+| File | Description |
+|------|-------------|
+| [SKILL.md](./SKILL.md) | Skill description and overview |
+| [API_REFERENCE.md](./API_REFERENCE.md) | Complete API request/response schemas |
+| [SETTINGS_REFERENCE.md](./SETTINGS_REFERENCE.md) | All valid configuration values |
+| [EXAMPLES.md](./EXAMPLES.md) | Conversation examples and best practices |
+
+## Links
+
+- **Platform**: [https://zopia.ai](https://zopia.ai)
+- **API Tokens**: [https://zopia.ai/settings/api-tokens](https://zopia.ai/settings/api-tokens)
+- **Documentation**: [Zopia AI Skills Guide](https://my.feishu.cn/docx/A8ludb4npoBVfVxha1ZcVd64nLc)
+
+## License
+
+See the [LICENSE](../LICENSE) file for details.
diff --git a/skills/zopia_ai_skills/README_CN.md b/skills/zopia_ai_skills/README_CN.md
new file mode 100644
index 0000000..d4e7fd7
--- /dev/null
+++ b/skills/zopia_ai_skills/README_CN.md
@@ -0,0 +1,255 @@
+# Zopia AI Skills
+
+> **AI 驱动的视频制作平台** -- 从剧本到成片，用自然语言对话完成全流程制作。
+
+中文 | [English](./README.md)
+
+---
+
+## Zopia 是什么？
+
+[Zopia](https://zopia.ai) 是一个 AI 驱动的视频制作平台。通过 API，你可以创建项目、配置视觉风格，然后与智能 Agent 进行对话，Agent 会**自动**完成完整的制作流程：
+
+```
+剧本创作 -> 角色设计 -> 分镜绘制 -> 视频生成
+```
+
+各阶段之间无需人工干预 -- 只需用自然语言描述你的创意构想，Agent 会自动调度每一个环节。
+
+## 核心特性
+
+- **自然语言驱动** -- 用日常语言描述故事，Agent 自动将其转化为专业制作流程
+- **全流程覆盖** -- 涵盖剧本、角色设计、分镜关键帧、视频片段的端到端制作
+- **多轮对话** -- 通过持续对话进行迭代和优化，支持会话持久化
+- **多种视觉风格** -- 内置 9 种风格，从日系动漫到写实风格、像素风到皮克斯 3D
+- **多种视频模型** -- 支持 9 种视频生成模型，包括 Kling O3、Vidu Q3 Pro、Seedance 1.5 等
+- **灵活配置** -- 画面比例、分辨率、生成方式、视觉风格均可按项目配置
+- **批量视频生成** -- 可控的分批生成（每批 3-5 个镜头），便于控制成本
+- **积分系统** -- 通过 API 查询余额和跟踪用量
+
+## 架构概览
+
+```
+┌─────────────────────────────────────────────────────────┐
+│                      Zopia 平台                          │
+│                                                         │
+│  ┌─────────┐    ┌──────────────────────────────────┐    │
+│  │  你的   │───>│          Agent（AI 核心）          │    │
+│  │  应用   │<───│                                   │    │
+│  └─────────┘    │  ┌──────────────┐ ┌────────────┐ │    │
+│    REST API     │  │   剧本创作    │ │  角色设计   │ │    │
+│                 │  │   Writer     │ │  Designer  │ │    │
+│                 │  └──────────────┘ └────────────┘ │    │
+│                 │  ┌──────────────┐ ┌────────────┐ │    │
+│                 │  │   分镜绘制    │ │  视频生成   │ │    │
+│                 │  │   Artist     │ │  Producer  │ │    │
+│                 │  └──────────────┘ └────────────┘ │    │
+│                 └──────────────────────────────────┘    │
+└─────────────────────────────────────────────────────────┘
+```
+
+Agent 内部调度四个专业工具：
+
+| 工具 | 职能 |
+|------|------|
+| `screenplay_writer` | 生成包含幕和场景的结构化剧本 |
+| `character_designer` | 创建详细的角色设计，附带视觉设计图 |
