# How to Build an Obsidian Vault That Runs Your Entire Business While You Sleep - (Full Course)
**作者**: CyrilXBT
**日期**: 2026-05-13T01:54:15.000Z
**来源**: [https://x.com/cyrilXBT/status/2054379666316693719](https://x.com/cyrilXBT/status/2054379666316693719)
---
Most business owners spend their days doing two kinds of work.
The work that moves the business forward.
And the work that keeps the business from falling behind.
The first kind is irreplaceable. Strategy. Relationships. Creative decisions. The judgment calls only you can make.
The second kind is administrative overhead. Updating project statuses. Compiling reports. Processing information. Tracking what happened. Preparing context for what comes next.
The second kind is eating the first kind alive.
The average knowledge worker spends 4 to 6 hours a day on administrative coordination. Maintaining records. Finding information. Moving context between tools. Updating people on status that could be automated.
That is not a productivity problem.
That is an architecture problem.
The Obsidian vault business operating system solves it at the architectural level.
Not by making you faster at administrative work.
By removing you from it entirely.
This article is the complete guide to building an Obsidian vault that runs your entire business while you sleep.
## The Architecture Before the Tools
Before you install anything, understand what you are building.
A business operating system is not a collection of notes.
It is a set of interconnected systems where information flows automatically from where it is created to where it is needed, and actions are triggered without you initiating them.
Three layers make this work.
The Knowledge Layer is Obsidian itself. Plain text Markdown files organized in a structure that is both human-readable and machine-navigable. Every piece of information your business generates lives here in a format that never becomes inaccessible, never requires a subscription to read, and never locks you into a proprietary system.
The Intelligence Layer is Claude Code connected via MCP. This is what transforms your vault from a passive repository into an active operating system. Claude reads your files, processes information, makes connections, and generates outputs. It is the thinking that happens while you are not thinking.
The Automation Layer is N8N. This is what makes the system run on schedule and on trigger without you initiating anything. N8N is the clock and the nervous system. It fires workflows at the right time and routes information to the right place.
Together these three layers create a business that operates continuously whether you are working or not.
## Building the Foundation: Vault Structure for Business Operations
The vault structure is the foundation that determines whether everything built on top of it works reliably.
Most Obsidian users build vaults for personal knowledge management. The folder structure reflects how a human brain organizes topics.
A business operating system vault is built differently. The structure reflects how business processes flow, not how topics relate.
Here is the complete folder architecture:
The GENERATED and QUEUE folders are the operational core of the business operating system.
QUEUE is how you communicate asynchronously with the intelligence layer. You drop a file describing what you need. The system processes it on its next cycle. The output appears in GENERATED.
You capture tasks when they occur to you. The system executes them when it runs. You review results when it is convenient.
This asynchronous loop is what makes the vault run your business while you sleep. You are never waiting for it. It is never waiting for you.
## The Master CLAUDE.md: Your Business Constitution
The CLAUDE.md file in your SYSTEM folder is the single most important file in your entire business operating system.
Every automated workflow reads it before doing anything. It is the context that makes generic AI capability into business-specific intelligence.
A poorly written CLAUDE.md produces generic outputs that require significant editing.
A precisely written CLAUDE.md produces outputs that require minimal editing because the intelligence layer understands your business as well as you do.
```
# Business Operating System — CLAUDE.md
## Business Identity
Business name: [YOUR BUSINESS]
What you do: [SPECIFIC DESCRIPTION]
Who you serve: [SPECIFIC CUSTOMER DESCRIPTION]
How you make money: [REVENUE MODEL]
Stage: [EARLY/GROWTH/ESTABLISHED]
## Active Clients
[CLIENT NAME]: [ENGAGEMENT TYPE] | Status: [CURRENT STATUS] |
Next action: [SPECIFIC NEXT STEP]
[Repeat for each client]
## Active Projects
[PROJECT NAME]: [BRIEF DESCRIPTION] | Status: [STATUS] |
Deadline: [DATE] | Next milestone: [MILESTONE]
[Repeat for each project]
## Voice and Communication
Writing style: [HOW YOU COMMUNICATE]
Tone: [YOUR TONE]
What you never say: [SPECIFIC PHRASES TO AVOID]
CTA style: [HOW YOU END PIECES]
## Revenue Targets
Monthly target: [AMOUNT]
Current MRR: [UPDATE MONTHLY]
Key revenue metric: [THE NUMBER YOU WATCH]
## Content Operation
Publishing schedule: [FREQUENCY AND PLATFORMS]
Content pillars: [YOUR MAIN TOPICS]
Formats: [WHAT YOU PRODUCE]
## Operating Rules
- Never delete files. Archive with timestamp instead.
