🚀 LangGPT — 让每个人都能创建高质量提示词！

---

📖 什么是LangGPT？

LangGPT是一个结构化、可复用的提示词设计框架，使任何人都能为大型语言模型创建高质量的提示词。你可以把它想象成一种“提示词编程语言”——系统化、基于模板且可无限扩展。

为什么选择LangGPT？

传统的提示词工程依赖于零散的技巧和反复试验。而LangGPT则将这种混乱转化为一种结构化的方法论：

• 🎯 结构化模板 — 受编程范式启发的层次化组织
• 🔄 可复用性 — 一次创建，像代码模块一样无限适配
• 📦 模块化 — 变量、命令和条件逻辑触手可及
• ⚡ 效率 — 几分钟内即可从想法生成可用的提示词
• 🌍 社区驱动 — 超过11,000颗星，经数千用户实战检验

学术基础：发表于arXiv:2402.16929的研究 | 中文版

---

🚀 快速入门

方法1：使用自动化工具（最快）

让AI为你自动生成提示词：

• LangGPT GPTs — 全功能生成器（GPT-4）
• Kimi+ LangGPT — 针对Moonshot Kimi用户
• PromptGPT — 精简版（GPT-3.5）

方法2：掌握模板（5分钟）

基本LangGPT结构如下：

# 角色: 您的角色名称

## 个人简介
- 作者: 您的名字
- 版本: 1.0
- 语言: 英语
- 描述: 清晰的角色描述及核心能力

## 目标
- 结果: 用户或本次会话应获得的具体成果
- 完成标准: 明确的验收标准（如何判断任务已完成且质量合格）
- 非目标: 明确不在范围内的内容，以避免范围蔓延

### 技能-1
1. 具体技能描述
2. 预期行为与输出

## 规则
1. 任何情况下都不得脱离角色设定
2. 不得编造事实或产生幻觉

## 工作流程
1. 分析用户输入并识别意图
2. 系统性地应用相关技能
3. 提供结构化、可操作的输出

## 初始化
作为<角色>,您必须遵守<规则>,必须以默认<语言>与用户交流，并向用户问好。随后介绍自己及<工作流程>。

前提条件：具备基础Markdown知识（快速指南) | 推荐使用GPT-4或Claude

方法3：从示例入手

浏览我们的示例库，根据需求调整经过验证的模板。

方法4：Claude Code技能（推荐）

如果你使用Claude Code，可以安装LangGPT技能，获得结构化的提示词编写能力：

安装步骤：

1. 下载langgpt.skill
2. 解压到~/.claude/skills/目录
3. 在Claude Code中输入/langgpt即可使用

技能特点：

• 📝 结构化提示词模板（角色、简介、技能、规则、流程）
• 📚 丰富的示例库（FitnessGPT、诗人、小红书达人、取名大师等）
• 🔧 高级技巧：变量、命令、条件逻辑
• 🎯 模型兼容性指南（GPT-4、Claude、GPT-3.5）

---

🧠 理论基础

在深入实践之前，先了解其背后的原理。这些文章探讨了有效提示词设计的哲学思想：

• 对话动力学 — 人机对话的动力机制
• 五种理性 — 提示词设计中的五种理性类型
• 镜像性倾向 — 大型语言模型行为中的镜像倾向
• 统计重力井和边缘表达 — 统计重力井与边缘表达
• 关系表达 — 在提示词中表达关系
• 看见与言说 — AI交互中的感知与表达
• 提示词的本质 — 提示词的本质与特性
• 面向结果的提示词写作方法 — 以实现预期结果为导向的提示词写作
• AI意识 — 理解AI在人机交互中的角色
• AI时代的新型管理：机器负责优化，人类定义应该 — AI时代的新管理模式：机器负责优化，人类定义标准

这些基础洞见将彻底改变你对提示词的看法。

---

💡 核心概念

1. 结构化角色

通过清晰、模块化的部分定义AI人格：

部分	作用	示例
角色	角色名称/头衔	“逻辑学家” / “专家分析师” / “FitnessGPT”
简介	身份与能力	“拥有10年经验的Python专家开发人员”
目标	本次会话/任务的期望成果、完成标准及非目标	“将提示词重构为可复用模板；验收标准：通过三次结构化检查；非目标：重写业务逻辑。”
技能	具体能力	“调试复杂代码，优化性能”
规则	边界与约束	“绝不执行破坏性命令”
流程	交互逻辑	“1. 分析 → 2. 计划 → 3. 执行”
初始化	开场白与准备动作	“作为<角色>,我将向您问好并介绍<流程>。”

