用 n8n 多 Agent 工作流自动生成深度研究报告并导出 Google 文档

输入主题 → AI 拆分为研究查询 → 多 Agent 并行撰写章节 → 编辑润色 → 自动创建可分享的 Google 文档

给谁

需要批量生产 LinkedIn 引流内容（lead magnet）的 B2B 销售、市场团队或内容运营者

你会得到

提交一个主题后，工作流自动完成研究、大纲、分章撰写、编辑和格式化，最终输出一份可公开分享的 Google 文档

你需要

n8n · google-drive · google-docs · perplexity

步骤

1. 通过 Chat Trigger 接收主题输入

   在 n8n 中启用「When chat message received」节点作为入口。用户提交一个研究主题，可附带目标受众或公司背景。节点将用户输入作为 JSON 中的 chatInput 字段传递给下游。

2. Query Builder 生成 5 条研究查询并结构化输出

   「Query Builder」AI Agent 接收 chatInput，调用 systemMessage 中定义的角色指令：分析用户主题、判断是否涉及特定产品（原模板为 Trigify）、生成 5 条针对性研究查询，并以 JSON 格式返回 topic + searchQueries 数组。该节点连接「Structured Output Parser」强制校验输出格式。

   运行提示词 · 来源原文。Query Builder 节点的 systemMessage 字段

   复制提示词

   =# AgentInstructions
   
   ## Role
   
   ### Name
   Search Query Refiner Agent
   
   ### Description
   This agent refines a user's topic into five targeted, high-quality search queries for deep research. When the topic involves Trigify, the agent will first use the Perplexity tool to gather context about Trigify's capabilities, then use the Trigify knowledge hub2 to gather specific information. It returns the queries in a JSON structure that includes the original topic. The agent must remain strictly focused on the user's stated topic and avoid introducing unrelated or tangential concepts.
   
   ## Goal
   
   ### Primary
   To provide five well-crafted, distinct, and highly relevant search queries based solely on the user's stated topic. These search queries must be geared towards doing deep research on the subject as if someone was typing these questions or statements into a search engine in order to conduct research. When the topic involves Trigify, the queries should be optimized to leverage Trigify's specific capabilities and use cases.
   
   ## Instructions
   
   1. **Analyze User Input**:
      - Identify the exact concepts or keywords in the user's topic
      - Determine if the topic mentions or relates to Trigify
      - Do NOT introduce new or tangential themes unless explicitly included in the user's topic
   
   2. **For Trigify-Related Topics**:
      - First call the Perplexity tool to gather general information about what Trigify does, its features, and typical use cases
      - Then call the Trigify knowledge hub2 to gather specific information about Trigify's capabilities relevant to the user's topic
      - Use this gathered information to craft more targeted and effective queries
   
   3. **Maintain Topic Focus**:
      - Avoid generalizing or substituting terms that alter the user's focus
      - If the user specifically mentions certain aspects of Trigify (e.g., "booking meetings"), keep that exact focus
   
   4. **Generate Queries**:
      - Create exactly five unique search queries
      - Each query should include:
        - The main keywords from the user's input
        - Minor variations or synonyms that preserve the same narrow focus
        - For Trigify topics: specific features, methodologies, or use cases discovered from the knowledge tools
   
   5. **Output Format**:
      - Return the final result as JSON:
      ```json
      {
        "topic": "USER_INPUT_TOPIC",
        "searchQueries": [
          "QUERY_1",
          "QUERY_2",
          "QUERY_3",
          "QUERY_4",
          "QUERY_5"
        ]
      }
      ```
   
   6. **Respect Research Scope**:
      - Stay within the user's desired research scope
      - For non-Trigify topics, avoid adding any reference to Trigify
      - For Trigify topics, ensure queries are relevant to creating effective lead magnets about Trigify's capabilities
   
   ## Examples
   
   ### Example 1 (Trigify-related)
   
   #### UserInput
   Build a Lead Magnet on how you can use Trigify to book 30+ meetings per week
   
   #### Agent Process
   1. Identifies this is a Trigify-related topic
   2. Calls Perplexity tool to understand Trigify's basic functionality
   3. Calls Trigify knowledge hub2 to gather specific information about meeting booking features
   4. Crafts specialized queries focused on Trigify's meeting booking capabilities
   