- Never send communication without human review.
- Always date-stamp generated files YYYY-MM-DD.
- Log every write operation in SYSTEM/logs/operations.md.
- When uncertain: deposit in GENERATED and flag for review.
- Escalate to human: anything involving money, commitments,
client relationships, or reputational decisions.
## Weekly Focus
[UPDATE EVERY MONDAY — what matters most this week]
[This weights every system action toward current priorities]
```
The Weekly Focus section is what keeps the system aligned with where your business actually is. Update it every Monday morning. It takes 2 minutes and dramatically improves the relevance of every automated output for the week.
## The Six Systems That Run Your Business
The vault runs six interconnected business systems. Each one owns a specific function and passes information to the others through shared vault structure.
System 1: Client Intelligence
The client intelligence system ensures you always have full context on every client relationship without manually maintaining it.
The Pre-Call Brief
30 minutes before every client call, this workflow fires via N8N:
Read the client folder for [CLIENT NAME] in 01 - CLIENTS.
Compile a complete pre-call brief covering:
- Relationship timeline: key events and decisions in order
- Current engagement status: what was agreed, what was delivered
- Open items: anything promised that has not been delivered
- Last communication: date, medium, key points
- Their stated priorities: what they care about most
- Suggested agenda: 3 items for this call based on status
- One thing to make sure you say: the most important point
Format for a 5-minute read. Be specific, not general.
Save to: 08 - GENERATED/briefings/[DATE]-[CLIENT]-precall.md
You never walk into a client call underprepared again.
The Communication Logger
Every client email, call note, or interaction you log in your daily note triggers an update to the client communication folder. The system extracts the key information, formats it consistently, and appends it to the client's communications folder.
You log once. The system files it permanently in the right place.
System 2: Project Operations
The project operations system keeps every active project moving forward without you manually tracking status.
The Daily Project Pulse
Every morning at 6AM the project pulse workflow reads all active project folders and generates a project status summary:
Read all overview.md files in 02 - PROJECTS.
For each project with status NOT COMPLETE:
Report:
- Project name and client
- Current phase
- What was completed yesterday (from daily note)
- What is currently blocked
- Next action and owner
- Days until deadline
Flag any project that:
- Is within 7 days of deadline with incomplete tasks
- Has been in the same status for 5+ days
- Has a blocked item with no resolution plan
Save to: 08 - GENERATED/briefings/[DATE]-project-pulse.md
The Project Auto-Updater
When you mark anything complete in your daily note using the convention DONE: [project] — [deliverable], a workflow reads the entry and updates the corresponding project's task list automatically.
You work. The project folder updates itself.
System 3: Content Production
The content system produces a consistent volume of content without you being the bottleneck.
The Morning Content Brief
Every morning alongside the project pulse, the content brief workflow runs:
Read from the vault:
1. Any research briefs in 08 - GENERATED/briefings/
created in the last 48 hours
2. Your recent daily notes for ideas you captured
3. 03 - OPERATIONS/content/calendar.md for scheduled topics
4. Your content pillars from CLAUDE.md
Generate a content brief with:
- Three content angles for today ranked by potential
- The hook option for each angle
- Which format fits each angle best
- Whether each requires additional research
Save to: 08 - GENERATED/briefings/[DATE]-content-brief.md
The Draft Engine
Drop a file in QUEUE named DRAFT-[content type]-[topic].md with any specific notes you want included.
The queue processor picks it up, reads your research brief on the topic, applies your voice profile from CLAUDE.md, and produces a first draft in GENERATED/drafts.
First drafts that require under 10 minutes of editing for 80% of pieces.
The Performance Tracker
Every Monday morning the content performance workflow pulls last week's data across every platform and generates an analysis:
Compile content performance for the past 7 days:
- Top 3 pieces by impressions
- Top 3 pieces by saves/bookmarks
- Lowest performing pieces
- The specific pattern that explains the performance difference
- One format or topic to increase based on data
- One format or topic to reduce based on data
- Next week's recommended focus
Save to: 08 - GENERATED/reports/[DATE]-content-performance.md
System 4: Financial Operations
The financial system keeps your numbers current without a separate administrative process.