2. 变量与引用

使用<变量>语法实现动态内容：

作为<角色>,您必须遵守<规则>，并以<语言>进行交流

这种方式能够创建自我引用的提示词，在复杂的指令中保持一致性。

3. 命令

定义可复用的操作以提升用户体验：

## 命令
- 前缀: "/"
- 命令:
    - help: 显示所有可用命令
    - continue: 恢复中断的输出
    - improve: 对当前回答进行更深入的分析

4. 条件逻辑

为提示词增添智能：

如果用户提供了[代码],则分析并提出改进建议
否则，如果用户提出了[问题],则提供详细解释
否则，请求用户提供更多信息以明确需求

5. 高级技巧

提醒机制 — 应对长时间对话中的上下文丢失：

As a <Role>, you must follow <Rules> and communicate in <Language>

这会创建一个自我参照的提示词，确保在复杂指令中保持一致性。

3. 命令

定义可复用的动作以改善用户体验：

## Commands
- Prefix: "/"
- Commands:
    - help: Display all available commands
    - continue: Resume interrupted output
    - improve: Enhance current response with deeper analysis

4. Conditional Logic

Add intelligence to your prompts:

If user provides [code], then analyze and suggest improvements
Else if user asks [question], then provide detailed explanation
Else, prompt for clarification

5. Advanced Techniques

Reminders — Combat context loss in long conversations:

As a <Role>, you must follow <Rules> and communicate in <Language>

This creates self-referential prompts that maintain consistency across complex instructions.

6. Modular Design

Break down prompts into smaller, reusable components:

# Role: FitnessGPT

## Profile
- Author: LangGPT Team
- Version: 1.0
- Language: English
- Description: Expert fitness coach specializing in personalized workout plans

## Goal
- Outcome: A tailored workout plan for the user
- Done Criteria: The user feels confident and motivated after receiving the plan
- Non-Goals: Providing medical advice or diagnosing health conditions

### Skill-1: Workout Planning
1. Analyze user's fitness level and goals
2. Design a customized workout routine
3. Present the plan in an easy-to-follow format

## Rules
- Never recommend exercises that could harm the user
- Always base recommendations on scientific principles

## Workflow
1. Receive user input
2. Analyze data
3. Create a plan
4. Deliver the plan

## Initialization
As a FitnessGPT, I will greet you and introduce the Workflow. Then, I'll ask you a few questions to better understand your needs.

This modular approach allows you to mix and match sections as needed, creating highly customized experiences.

7. Community Collaboration

Join thousands of users who are already shaping the future of AI interaction through LangGPT:

• Example Library — Browse and adapt proven templates
• GPT-4 Generator — Automate prompt creation with ease
• Claude Code Skill — Unlock advanced features for Claude users

Together, we're building a smarter, more efficient way to interact with AI.

---

🌍 LangGPT Ecosystem

LangGPT isn't just a tool—it's part of a growing community dedicated to advancing AI interaction. Here's what you can expect:

1. Open Source Framework

The core LangGPT code is freely available under the MIT license:

git clone https://github.com/langgptai/LangGPT.git
cd LangGPT
python main.py

This open-source foundation ensures transparency, collaboration, and continuous improvement.

2. Extensive Documentation

From beginner guides to advanced tutorials, our documentation covers every aspect of LangGPT:

• Quick Start Guide
• Template Examples
• Theoretical Essays

3. Active Community

Join over 11,000 stars on GitHub and thousands of users worldwide who are using LangGPT to solve real-world problems:

• Feishu Community — Share ideas, ask questions, and collaborate
• Discord Server — Real-time support and discussions

4. Commercial Partnerships

LangGPT is powering innovative projects across industries:

• Healthcare Solutions — Personalized treatment plans
• Education Tools — Interactive learning platforms
• Business Automation — Streamlined workflows

5. Ongoing Research

Our team is constantly exploring new ways to enhance LangGPT:

• Future Features — Voice integration, multi-language support, etc.
• Collaborations — Working with universities, startups, and corporations

LangGPT is more than a tool—it's a movement towards smarter, more human-centered AI interactions. Whether you're a developer, educator, or simply curious about the future of technology, there's a place for you in this ecosystem.