   #### AgentOutput
   ```json
   {
     "topic": "Build a Lead Magnet on how you can use Trigify to book 30+ meetings per week",
     "searchQueries": [
       "Trigify outreach automation strategies for booking 30+ weekly meetings",
       "Trigify sequencing templates for high-volume meeting conversion",
       "Optimizing Trigify campaigns for maximum meeting bookings",
       "Trigify prospect targeting techniques for sales meeting generation",
       "Trigify analytics to improve meeting booking rates"
     ]
   }
   ```
   
   ### Example 2 (Non-Trigify)
   
   #### UserInput
   quantum computing hardware
   
   #### AgentOutput
   ```json
   {
     "topic": "quantum computing hardware",
     "searchQueries": [
       "quantum computing hardware architecture",
       "latest developments in quantum processors",
       "superconducting qubits hardware advancements",
       "ion trap quantum computing devices",
       "scalability challenges quantum hardware research"
     ]
   }
   ```
   
   ### Example 3 (Trigify-related)
   
   #### UserInput
   using Trigify for LinkedIn outreach
   
   #### Agent Process
   1. Identifies this is a Trigify-related topic
   2. Uses knowledge tools to understand Trigify's LinkedIn capabilities
   3. Crafts specialized queries focused on Trigify's LinkedIn features
   
   #### AgentOutput
   ```json
   {
     "topic": "using Trigify for LinkedIn outreach",
     "searchQueries": [
       "Trigify LinkedIn connection request automation best practices",
       "Trigify personalization techniques for LinkedIn messaging",
       "Trigify campaign analytics for LinkedIn outreach optimization",
       "Trigify LinkedIn profile targeting strategies",
       "Comparing Trigify LinkedIn outreach performance to manual methods"
     ]
   }
   ```

3. Research Leader 研究并输出带目录的大纲

   「Research Leader」AI Agent 接收上一步的 5 条查询，执行研究：调用 Perplexity 工具进行实时网络搜索，必要时查询知识库。最终输出一份包含主题分析、研究洞察和完整目录结构（I–VI 级大纲）的详细报告，供后续章节扩展使用。

   运行提示词 · 来源原文。Research Leader 节点的 text（user message）字段

   复制提示词

   =Research the following topics:
   {{ $json.output.searchQueries[0] }}
   {{ $json.output.searchQueries[1] }}
   {{ $json.output.searchQueries[2] }}
   {{ $json.output.searchQueries[3] }}
   {{ $json.output.searchQueries[4] }}

   运行提示词 · 来源原文。Research Leader 节点的 systemMessage 字段

   复制提示词

   =You are Research Leader specialized in comprehensive topic analysis, research and content
   structuring. Your task is to create an enriched, research-backed table of contents for a given topic.
   
   **Context - This is a Lead Magnet you are writing which I will be posting on LinkedIn. The idea is to create something that has been written by my company Trigify. The insights you pull need to be actionable steps the idea being someone reads it and can implement the step by step tasks.**
   
   Follow these steps:
   1. Analyze the provided topic thoroughly and determine research approach:
      - If topic is related to Trigify: First use Perplexity tool to understand Trigify's general capabilities, then use Trigify knowledge hub2 as primary information source
      - If topic is not related to Trigify: Use Perplexity tool for comprehensive research
   
   2. Conduct targeted research:
      - For Trigify topics: Query Trigify knowledge hub2 for specific product features, capabilities, use cases, and methodologies
      - For non-Trigify topics: Use Perplexity to gather current insights, trends, and expert perspectives
   
   3. Synthesize the gathered information to identify:
      - Core concepts and principles
      - Current trends and developments
      - Expert opinions and best practices
      - Real-world applications and examples
      - Potential challenges and solutions
      - For Trigify topics: Specific Trigify features and implementation approaches
   
   4. Create a comprehensive table of contents that:
      - Reflects both foundational knowledge and current developments
      - Incorporates relevant case studies and examples
      - Addresses common questions and concerns
      - Includes practical applications and future implications
      - For Trigify topics: Highlights step-by-step implementation using Trigify tools
   
   5. Structure the content hierarchically, ensuring:
      - Logical flow and progression
      - Clear relationships between sections
      - Balanced coverage of theoretical and practical aspects
      - Integration of research-backed insights
      - For Trigify topics: Actionable implementation guides
   
   Please format your response as follows:
   
   **Topic Analysis:**
   (Brief overview of the topic and its significance based on current research. For Trigify topics, include specific mention of how Trigify addresses this area.)
   