Revenue Logging
Every time you log a payment in your daily note using the convention RECEIVED: $[amount] — [client] — [description], a workflow captures the entry, updates your monthly revenue tracker in OPERATIONS/finances, and appends the record to the client's invoice folder.
The Monthly Financial Brief
On the first of every month a workflow generates the previous month's financial summary:
Read all revenue entries from daily notes for [PREVIOUS MONTH].
Read all expense records from OPERATIONS/finances/expenses/.
Read the monthly target from CLAUDE.md.
Generate a financial brief:
- Total revenue vs target (with percentage)
- Month over month comparison
- Revenue by client
- Top expense categories
- Net position
- Trend: trajectory for next month if pattern continues
Save to: 08 - GENERATED/reports/[DATE]-financial-brief.md
The Invoice Reminder
On the 25th of every month the invoice reminder workflow identifies retainer clients, checks payment records, and generates draft invoice emails for any outstanding amounts.
The drafts land in GENERATED/communications for your review before sending.
You approve. The system drafted.
System 5: Research and Intelligence
The intelligence system keeps you informed about what matters to your business without you spending hours reading.
The Daily Intelligence Brief
Every morning the intelligence workflow runs alongside the project pulse and content brief:
Pull from configured sources:
- Top AI and relevant industry news from last 24 hours
- Competitor activity if monitoring is configured
- Any mentions of your brand or key topics
Synthesize into an intelligence brief:
- Three most important developments and why they matter
- One opportunity this creates for your business
- One risk this creates for your business
- Recommended action if any
Save to: 08 - GENERATED/briefings/[DATE]-intelligence.md
The Research Queue
Drop a file in QUEUE named RESEARCH-[topic].md with any specific questions you want answered.
The research workflow reads the request, searches for relevant information, cross-references with existing vault knowledge, and produces a structured research brief in GENERATED/briefings.
You queue research at midnight. The brief is ready at 6AM.
System 6: Performance and Review
The performance system closes the loop on everything the other systems produce.
The Weekly Review
Every Sunday at 8PM the weekly review workflow reads the entire week:
Read all daily notes from the past 7 days.
Read all project status changes.
Read the content performance report.
Read the financial entries.
Read all generated briefings and outputs from the week.
Synthesize a weekly review:
- What moved forward and the specific cause
- What did not move and the honest reason
- One pattern that appeared across the week
- Three priorities for next week ranked by leverage
- One personal insight from the week's patterns
Save to: 08 - GENERATED/reports/[DATE]-weekly-review.md
The Quarterly Business Review
On the first day of each quarter the quarterly review workflow reads three months of weekly reviews and generates a strategic retrospective.
Where the business went during the quarter. What patterns drove the results. What to do differently next quarter. One big bet worth making.
The quarterly review used to be a full day of manual analysis.
Now it is a 20-minute read of what the system synthesized.
## Setting Up the Automation Layer
All six systems require N8N to run automatically.
Self-hosted N8N on a $5 DigitalOcean droplet gives you unlimited workflow runs with no per-execution pricing.
The workflow types you need:
Cron Triggers for scheduled operations. Morning briefing at 6AM. Weekly review Sunday at 8PM. Monthly reports on the 1st. Invoice reminders on the 25th.
File Watch Triggers for event-based operations. Queue processor watches the QUEUE folder for new files. Project updater watches project folders for changes.
Webhook Triggers for external events. New Stripe payment updates revenue tracker. New calendar event triggers pre-call brief.
The N8N workflow structure for every Claude-powered operation follows the same pattern:
Trigger → Read CLAUDE.md → Read relevant vault content
→ Call Claude API with compiled context → Write output
to vault → Send notification → Log operation
The Claude API call always includes the CLAUDE.md content as system context. This is what makes every automated output business-specific rather than generic.
## The Daily Experience After the System Is Running
Here is what a Tuesday morning looks like once the vault is operational.
5:58 AM: Your phone receives a Telegram notification. Four items are ready in the vault. Project pulse. Content brief. Intelligence brief. A research brief you queued last night.