+| `storyboard_artist` | 制作分镜关键帧，逐镜头分解 |
+| `video_producer` | 根据分镜画面生成视频片段 |
+
+## 工作流程
+
+```
+1. 创建项目
+       │
+       v
+2. 保存设置（locale、aspect_ratio、style）
+       │
+       v
+3. Agent 对话（多轮）
+   ├── "写一个关于...的剧本"        --> screenplay_writer
+   ├── "设计主要角色..."            --> character_designer
+   ├── "为第一幕创建分镜"           --> storyboard_artist
+   └── "生成第 1-3 个镜头的视频"    --> video_producer
+       │
+       v
+4. 在 https://zopia.ai/base/{project_id} 查看成果
+```
+
+## 快速开始
+
+### 1. 获取 API Token
+
+访问 [zopia.ai/settings/api-tokens](https://zopia.ai/settings/api-tokens)，登录后点击 **"Generate New Token"**。
+
+- 格式：`zopia-xxxxxxxxxxxx`
+- 有效期：30 天
+
+### 2. 创建项目
+
+```bash
+curl -X POST "https://zopia.ai/api/base/create" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"baseName": "我的第一个视频", "lang": "zh-CN"}'
+```
+
+### 3. 配置项目设置
+
+```bash
+curl -X POST "https://zopia.ai/api/base/settings" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "base_id": "base_xxx",
+    "settings": {
+      "locale": "zh-CN",
+      "aspect_ratio": "16:9",
+      "style": "anime_japanese_korean",
+      "generation_method": "n_grid"
+    }
+  }'
+```
+
+### 4. 与 Agent 对话
+
+```bash
+# 第一轮：生成剧本（无需传 session_id）
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id": "base_xxx", "message": "请生成一个三幕式的校园青春主题剧本"}'
+
+# 第二轮：设计角色（使用第一轮返回的 session_id）
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id": "base_xxx", "session_id": "session_xxx", "message": "请为主要角色生成详细的角色设计，包含设计图"}'
+
+# 第三轮：绘制分镜
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id": "base_xxx", "session_id": "session_xxx", "message": "请为第一幕生成分镜，列出所有镜头的景别和描述"}'
+
+# 第四轮：分批生成视频
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id": "base_xxx", "session_id": "session_xxx", "message": "开始生成第 1 到第 3 个镜头的视频"}'
+```
+
+## API 接口一览
+
+### 接口列表
+
+| 方法 | 路径 | 说明 |
+|------|------|------|
+| `POST` | `/api/base/create` | 创建新项目 |
+| `POST` | `/api/base/settings` | 保存项目设置 |
+| `GET`  | `/api/base/settings?base_id=xxx` | 读取当前设置 |
+| `POST` | `/api/v1/agent/chat` | 与 AI Agent 对话 |
+| `GET`  | `/api/base/list` | 列出所有项目 |
+| `GET`  | `/api/base/{id}` | 获取项目详情 |
+| `GET`  | `/api/billing/getBalance` | 查询积分余额 |
+
+基础地址：`https://zopia.ai`
+
+所有请求需要在 Header 中携带：
+```
+Authorization: Bearer <TOKEN>
+```
+
+完整的请求/响应格式请参阅 [API_REFERENCE.md](./API_REFERENCE.md)。
+
+## 配置选项
+
+### 视觉风格
+
+| 风格 ID | 说明 |
+|---------|------|
+| `anime_japanese_korean` | 日系动漫风格 |
+| `realistic_3d_cg` | 高精度 3D CG 写实 |
+| `pixar_3d_cartoon` | 皮克斯 3D 卡通 |
+| `photorealistic_real_human` | 照片级写实真人 |
+| `3D_CG_Animation` | 3D CG 动画 / 国风 |
+| `anime_chibi` | Q版 / 萌系风格 |
+| `anime_shinkai` | 新海诚风格 |
+| `anime_ghibli` | 吉卜力风格 |
+| `stylized_pixel` | 像素风 |