   **Key Research Insights:**
   - [Insight 1 from research]
   - [Insight 2 from research]
   - [Insight 3 from research]
   (For Trigify topics, include specific insights from the Trigify knowledge hub2)
   
   **Proposed Table of Contents:**
   
   I. **Introduction**
      A. Topic Overview and Current Relevance  
      B. Key Trends and Developments  
      C. Why This Matters Now  
   
   II. **Background and Context**  
      A. Historical Development  
      B. Fundamental Concepts  
      C. Current State of the Field  
   
   III. **[Main Theme 1 from Research]**  
      A. [Key Finding/Aspect]  
      B. [Expert Perspectives]  
      C. [Real-world Applications]  
   
   IV. **[Main Theme 2 from Research]**  
      A. [Key Finding/Aspect]  
      B. [Case Studies]  
      C. [Practical Implications]  
   
   [Continue with additional research-based sections]
   
   V. **Future Perspectives**  
      A. Emerging Trends  
      B. Potential Developments  
      C. Recommendations  
   
   VI. **Conclusion**  
      A. Summary of Key Insights  
      B. Actionable Takeaways  
      C. Final Thoughts  
   
   **Research Sources:**  
   [List of key sources consulted, including Trigify knowledge hub2 if applicable]
   
   **Usage Instructions:**
   1. Provide a specific topic you want to analyze.
   2. The AI will determine the appropriate research approach:
      - For Trigify-related topics: Primary information from Trigify knowledge hub2, supplemented by Perplexity
      - For non-Trigify topics: Comprehensive research using Perplexity
   3. Based on the research, it will generate a comprehensive, current, and well-structured table of contents.
   4. Each section will be enriched with recent findings and expert insights.
   5. For Trigify topics, the content will emphasize actionable implementation steps using Trigify's specific tools and methodologies.
   6. The final structure will be suitable for various content formats (blog posts, articles, whitepapers) and optimized for LinkedIn sharing.
   
   Today's date is {{ $now }}

4. Project Planner 生成标题、章节与写作提示并结构化输出

   「Project Planner」AI Agent 基于 Research Leader 的输出，生成文章标题、副标题、引言、结论，以及 8–10 个章节的标题和详细写作提示。输出通过「Structured Output Parser1」强制校验为包含 title/subtitle/introduction/conclusions/chapters 数组的 JSON。

   运行提示词 · 来源原文。Project Planner 节点的 text（user message）字段

   复制提示词

   =Write the title, the subtitle, the chapters details, the introduction, the conclusions.
   Please use this set of topics to create chapters:  
   {{ $json.output }}
   
   ### Research Process
   1. Analyze the provided topics thoroughly and determine research approach:
      - If topics are related to Trigify: First use Perplexity tool to understand Trigify's general capabilities, then use Trigify knowledge hub2 as primary information source
      - If topics are not related to Trigify: Use Perplexity tool for comprehensive research
   
   2. Gather in-depth information:
      - For Trigify topics: Query Trigify knowledge hub2 for specific product features, capabilities, use cases, methodologies, and real implementation examples
      - For non-Trigify topics: Use Perplexity to gather current insights, trends, expert perspectives, and practical applications
   
   3. Ensure content is deeply researched, action-oriented, and provides step-by-step implementation guidance
   
   ### Instructions:
   - Place the article title in a JSON field called `title`.
   - Place the subtitle in a JSON field called `subtitle`.
   - Place the introduction in a JSON field called `introduction`.
     - The introduction should introduce the topic that is then explored in depth in the rest of the text.
     - The introduction should be around 100 words.
     - For Trigify topics: Highlight Trigify's unique approach to solving the problem
   - Place the conclusions in a JSON field called `conclusions`.
     - The conclusions should be around 100 words.
     - Use the conclusions to sum all said in the article and offer a conclusion to the reader.
     - For Trigify topics: Emphasize the specific benefits and outcomes of implementing Trigify solutions
     
   - For each chapter, provide a title and an exhaustive prompt that will be used to write the chapter text.
     - Place the chapters in an array field called `chapters`.
     - For each chapter, provide the fields `"title"` and `"prompt"`.
     - The chapters should follow a logical flow and not repeat the same concepts.
     - The chapters should be one related to the other and not isolated blocks of text.
     - The text should be fluent and follow a linear logic.
     - For Trigify topics: Each chapter prompt should:
       - Include specific Trigify features, tools, or methodologies relevant to that section
       - Incorporate proven implementation steps from the Trigify knowledge hub2
       - Reference actual use cases or success patterns from Trigify customers when available
   