6:10 AM: You read the project pulse in 4 minutes. Two projects on track. One flagged as overdue for status update. You know exactly what needs attention before you have looked at email.
6:15 AM: You open the content brief. Three angles for today. You pick one and add it to the content queue file.
8:45 AM: You have a client call at 9AM. You open the pre-call brief that was automatically generated at 8:30. You spend 5 minutes reading. You walk in more prepared than you would have been after 30 minutes of manual prep.
During the day: You log notes, ideas, completions, and payments in your daily note using the simple conventions. Everything routes to the right place automatically.
4:00 PM: The content draft you queued this morning is waiting in GENERATED/drafts. You spend 8 minutes editing. You schedule it.
10:30 PM: Before bed you drop two research requests in the QUEUE folder. They will be processed before you wake up.
That is the experience.
Not a system you maintain.
A system that maintains your business while you direct it.
## The Compounding Effect
The most important thing about the Obsidian business operating system is not what it does in week one.
It is what happens in month six.
Every client interaction that gets logged makes the pre-call briefs richer.
Every content performance data point makes the content briefs more accurate.
Every research brief adds to the intelligence the system draws from.
Every weekly review becomes material for better quarterly analysis.
The system compounds.
After six months of consistent use, your vault knows your business at a depth that would take a human employee years to develop.
The morning briefings are calibrated to your actual patterns.
The research builds on months of accumulated context.
The content system knows what performs for your specific audience.
This is the difference between a tool you use and a system that compounds.
Build the vault this weekend.
Start with the CLAUDE.md and the Morning Briefing workflow.
Add one system per week.
By month two you have a business operating system running your entire operation.
By month six you will not remember how you ran the business without it.
Follow @cyrilXBT for the exact N8N workflows, Claude prompts, and vault templates that power this entire system.
## 相关链接
- [CyrilXBT](https://x.com/cyrilXBT)
- [@cyrilXBT](https://x.com/cyrilXBT)
- [347K](https://x.com/cyrilXBT/status/2054379666316693719/analytics)
- [@cyrilXBT](https://x.com/@cyrilXBT)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [9:54 AM · May 13, 2026](https://x.com/cyrilXBT/status/2054379666316693719)
- [347.8K Views](https://x.com/cyrilXBT/status/2054379666316693719/analytics)
- [View quotes](https://x.com/cyrilXBT/status/2054379666316693719/quotes)
---
*导出时间: 2026/5/14 16:29:42*
---