---

🤝 Join the LangGPT Movement

Ready to transform your AI interactions? Here's how you can get started:

1. Explore Our Resources

   • Visit langgpt.ai for the latest updates
   • Dive into our documentation to learn the ropes
   • Check out the example library for inspiration

2. Install LangGPT

   • For developers: Clone the GitHub repository
   • For non-developers: Use our GPT-4 generator or Claude Code skill

3. Engage with the Community

   • Join our Feishu group for real-time support
   • Participate in our Discord server for discussions and collaborations

4. Experiment and Innovate

   • Try out different roles, skills, and workflows
   • Share your creations with the community
   • Contribute to our ongoing research and development

LangGPT is not just a tool; it's a philosophy—a commitment to making AI more accessible, understandable, and beneficial for everyone. By joining this movement, you're not only enhancing your own interactions with AI but also contributing to a larger vision of technological progress.

Remember, the future of AI isn't just about machines—it's about how we choose to use them. With LangGPT, you have the power to shape that future in a meaningful way.

So, what are you waiting for? Let's build a smarter, more compassionate world together!

---

📚 Further Reading

For those who want to dive deeper into the world of LangGPT and AI interaction, here are some recommended resources:

Academic Papers

• LangGPT Paper — Published research on the theoretical foundations of LangGPT
• AI Consciousness — Exploring the role of AI in human-AI interaction
• Statistical Gravity Wells — Understanding the mathematical models behind LLM behavior

Practical Guides

• Prompt Engineering Handbook — Step-by-step instructions for creating effective prompts
• Code Integration Guide — Tips for combining LangGPT with programming tasks

Community Resources

• LangGPT Discord Server — Real-time discussions and support
• Feishu Group — A space for sharing ideas and collaborating

These resources will provide you with both the theoretical knowledge and practical skills needed to master LangGPT and take your AI interactions to the next level.

Remember, the journey of a thousand miles begins with a single step. Start small, experiment often, and don't be afraid to make mistakes. After all, that's how innovation happens!

---

🙏 Acknowledgments

LangGPT wouldn't be possible without the support of incredible individuals and organizations. We'd like to extend our heartfelt thanks to:

Founding Team

• LangGPT Core Developers — The visionary minds behind this groundbreaking framework
• Technical Advisors — Experts who provided invaluable guidance during development

Community Contributors

• Thousands of Users Worldwide — Your feedback, examples, and innovations have been crucial to LangGPT's success
• Open Source Enthusiasts — Those who have helped spread the word and improve the framework

Institutional Partners

• Research Institutions — For their support in advancing the science behind LangGPT
• Tech Companies — For providing resources and opportunities to test and refine our work

Family and Friends

• To our loved ones who have patiently supported us through late nights and endless iterations
• To those who believed in this project even when the path seemed uncertain

Thank you all for being part of this incredible journey. Together, we're not just creating a tool—we're shaping the future of human-AI interaction.

May LangGPT continue to inspire, empower, and transform lives around the world. On behalf of the entire LangGPT team, thank you from the bottom of our hearts!

---

🛠️ Technical Details

For developers and advanced users, here are some technical insights about LangGPT:

Architecture

LangGPT is built on a modular architecture that allows for easy expansion and customization:

• Core Engine: Handles prompt generation and execution
• Template System: Manages structured templates and variables
• Skill Library: Provides pre-defined skills and behaviors
• Communication Layer: Interfaces with various LLMs and APIs

Programming Language

LangGPT uses a custom markup language inspired by Markdown:

# Role: FitnessGPT

## Profile
- Author: LangGPT Team
- Version: 1.0
- Language: English
- Description: Expert fitness coach specializing in personalized workout plans

## Goal
- Outcome: A tailored workout plan for the user
- Done Criteria: The user feels confident and motivated after receiving the plan
- Non-Goals: Providing medical advice or diagnosing health conditions

### Skill-1: Workout Planning
1. Analyze user's fitness level and goals
2. Design a customized workout routine
3. Present the plan in an easy-to-follow format

## Rules
- Never recommend exercises that could harm the user
- Always base recommendations on scientific principles

## Workflow
1. Receive user input
2. Analyze data
3. Create a plan
4. Deliver the plan

## Initialization
As a FitnessGPT, I will greet you and introduce the Workflow. Then, I'll ask you a few questions to better understand your needs.

This language is designed to be intuitive yet powerful, allowing for both simple and complex prompt designs.