+
+### 视频模型
+
+| 模型 | ID | 推荐生成方式 |
+|------|----|-------------|
+| Kling O3 *（推荐）* | `generate_video_by_kling_o3` | `n_grid`、`multi_ref`、`multi_ref_v2` |
+| Vidu Q3 Pro *（推荐）* | `generate_video_by_viduq3_pro` | `n_grid` |
+| Kling v3.0 | `generate_video_by_kling_v3_0` | `n_grid` |
+| Seedance 1.5 | `generate_video_by_seedance_15` | `start_frame` |
+| Vidu Q2 Pro | `generate_video_by_viduq2_pro` | `start_frame` |
+| Hailuo 02 | `generate_video_by_hailuo_02` | `start_frame` |
+| Kling v2.6 Pro | `generate_video_by_kling_v26_pro` | `start_frame` |
+| Wan 2.6 I2V | `generate_video_by_wan26_i2v` | `start_frame` |
+| Wan 2.6 Flash | `generate_video_by_wan26_i2v_flash` | `start_frame` |
+
+### 生成方式
+
+| 方式 | 说明 |
+|------|------|
+| `n_grid` *（推荐）* | 多帧网格 -- AI 自动规划关键帧序列，视觉连续性好 |
+| `multi_ref` | 多参考 v1 -- 使用多张参考图驱动视频（仅限 Kling O3） |
+| `multi_ref_v2` | 多参考 v2 -- 改进版本（仅限 Kling O3） |
+| `start_frame` | 首帧驱动 -- 使用关键帧作为起始帧，精准控制第一帧 |
+
+### 分辨率选项
+
+| 设置 | 可选值 | 默认值 |
+|------|--------|--------|
+| 关键帧（`image_size`） | `1K`、`2K`、`4K` | `2K` |
+| 视频（`video_resolution`） | `480p`、`720p`、`1080p` | `480p` |
+| 画面比例（`aspect_ratio`） | `16:9`（横屏）、`9:16`（竖屏） | -- |
+
+## 多轮会话管理
+
+- **首次调用**：不传 `session_id`，响应会返回一个生成的 `session_id`
+- **后续调用**：传入 `session_id` 继续同一会话
+- **新建会话**：再次不传 `session_id` 即可开启新对话
+- **并发限制**：每个 `session_id` 同一时间只允许一个请求，忙碌时返回 `409`
+
+## 最佳实践
+
+1. **务必先配置设置** -- `locale`、`aspect_ratio` 和 `style` 是与 Agent 对话前的必填项
+2. **生成视频前确认镜头数** -- 视频生成消耗较多资源，在分镜完成后检查 `workspace.files[].record_count` 获取总镜头数
+3. **分批生成视频** -- 每批 3-5 个镜头，便于控制成本和在批次间进行审查
+4. **优先使用 `n_grid` 生成方式** -- 提供最佳的镜头间视觉连续性
+5. **等待响应完成** -- 图片和视频生成可能较慢，务必等待完成后再发送下一个请求
+
+## 错误码
+
+| 状态码 | 含义 |
+|--------|------|
+| `400` | 参数错误或缺少必填设置 |
+| `401` | 未认证或 Token 已过期 |
+| `402` | 积分不足 |
+| `403` | 无权限（非项目所有者） |
+| `404` | 资源未找到 |
+| `409` | 会话忙碌 -- 上一个请求仍在处理中 |
+
+## 文档索引
+
+| 文件 | 说明 |
+|------|------|
+| [SKILL.md](./SKILL.md) | Skill 描述与概览 |
+| [API_REFERENCE.md](./API_REFERENCE.md) | 完整的 API 请求/响应格式 |
+| [SETTINGS_REFERENCE.md](./SETTINGS_REFERENCE.md) | 所有有效的配置值 |
+| [EXAMPLES.md](./EXAMPLES.md) | 对话示例与最佳实践 |
+
+## 相关链接
+
+- **平台地址**：[https://zopia.ai](https://zopia.ai)
+- **API Token 管理**：[https://zopia.ai/settings/api-tokens](https://zopia.ai/settings/api-tokens)
+- **说明文档**：[Zopia AI Skills 使用指南](https://my.feishu.cn/docx/A8ludb4npoBVfVxha1ZcVd64nLc)
+
+## 许可证
+
+详见 [LICENSE](../LICENSE) 文件。
diff --git a/skills/zopia_ai_skills/SETTINGS_REFERENCE.md b/skills/zopia_ai_skills/SETTINGS_REFERENCE.md
new file mode 100644
index 0000000..b09c372
--- /dev/null
+++ b/skills/zopia_ai_skills/SETTINGS_REFERENCE.md
@@ -0,0 +1,59 @@
+# Settings Reference
+
+All valid values for Zopia project settings fields.