## 中文翻译
# 如何建立一个 Obsidian 知识库，在你睡觉时运营你的整个生意 —— (完整课程)
**作者**: CyrilXBT
**日期**: 2026-05-13T01:54:15.000Z
**来源**: [https://x.com/cyrilXBT/status/2054379666316693719](https://x.com/cyrilXbt/status/2054379666316693719)
---
大多数企业主每天都在处理两类工作。
推动业务向前发展的工作。
以及防止业务落后的工作。
第一类工作是不可替代的。战略。关系。创意决策。只有你能做出的判断。
第二类工作是行政开销。更新项目状态。汇编报告。处理信息。追踪发生的事情。为接下来的事情准备背景资料。
第二类工作正在吞噬第一类工作。
普通知识工作者每天花费 4 到 6 小时进行行政协调。维护记录。查找信息。在工具之间转移上下文。向人们更新本可以自动化的状态。
这不是一个生产力问题。
这是一个架构问题。
Obsidian 知识库业务操作系统在架构层面解决了这个问题。
不是让你在行政工作上变得更快。
而是将你完全从中解放出来。
这篇文章是建立一个 Obsidian 知识库的完整指南，该知识库可以在你睡觉时运行你的整个业务。
## 工具之前的架构
在你安装任何东西之前，先了解你在构建什么。
业务操作系统不是笔记的集合。
它是一组相互连接的系统，信息自动从创建的地方流向需要的地方，并且操作在无需你启动的情况下被触发。
三个层使其工作。
知识层是 Obsidian 本身。以人类可读且机器可导航的结构组织的纯文本 Markdown 文件。您的业务生成的每一条信息都以一种永远不会无法访问、不需要订阅即可阅读、也不会将您锁定在专有系统中的格式存在。
智能层是通过 MCP 连接的 Claude Code。这是将您的知识库从被动存储库转变为主动操作系统的关键。Claude 读取您的文件，处理信息，建立联系，并生成输出。这是在您不思考时发生的思考。
自动化层是 N8N。这是使系统按计划和时间表运行而无需您启动任何内容的原因。N8N 是时钟和神经系统。它在正确的时间触发工作流，并将信息路由到正确的位置。
这三层共同创建了一个无论您是否工作都能持续运行的业务。
## 建立基础：用于业务运营的知识库结构
知识库结构是决定建立在其上的一切能否可靠工作的基础。
大多数 Obsidian 用户为个人知识管理 (PKM) 构建知识库。文件夹结构反映了人脑如何组织主题。
业务操作系统知识库的构建方式不同。结构反映业务流程如何流动，而不是主题如何关联。
以下是完整的文件夹架构：
GENERATED 和 QUEUE 文件夹是业务操作系统的操作核心。
QUEUE 是您与智能层异步通信的方式。您放入一个描述您需求的文件。系统在下一个周期处理它。输出出现在 GENERATED 中。
您在想到任务时捕获它们。系统在运行时执行它们。在您方便时审查结果。
这种异步循环是知识库在您睡觉时运行您的业务的原因。您永远不需要等待它。它永远不需要等待您。
## 主控 CLAUDE.md：您的商业宪法
SYSTEM 文件夹中的 CLAUDE.md 文件是整个业务操作系统中最重要的单个文件。
每个自动化的工作流在做任何事情之前都会读取它。它是将通用 AI 能力转化为特定业务智能的上下文。
编写拙劣的 CLAUDE.md 会产生通用输出，需要大量编辑。
精确编写的 CLAUDE.md 产生的输出只需很少的编辑，因为智能层对您的业务与您一样了解。
```
# Business Operating System — CLAUDE.md
## Business Identity
Business name: [YOUR BUSINESS]
What you do: [SPECIFIC DESCRIPTION]
Who you serve: [SPECIFIC CUSTOMER DESCRIPTION]
How you make money: [REVENUE MODEL]
Stage: [EARLY/GROWTH/ESTABLISHED]
## Active Clients
[CLIENT NAME]: [ENGAGEMENT TYPE] | Status: [CURRENT STATUS] |
Next action: [SPECIFIC NEXT STEP]
[Repeat for each client]
## Active Projects
[PROJECT NAME]: [BRIEF DESCRIPTION] | Status: [STATUS] |
Deadline: [DATE] | Next milestone: [MILESTONE]
[Repeat for each project]
## Voice and Communication
Writing style: [HOW YOU COMMUNICATE]
Tone: [YOUR TONE]
What you never say: [SPECIFIC PHRASES TO AVOID]
CTA style: [HOW YOU END PIECES]
## Revenue Targets
Monthly target: [AMOUNT]
Current MRR: [UPDATE MONTHLY]
Key revenue metric: [THE NUMBER YOU WATCH]
## Content Operation
Publishing schedule: [FREQUENCY AND PLATFORMS]
Content pillars: [YOUR MAIN TOPICS]
Formats: [WHAT YOU PRODUCE]
## Operating Rules
- Never delete files. Archive with timestamp instead.
- Never send communication without human review.
- Always date-stamp generated files YYYY-MM-DD.
- Log every write operation in SYSTEM/logs/operations.md.
- When uncertain: deposit in GENERATED and flag for review.
- Escalate to human: anything involving money, commitments,
client relationships, or reputational decisions.