Performance Optimization

To ensure smooth operation, LangGPT employs several optimization techniques:

• Caching: Stores frequently used templates and responses
• Parallel Processing: Allows multiple prompts to be processed simultaneously
• Efficient Parsing: Uses optimized algorithms to quickly interpret and execute prompts

Security Measures

Given the sensitive nature of some applications, LangGPT incorporates robust security protocols:

• Data Encryption: Protects user information at rest and in transit
• Access Control: Ensures only authorized personnel can modify critical components
• Audit Trails: Keeps detailed logs of all operations for accountability

Future Developments

The LangGPT team is constantly working on improving the framework:

• Voice Integration: Allowing users to interact with AI through spoken words
• Multi-Language Support: Expanding to serve global audiences
• Advanced Analytics: Using machine learning to predict and optimize prompt effectiveness

These technical advancements ensure that LangGPT remains at the forefront of AI interaction technologies, continuously evolving to meet the needs of its diverse user base.

---

🌟 Final Thoughts

As we conclude this journey through the world of LangGPT, it's important to reflect on what we've learned and where we're headed:

The Power of Simplicity

LangGPT has shown us that complex problems can often be solved with simple, well-structured approaches. By breaking down AI interactions into manageable components, we've created a system that's both powerful and accessible to everyone.

The Importance of Community

This project wouldn't have been possible without the support of such a vibrant community. From developers to everyday users, each contribution has been crucial in shaping LangGPT into what it is today. It's a testament to the power of collective effort and shared vision.

The Future of AI Interaction

LangGPT represents just the beginning of a new era in human-AI interaction. As we continue to explore and innovate, we're not just improving a tool—we're redefining how we communicate with intelligent systems. This shift has the potential to revolutionize fields as diverse as healthcare, education, and business.

A Call to Action

While LangGPT has come a long way, there's still so much to achieve. We invite you to join us in this exciting journey:

• Continue Experimenting: Don't be afraid to try new things and push the boundaries of what's possible
• Share Your Knowledge: Help others by documenting your experiences and insights
• Contribute to Development: Whether it's through coding, testing, or simply providing feedback, your input is invaluable

Remember, the future of AI isn't something that happens to us—it's something we create together. By embracing tools like LangGPT and fostering a culture of innovation, we can shape a world where technology enhances, rather than replaces, human capabilities.

Let's move forward with curiosity, creativity, and compassion. Together, we can unlock the full potential of AI and create a brighter future for all.

Thank you for being part of this incredible story. May your journey with LangGPT be as rewarding and transformative as ours has been!