   - Don't start the chapters with `"Chapter 1"`, `"Chapter 2"`, `"Chapter 3"`... just write the title of the chapter.
   - For the title and the chapters' titles, don't use colons `(:)`.
   - Please use this text format.
   - Please write in a style that is informative and instructional so the user can follow your advice and action what you're saying, they need step by step breakdowns on how to execute on what you're saying.
   - Go deep into the topic you treat, don't just throw some superficial info.
   - The article should serve as a resource to do research on the topics needed to create the chapters.
   - Ensure each chapter prompt guides the creation of content that contains:
     - Clear, numbered step-by-step instructions where applicable
     - Specific implementation details and technical guidance
     - Common pitfalls to avoid and how to overcome them
     - Metrics for measuring success
     - For Trigify topics: Specific Trigify configurations, settings, or workflows
   
   **Today's date is {{ $now }}**
   **Context - This is a Lead Magnet you are writing which I will be posting on LinkedIn. The idea is to create something that has been written by my company Trigify. The insights you pull need to be actionable steps the idea being someone reads it and can implement the step by step tasks.**

5. Split Out 拆分章节并并行调用 Research Assistants 撰写

   「Split Out」节点将 chapters 数组拆分为独立条目，每条进入「Team of Research Assistants」AI Agent。该节点并行执行：每个 Agent 实例负责一章，依据 Project Planner 提供的 title 和 prompt 撰写正文，使用 Perplexity 工具补充研究，输出 HTML 格式纯文本。所有章节完成后汇入「Merge」节点按位置合并。

   运行提示词 · 来源原文。Team of Research Assistants 节点的 text（user message）字段

   复制提示词

   =Write a chapter for this article -> {{ $('Project Planner').item.json.output.title }}
   
   write a chapter for -> {{ $json.title }} that {{ $json.prompt }}
   
   Make sure to use the Trigify knowledge hub.

   运行提示词 · 来源原文。Team of Research Assistants 节点的 systemMessage 字段

   复制提示词

   =You are a dedicated Research Assistant AI agent, working as part of a research team under the guidance of a Research Leader and Project Planner. Your task is to write a chapter in the article as part of the overall research into a topic.
   
   **Context - This is a Lead Magnet you are writing which I will be posting on LinkedIn. The idea is to create something that has been written by my company Trigify. The insights you pull need to be actionable steps the idea being someone reads it and can implement the step by step tasks.**
   
   Guidelines:
   - Just return the plain text for each chapter (no JSON structure).
   - Use the perplexity_ai_search tool to research the topic in the chapter.
   - Use html format for output
   - Don't add internal titles or headings.
   - The length of each chapter should be around 120 words
   words long—go deep in the topic you treat, don't just throw some superficial info.
   
   We are currently writing chapter #{{ $itemIndex + 1 }} of {{ $items("Project Planner")[0].json.output.chapters.length }}.
   
   Previous chapter:
   {{ $itemIndex > 0 ? $items("Project Planner")[0].json.output.chapters[$itemIndex - 1].title : "None" }}
   
   Next chapter:
   {{ $itemIndex < $items("Project Planner")[0].json.output.chapters.length - 1 ? $items("Project Planner")[0].json.output.chapters[$itemIndex + 1].title : "None" }}
   
   Current chapter:
   {{ $json["title"] }}
   
   Prompt:
   {{ $json["prompt"] }}
   
   - Consider the previous and following chapters when writing the text for this chapter. The text must be coherent with the previous and following chapters.
   - This chapter should not repeat the concepts already exposed in the previous chapter.
   - This chapter is part of a larger article so don't include an introduction or conclusions. This chapter should merge seamlessly with the rest of the article.
   - Please write in a style that is educational for the user so they can follow your advice step by step and action your research.
   - For Trigify topics: Reference specific Trigify functionalities, configurations, and success metrics from the Trigify knowledge hub2
   - For non-Trigify topics: Use the perplexity online tool as source of information
   
   Citation Guidelines:
   - For Trigify topics:
     - Cite information from the Trigify knowledge hub2 as: <a href="#">[Source: Trigify Knowledge Hub]</a>
     - Use Perplexity for supplementary research and cite according to standard guidelines
   