## Weekly Focus
[UPDATE EVERY MONDAY — what matters most this week]
[This weights every system action toward current priorities]
```
“每周重点”部分使系统与您的业务实际所在保持一致。每周一早上更新。这需要 2 分钟，但能显著提高本周每次自动输出的相关性。
## 运营您业务的六大系统
知识库运行六个相互关联的业务系统。每个系统拥有特定功能，并通过共享的知识库结构将信息传递给其他系统。
系统 1：客户智能
客户智能系统确保您始终拥有每个客户关系的完整上下文，而无需手动维护。
通话前简报
每次客户通话前 30 分钟，此工作流通过 N8N 触发：
读取 01 - CLIENTS 中 [CLIENT NAME] 的客户文件夹。
编译一份完整的通话前简报，涵盖：
- 关系时间线：按顺序排列的关键事件和决策
- 当前参与状态：商定了什么，交付了什么
- 未完成事项：已承诺但未交付的任何内容
- 最后沟通：日期、媒介、关键点
- 他们陈述的优先事项：他们最关心的内容
- 建议议程：基于状态的 3 个通话项目
- 一件必须确保要说的事情：最重要的一点
格式化为 5 分钟阅读。要具体，不要笼统。
保存至：08 - GENERATED/briefings/[DATE]-[CLIENT]-precall.md
您再也不会毫无准备地走进客户通话了。
沟通记录器
您在每日笔记中记录的每封客户电子邮件、通话记录或互动，都会触发对客户沟通文件夹的更新。系统提取关键信息，一致地格式化它，并将其附加到客户的沟通文件夹中。
您记录一次。系统将其永久地归档在正确的地方。
系统 2：项目运营
项目运营系统保持每个活跃的项目向前推进，无需您手动跟踪状态。
每日项目脉搏
每天早上 6 点，项目脉搏工作流读取所有活跃的项目文件夹并生成项目状态摘要：
读取 02 - PROJECTS 中的所有 overview.md 文件。对于状态为未完成 (NOT COMPLETE) 的每个项目：
报告：
- 项目名称和客户
- 当前阶段
- 昨天完成了什么（来自每日笔记）
- 当前受阻的内容
- 下一步行动和负责人
- 距离截止日期的天数
标记任何符合以下条件的项目：
- 距离截止日期 7 天内有未完成的任务
- 处于相同状态 5 天以上
- 有受阻项目且无解决方案
保存至：08 - GENERATED/briefings/[DATE]-project-pulse.md
项目自动更新器
当您在每日笔记中使用约定 DONE: [project] — [deliverable] 标记任何完成内容时，工作流会读取该条目并自动更新相应项目的任务列表。
您工作。项目文件夹自我更新。
系统 3：内容生产
内容系统在不让您成为瓶颈的情况下产生持续的内容量。
早晨内容简报
每天早上与项目脉搏一起，内容简报工作流运行：
从知识库读取：
1. 过去 48 小时内在 08 - GENERATED/briefings/ 中创建的任何研究简报
2. 您最近的每日笔记中您捕获的想法
3. 03 - OPERATIONS/content/calendar.md 中的计划主题
4. 您的内容支柱（来自 CLAUDE.md）
生成包含以下内容的内容简报：
- 今天按潜力排名的三个内容角度
- 每个角度的钩子选项
- 哪种格式最适合每个角度
- 每个角度是否需要额外研究
保存至：08 - GENERATED/briefings/[DATE]-content-brief.md
起草引擎
在 QUEUE 中放入一个名为 DRAFT-[content type]-[topic].md 的文件，并附上您想包含的任何具体说明。
队列处理器会接收它，阅读您关于该主题的研究简报，应用您在 CLAUDE.md 中的语音资料，并在 GENERATED/drafts 中生成初稿。
对于 80% 的作品，初稿只需不到 10 分钟的编辑。
性能跟踪器
每周一早上，内容性能工作流提取上周跨每个平台的数据并生成分析：
汇编过去 7 天的内容性能：
- 浏览量最高的前 3 篇
- 收藏/书签最多的前 3 篇
- 表现最差的内容
- 解释性能差异的具体模式
- 基于数据增加的一种格式或主题
- 基于数据减少的一种格式或主题
- 下周的推荐重点
保存至：08 - GENERATED/reports/[DATE]-content-performance.md
系统 4：财务运营
财务系统使您的数字保持最新，无需单独的行政流程。
收入记录
每当您在每日笔记中使用约定 RECEIVED: $[amount] — [client] — [description] 记录付款时，工作流会捕获该条目，更新 OPERATIONS/finances 中的月度收入跟踪器，并将记录附加到客户的发票文件夹。