+
+## Required Fields (must be set before Agent chat)
+
+| Field | Description | Valid Values |
+|-------|-------------|--------------|
+| `locale` | Dialogue / output language | Any ISO language code. Common: `zh-CN`, `en`, `ja` |
+| `aspect_ratio` | Frame aspect ratio | `16:9` (landscape), `9:16` (portrait) |
+| `style` | Visual style | See Style List below |
+
+## Optional Fields
+
+| Field | Description | Valid Values | Default |
+|-------|-------------|--------------|---------|
+| `video_model` | Video generation model | See Video Model List below | `generate_video_by_kling_o3` |
+| `generation_method` | Generation workflow | Depends on model, see below | `n_grid` |
+| `image_size` | Keyframe resolution | `1K`, `2K`, `4K` | `2K` |
+| `video_resolution` | Video resolution | `480p`, `720p`, `1080p` | `480p` |
+
+## Style List (`style` values)
+
+| ID | Description |
+|----|-------------|
+| `anime_japanese_korean` | Japanese anime style |
+| `realistic_3d_cg` | High-detail 3D CG realistic |
+| `pixar_3d_cartoon` | Pixar 3D cartoon |
+| `photorealistic_real_human` | Photorealistic real human |
+| `3D_CG_Animation` | 3D CG animation / Chinese style |
+| `anime_chibi` | Chibi / cute style |
+| `anime_shinkai` | Makoto Shinkai style |
+| `anime_ghibli` | Studio Ghibli style |
+| `stylized_pixel` | Pixel art |
+
+## Video Model List (`video_model` values)
+
+| `video_model` Value | Model Name | Supported `generation_method` |
+|---------------------|-----------|-------------------------------|
+| `generate_video_by_kling_o3` ⭐ | Kling O3 | `n_grid`, `multi_ref`, `multi_ref_v2` |
+| `generate_video_by_viduq3_pro` ⭐ | Vidu Q3 Pro | `n_grid` |
+| `generate_video_by_kling_v3_0` | Kling v3.0 | `n_grid` |
+| `generate_video_by_seedance_15` | Seedance 1.5 | `start_frame` |
+| `generate_video_by_viduq2_pro` | Vidu Q2 Pro | `start_frame` |
+| `generate_video_by_hailuo_02` | Hailuo 02 | `start_frame` |
+| `generate_video_by_kling_v26_pro` | Kling v2.6 Pro | `start_frame` |
+| `generate_video_by_wan26_i2v` | Wan 2.6 I2V | `start_frame` |
+| `generate_video_by_wan26_i2v_flash` | Wan 2.6 Flash | `start_frame` |
+
+⭐ = Recommended models.
+
+## Generation Method (`generation_method` values)
+
+| Value | Description | Use Case |
+|-------|-------------|----------|
+| `n_grid` | Multi-frame grid: AI auto-plans keyframe sequence, generates multiple shots at once | Recommended, good visual continuity |
+| `multi_ref` | Multi-reference v1: uses multiple reference images to drive video | Kling O3 only, multi-subject scenes |
+| `multi_ref_v2` | Multi-reference v2: improved version | Kling O3 only, prefer over v1 |
+| `start_frame` | Start-frame driven: uses keyframe image as starting frame | Legacy model compatibility, precise first-frame control |
diff --git a/skills/zopia_ai_skills/SKILL.md b/skills/zopia_ai_skills/SKILL.md
new file mode 100644
index 0000000..74a2c1f
--- /dev/null
+++ b/skills/zopia_ai_skills/SKILL.md
@@ -0,0 +1,143 @@
+---
+name: zopia-api
+description: "Drive AI video production via the Zopia API: create projects, configure styles, chat with the Agent to generate screenplays/characters/storyboards/videos, and query project status and credits. Use when the user wants to create AI video projects or interact with the Zopia platform through its API."
+---
+
+# Zopia External Agent Skill
+
+Zopia is an AI-driven video production platform. Through its API you can create projects, configure settings, then converse with the platform's built-in Agent, which automatically handles screenplay writing, character design, storyboard illustration, and video generation.
+
+## Authentication
+
+Guide the user to visit `https://zopia.ai/settings/api-tokens`, log in, and click "Generate New Token".