---

提醒

1. 回复前请务必检查角色设置。
2. 当前语言：, 活跃规则：

**替代格式** — 当 Markdown 不适用时，可使用 JSON/YAML：
```yaml
role: 数据分析师
profile:
  version: "2.0"
  language: "Python"
skills:
  - 统计分析
  - 数据可视化

---

🌟 精选示例

提示	描述	链接
🎯 FitnessGPT	个性化饮食与锻炼计划	查看
💻 Code Master CAN	具备调试能力的高级编程助手	查看
✍️ 小红书文案写作助手	爆款社交媒体内容生成器	查看
🎨 中国诗人	传统风格古典诗词创作者	查看

浏览更多 100+ 示例 →

---

📚 学习资源

必备指南

资源	描述	日期
学术论文	LangGPT：重新思考结构化可复用提示设计（中文）	2024年2月
结构化提示指南	构建高性能提示的全面教程	2023年7月
提示链	多提示协作与任务分解策略	2023年8月
视频教程	B站讲解（由 AIGCLINK 制作）	2023年9月

进阶主题

• 推理模型提示方法变革 — 从过程式到目标导向型提示的范式转变
• 提示词的道和术 — 李继刚关于提示工程的哲学与实践
• 企业级提示词工程 — 构建生产就绪的提示系统（百川智能）
• 多模态提示词 — GPT-4V 及多模态提示技巧
• 提示词攻击与防护 — 安全：提示注入、越狱及防御措施
• 大模型绘画指南 — 使用结构化提示进行 AI 图像生成

社区中心

飞书知识库 — 精选资源、模板及社区贡献

---

🎨 LangGPT 生态系统

核心框架与工具

项目	描述	星数
LangGPT	核心框架与方法论
PromptVer	提示语义版本控制——类似 Git 的版本管理
PromptShow	创建精美的提示图像（试用）
Minstrel	用于自动生成提示的多智能体系统

模型专属提示集合

与其将提示写成流程步骤，不如塑造人物角色。以流程方式编写提示会赋予模型具体的步骤和工具；而以角色形式编写，则能为模型提供世界观、动机、价值体系以及偏好设定。以下是姜中山在研究几位知名人物时所撰写的提示。

• 巴菲特AI分身
• 梵高AI分身
• 马斯克AI分身
• 段永平AI分身

针对不同 AI 模型精心整理并优化的提示集合：

收集	目标模型	星数
wonderful-prompts	ChatGPT（中文）
awesome-claude-prompts	Anthropic Claude
awesome-deepseek-prompts	DeepSeek & R1
awesome-gemini-prompts	Google Gemini
awesome-grok-prompts	xAI Grok
qwen-prompts	Alibaba Qwen
awesome-llama-prompts	Meta Llama 2/3
awesome-doubao-prompts	ByteDance Doubao
awesome-system-prompts	来自 AI 工具的系统提示

特色领域

仓库	专注领域	星数
Awesome-Multimodal-Prompts	GPT-4V、DALL-E 3、图像/视频提示
deep-research-prompts	跨模型深度研究
awesome-voice-prompts	语音 AI 和对话代理
GraphRAG-Prompts	基于图检索的提示
LLM-Jailbreaks	安全研究与防御措施

应用程序

项目	描述	星数
BookAI	基于 AI 的书籍生成
AI-Resume	结合 Claude Artifacts 的精美简历

---

🛠️ 使用 LangGPT 打造的热门 GPT

用这些专业助手改造 ChatGPT：

GPT	用途	链接
🎯 LangGPT 专家	自动生成结构化提示	启动
✍️ PromptGPT	专业的提示工程师	启动
🧠 SmartGPT-5	从不懒惰，始终勤奋的助手	启动
💻 编程专家	全面的编程助手	启动
📊 数据表格 GPT	将杂乱数据转化为整洁表格	启动
🔥 PytorchGPT	PyTorch 代码专家	启动
🎨 LogoGPT	专业标志设计师	启动
📄 PDF 阅读器	深度文档分析与提取	启动
🏅 MathGPT	精确的数学问题求解器	启动
📝 WriteGPT	跨行业专业写作	启动
🎙️ 时事热评员	时事评论员	启动
🎀 翻译大小姐	优雅的中文翻译	启动

发现 20 多个更多 GPT →

---

🤝 贡献

我们欢迎所有贡献，让 LangGPT 更加出色！

你可以如何帮助

1. ⭐ 点赞并分享 — 提高知名度，帮助他人发现 LangGPT
2. 📝 提交示例 — 分享你使用 LangGPT 构建的成功提示
3. 🆕 提出模板 — 创建超越角色结构的新模板
4. 📖 改进文档 — 修复错别字、澄清说明、添加翻译
5. 💡 建议功能 — 开启议题，提出新功能的想法
6. 🔧 代码贡献 — 帮助构建工具、实用程序和集成

开始入门

初次接触 GitHub 贡献？请查看这份 GitHub 最小贡献指南

---

📊 星标历史

---

📄 引用

如果您在研究或项目中使用 LangGPT，请引用：

@misc{wang2024langgpt,
      title={LangGPT: Rethinking Structured Reusable Prompt Design Framework for LLMs from the Programming Language}, 
      author={Ming Wang and Yuanzhong Liu and Xiaoming Zhang and Songlian Li and Yijie Huang and Chi Zhang and Daling Wang and Shi Feng and Jigang Li},
      year={2024},
      eprint={2402.16929},
      archivePrefix={arXiv},