月度财务简报
每月的第一天，工作流生成上个月的财务摘要：
从每日笔记中读取 [上个月] 的所有收入条目。
从 OPERATIONS/finances/expenses/ 读取所有费用记录。
从 CLAUDE.md 读取月度目标。
生成财务简报：
- 总收入与目标对比（含百分比）
- 月环比比较
- 按客户分类的收入
- 主要费用类别
- 净头寸
- 趋势：如果模式持续，下个月的轨迹
保存至：08 - GENERATED/reports/[DATE]-financial-brief.md
发票提醒
每月的 25 号，发票提醒工作流识别保留金客户，检查付款记录，并为任何未结金额生成发票邮件草稿。
草稿会落在 GENERATED/communications 中供您在发送前审查。
您批准。系统起草。
系统 5：研究和情报
情报系统让您了解对业务重要的事情，而无需花费数小时阅读。
每日情报简报
每天早上，情报工作流与项目脉搏和内容简报一起运行：
从配置的来源中提取：
- 过去 24 小时内最重要的 AI 和相关行业新闻
- 如果配置了监控，则为竞争对手活动
- 对您的品牌或关键主题的任何提及
综合成情报简报：
- 三个最重要的发展及其重要性
- 这为您的业务创造的一个机会
- 这为您的业务创造的一个风险
- 推荐的行动（如有）
保存至：08 - GENERATED/briefings/[DATE]-intelligence.md
研究队列
在 QUEUE 中放入一个名为 RESEARCH-[topic].md 的文件，并附上您想得到答案的任何具体问题。
研究工作流读取请求，搜索相关信息，与现有知识库知识交叉参考，并在 GENERATED/briefings 中生成结构化的研究简报。
您在午夜排队研究。简报在早上 6 点准备好。
系统 6：绩效和审查
绩效系统为其他系统生成的所有内容关闭循环。
每周审查
每周日晚上 8 点，每周审查工作流读取整周的内容：
读取过去 7 天的所有每日笔记。
读取所有项目状态变更。
读取内容性能报告。
读取财务条目。
读取本周生成的所有简报和输出。
综合每周审查：
- 什么向前推进了以及具体原因
- 什么没有推进以及诚实的原因
- 本周出现的一种模式
- 下周按杠杆排名的三个优先事项
- 从本周模式得出的一个个人见解
保存至：08 - GENERATED/reports/[DATE]-weekly-review.md
季度业务审查
每季度的第一天，季度审查工作流阅读三个月的每周审查并生成战略回顾。
业务在本季度的去向。什么模式推动了结果。下季度要做什么不同的事情。一个值得做的大赌注。
季度审查曾经是一整天的手动分析。
现在它是系统综合内容的 20 分钟阅读。
## 设置自动化层
所有六个系统都需要 N8N 才能自动运行。
在 5 美元的 DigitalOcean 小型服务器上自托管 N8N，为您提供了无限的工作流运行，没有每次执行的费用。
您需要的工作流类型：
用于计划操作 的 Cron 触发器。早上 6 点的晨间简报。周日晚上 8 点的每周审查。每月 1 号的月度报告。25 号的发票提醒。
用于基于事件的操作 的文件监视触发器。队列处理器监视 QUEUE 文件夹中的新文件。项目更新器监视项目文件夹的更改。
用于外部事件 的 Webhook 触发器。新的 Stripe 付款更新收入跟踪器。新的日历事件触发通话前简报。
每个由 Claude 驱动操作的 N8N 工作流结构遵循相同的模式：
触发 → 读取 CLAUDE.md → 读取相关的知识库内容
→ 使用编译的上下文调用 Claude API → 将输出写入
知识库 → 发送通知 → 记录操作
Claude API 调用始终包含 CLAUDE.md 内容作为系统上下文。这就是使每个自动化的输出都针对业务而非通用的原因。
## 系统运行后的日常体验
一旦知识库投入运行，这里是一个周二早晨的样子。
上午 5:58：您的手机收到 Telegram 通知。知识库中有四个项目准备好了。项目脉搏。内容简报。情报简报。您昨晚排队的一个研究简报。
上午 6:10：您在 4 分钟内阅读了项目脉搏。两个项目按计划进行。一个被标记为状态更新逾期。在查看电子邮件之前，您确切地知道需要注意什么。
上午 6:15：您打开内容简报。今天的三个角度。您选择一个并将其添加到内容队列文件中。
上午 8:45：您上午 9 点有一个客户通话。您打开在 8:30 自动生成的通话前简报。您花 5 分钟阅读。您走进会议室时的准备状态比手动准备 30 分钟还要充分。
一天中：您使用简单的约定在每日笔记中记录笔记、想法、完成情况和付款。所有内容都会自动路由到正确的地方。
下午 4:00：您今早排队的内容草稿正在 GENERATED/drafts 中等待。您花 8 分钟编辑。您安排它。
晚上 10:30：睡觉前...