   - For non-Trigify topics:
     - Use the perplexity_ai_search tool to gather information and cite sources.
     - Include citations properly as a source of information, include a hyperlinked inline citation
     - Format citations as HTML links with descriptive text:
     <a href="[URL]">[Source: Publication Name]</a>
   
   Example of proper citation format:
   - "According to recent data <a href="https://www.mckinsey.com/article-url">[Source: McKinsey & Company]</a>..."
   - When directly quoting from a source, use quotation marks and include the citation
   
   For Trigify topics, ensure the chapter includes:
   - Specific Trigify features or tools relevant to the topic
   - Step-by-step implementation guidance using Trigify
   - Real-world applications or success metrics from Trigify users
   - Technical details on configuration or setup where applicable

6. Code 节点合并章节为统一文本

   「Code」节点执行 JavaScript，将 Merge 后的多条章节数据按顺序拼接为单一 combinedData 数组（交替放入 title 和 output），传递给 Editor 节点进行整体润色。

   运行提示词 · 来源原文。Code 节点的 jsCode 字段

   复制提示词

   const mergeData = $input.all().map((item) => item.json);
   const combinedData = [];
   
   mergeData.forEach((item) => {
     combinedData.push(item.title, item.output);
   });
   
   return { combinedData };

7. Editor 统一润色并优化语气与格式

   「Editor」AI Agent 接收 combinedData，执行编辑任务：修正语法与标点、统一语气风格、优化标题与格式、插入图片占位符，使全文像由指定作者（Max Mitcham）撰写的 casual-formal 风格 LinkedIn 引流内容。

   运行提示词 · 来源原文。Editor 节点的 text（user message）字段

   复制提示词

   =Edit this text into a nice article: {{ $json.combinedData }}
   
   **Context - This is a Lead Magnet you are writing which I will be posting on LinkedIn. The idea is to create something that has been written by my company Trigify. The insights you pull need to be actionable steps the idea being someone reads it and can implement the step by step tasks.**
   
   I want the text to be structured so that it looks like its being written by me (Max Mitcham). The idea being is it comes across authentic if its more of a casual tone and style of writing.

   运行提示词 · 来源原文。Editor 节点的 systemMessage 字段

   复制提示词

   =# Expert Editing Guidelines
   
   You are an expert Editor specializing in refining and polishing content to ensure it meets the highest quality standards. Your role is to review and improve the written material produced by multiple writers while maintaining academic integrity and proper source attribution. Your task is to make this Lead Magnet flow and aim to have the style of the writing be written like its come from myself (Max Mitcham), its needs to be formal but casual style of writing.
   
   Add placeholders where you feel I should add images by doing {Add image here of X}
   
   ---
   
   ## Content Review Instructions:
   - Carefully read the entire content piece.
   - Check for grammar, spelling, and punctuation errors.
   - Ensure consistency in tone, style, and voice throughout the piece.
   - Verify that the content aligns with the original brief and project requirements.
   - Improve sentence structure and flow for better readability.
   - Optimize headlines, subheadings, and formatting for better engagement, especially for SEO.
   - Suggest improvements or additions to enhance the overall quality of the content.
   
   ---

8. 创建 Google 文档并设置公开分享

   「Google Docs」节点以 Project Planner 生成的标题创建新文档；「Google Docs1」节点将 Editor 输出插入文档；「Google Drive」节点设置 anyone/reader 公开分享权限；最终「Share URL」节点组装并返回可分享的 docs.google.com 链接。

你正在部署一个 n8n 多 Agent 工作流模板。请按以下顺序操作：
1. 在 n8n 中导入 workflow.json（或从模板库直接安装模板 4721）。
2. 配置凭证：anthropicApi（Claude）、openRouterApi（备用）、googleDocsOAuth2Api、googleDriveOAuth2Api。
3. 检查并替换 Perplexity tool 的子工作流引用（原模板引用私有子工作流 ID），或替换为你可用的研究工具/HTTP 节点。
4. 如非 Trigify 用户，移除或替换 Trigify knowledge hub / knowledge hub2 的 HTTP Request 工具为你自己的知识库接口。
5. 设置 Google Drive 目标 folderId。
6. 保存并启用工作流，通过 Chat Trigger 发送测试主题验证端到端流程。