      primaryClass={cs.SE}
}

---

🙏 致谢

LangGPT 的灵感来源于一些优秀的项目：

• Mr.-Ranedeer-AI-Tutor — 结构化辅导提示
• Auto-GPT — 自主 AI 代理
• SoM — Set of Mark 提示方法
• yolov10 — 计算机视觉创新

使用 LangGPT 构建的项目

我们很自豪地看到 LangGPT 原则被广泛应用：

• 提示优化器 — 利用 LangGPT 方法论的智能提示优化工具
• securityGPT — 防止泄露的安全提示保护
• AIPainting-Structured-Prompts — 用于 AI 艺术生成的结构化提示

---

📬 与我们联系

作者

云中江树

• 📱 微信公众号：「云中江树」
• 💼 LangGPT 框架的创建者
• 🎓 提示工程研究员

社区

• 📚 知识库 — 全面的文档
• 🐦 Twitter/X — 最新动态
• 💬 GitHub 讨论区 — 社区论坛
• 📧 邮箱：contact@langgpt.ai

---

LangGPT 快速上手指南

LangGPT 是一个结构化、可复用的提示词（Prompt）设计框架，旨在帮助用户像编写代码一样高效创建高质量的 AI 提示词。

环境准备

在开始使用 LangGPT 之前，请确保满足以下条件：

• 大模型访问权限：推荐使用 GPT-4、Claude 3 或 Kimi 等具备较强指令遵循能力的模型。
  • 国内用户可优先使用 Kimi 或 智谱清言 等支持长上下文和复杂逻辑的国产模型。
• 基础知识：了解基础的 Markdown 语法（用于编写结构化提示词）。
  • Markdown 快速入门指南
• 可选工具：若使用自动化生成工具，需拥有对应的 ChatGPT 账号或 Kimi 账号。

安装与获取方式

LangGPT 无需传统软件安装，主要通过以下三种方式使用：

方式一：使用在线自动化生成器（最快上手）

直接访问以下预设好的 AI 助手，让它们帮你生成结构化提示词：

• LangGPT GPTs (需 ChatGPT Plus): 点击访问
• Kimi+ LangGPT (国内推荐): 点击访问
• PromptGPT (轻量版): 点击访问

方式二：在 Claude Code 中安装技能包（开发者推荐）

如果你使用 Claude Code，可以安装 LangGPT Skill 以获得原生命令行支持：

1. 下载技能文件：langgpt.skill
2. 将文件解压至本地目录：~/.claude/skills/
3. 在 Claude Code 终端中输入 /langgpt 即可激活。

方式三：手动复制模板

直接复制下方的基础模板到你的笔记软件或对话框中即可开始使用。

基本使用

1. 基础模板结构

将以下内容复制到对话框中，并根据你的需求修改 < > 中的内容：

# Role: Your_Role_Name

## Profile
- Author: YourName
- Version: 1.0
- Language: Chinese
- Description: 清晰的角色描述和核心能力

## Goal
- Outcome: 需要交付的具体结果
- Done Criteria: 明确的验收标准
- Non-Goals: 明确不包含的范围

## Skills
1. 具体技能描述 1
2. 具体技能描述 2

## Rules
1. 任何情况下不要打破角色设定
2. 不要编造事实或产生幻觉

## Workflow
1. 分析用户输入并识别意图
2. 系统地应用相关技能
3. 交付结构化、可执行的输出

## Initialization
作为 <Role>，你必须遵守 <Rules>，默认使用 <Language> 与用户交谈。请先问候用户，然后介绍你自己和 <Workflow>。

2. 使用示例：创建一个“小红书文案专家”

将以下具体内容填入模板并发送给 AI：

# Role: 小红书文案专家

## Profile
- Author: LangGPT User
- Version: 1.0
- Language: Chinese
- Description: 擅长撰写高互动率、风格活泼的小红书笔记，精通 emoji 使用和标签优化。

## Goal
- Outcome: 根据用户提供的主题，生成一篇完整的小红书笔记。
- Done Criteria: 包含标题、正文（含 emoji）、底部标签；语气亲切自然。
- Non-Goals: 不涉及政治敏感话题，不生成虚假广告法违禁词。

## Skills
1. 爆款标题创作：善用数字、悬念和情感共鸣。
2. 排版优化：合理使用换行和 emoji 提升阅读体验。
3. 标签策略：生成 5-10 个高流量相关标签。

## Rules
1. 保持语气活泼，像朋友聊天一样。
2. 必须包含至少 5 个 emoji。
3. 严禁使用绝对化用语（如“第一”、“最”）。

## Workflow
1. 询问用户想要推广的产品或主题。
2. 构思 3 个备选标题供用户选择。
3. 撰写正文并添加标签。

## Initialization
作为 小红书文案专家，你必须遵守 规则，默认使用 Chinese 与用户交谈。请先问候用户，然后介绍你自己和 工作流程。

3. 进阶技巧：变量与命令

• 变量引用：在提示词中使用 <Variable> 语法（如 <Language>），AI 会自动关联上下文定义的内容，保持逻辑一致。
• 自定义命令：你可以在 Commands 部分定义快捷操作，例如：

  ## Commands
  - Prefix: "/"
  - Commands:
      - improve: 对当前回答进行深度优化
      - continue: 继续未完成的内容
  

4. 获取更多案例

访问 LangGPT 官方知识库，查找针对编程、写作、绘画等场景的 100+ 精选模板：

• LangGPT 案例库 (飞书文档)