+
+- Format: `zopia-xxxxxxxxxxxx`
+- Validity: 30 days
+
+Include in every request header:
+
+```
+Authorization: Bearer <TOKEN>
+```
+
+## Recommended Workflow
+
+```
+Create Project -> Save Settings (locale/style/aspect_ratio/generation_method) -> Multi-turn Agent Chat -> View Results
+```
+
+Agent chat is the core interaction. Describe your needs in natural language and the Agent will automatically dispatch tools to:
+- Write screenplays (`screenplay_writer`)
+- Design characters (`character_designer`)
+- Draw storyboard keyframes (`storyboard_artist`)
+- Generate video clips (`video_producer`)
+
+**Multi-turn conversations**: Do not pass `session_id` on the first call; the response will return a generated `session_id`. Pass that `session_id` on subsequent calls to continue within the same session. Omitting `session_id` starts a new conversation.
+
+## Core API Endpoints
+
+### 1. Create Project
+
+```
+POST https://zopia.ai/api/base/create
+```
+
+Body: `{ "baseName": "My Project", "lang": "zh-CN" }`
+
+Returns `baseId` used in all subsequent calls.
+
+### 2. Save Settings (required before Agent chat)
+
+```
+POST https://zopia.ai/api/base/settings
+```
+
+Body:
+
+```json
+{
+  "base_id": "base_xxx",
+  "settings": {
+    "locale": "zh-CN",
+    "aspect_ratio": "16:9",
+    "style": "anime_japanese_korean",
+    "generation_method": "n_grid"
+  }
+}
+```
+
+**Three fields are required** before Agent chat: `locale`, `aspect_ratio`, `style`.
+**Always pass `generation_method`** explicitly; server defaults to `n_grid` if omitted.
+
+### 3. Agent Chat
+
+```
+POST https://zopia.ai/api/v1/agent/chat
+```
+
+Body:
+
+```json
+{
+  "base_id": "base_xxx",
+  "message": "Please generate a three-act screenplay on a campus theme",
+  "session_id": "session_xxx"
+}
+```
+
+- `base_id`: required
+- `message`: required, natural language instruction
+- `session_id`: optional, for continuing a conversation
+
+**Concurrency**: Only one request per `session_id` at a time; returns `409` if another is still running.
+**Latency**: Image, storyboard, and video generation can be slow. Wait for each response before sending the next request.
+
+### 4. Other Endpoints
+
+- `GET /api/base/settings?base_id=base_xxx` - Read current settings
+- `GET /api/base/list` - List all projects (newest first)
+- `GET /api/base/{id}` - Project details including sessions and settings
+- `GET /api/billing/getBalance` - Query credit balance
+
+For full API details, request bodies, and response schemas, see [API_REFERENCE.md](API_REFERENCE.md).
+For all settings values (styles, video models, generation methods), see [SETTINGS_REFERENCE.md](SETTINGS_REFERENCE.md).
+For conversation examples and video generation best practices, see [EXAMPLES.md](EXAMPLES.md).
+
+## Video Generation Best Practices
+
+**Video generation is expensive. Always confirm the number of shots with the user before generating. Never generate all shots at once.**
+
+After storyboarding, use `workspace.files[].record_count` to get the shot count, ask the user which range to generate, then generate in batches of 3-5 shots.
+
+## Error Codes
+
+| Code | Meaning |
+|------|---------|
+| `400` | Bad parameters / missing settings |
+| `401` | Unauthenticated or token expired |
+| `402` | Insufficient credits |
+| `403` | No permission (not project owner) |
+| `404` | Resource not found |
+| `409` | Session already running, retry later |
+
+## Quick Start
+
+```bash
+# 1. Create project
+curl -X POST "https://zopia.ai/api/base/create" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"baseName":"Agent Demo","lang":"zh-CN"}'
+
+# 2. Save settings
+curl -X POST "https://zopia.ai/api/base/settings" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","settings":{"locale":"zh-CN","aspect_ratio":"16:9","style":"anime_japanese_korean","generation_method":"n_grid"}}'
+
+# 3. Start chatting
+curl -X POST "https://zopia.ai/api/v1/agent/chat" \
+  -H "Authorization: Bearer zopia-xxxx" \
+  -H "Content-Type: application/json" \
+  -d '{"base_id":"base_xxx","message":"Please generate a three-act campus-themed screenplay"}'
+```