Section 1
Your AI Forgets Everything
Every time you start a new Claude session, it remembers nothing about you. Your brand voice, your customer, your projects, your decisions, gone. You spend the first ten minutes re-explaining who you are before any real work happens.
The problem isn't the model. The model is smart. The problem is context. Without a place to store it, every session is day one.
Obsidian fixes this. It's a free note-taking app where every note is a plain markdown file on your computer. When you point Claude Code at your Obsidian vault, Claude can read your notes, your projects, and your context at the start of every session. No more re-briefing.
This guide gives you the one prompt that builds the whole thing.
Section 2
What You Need Before You Start
Three things. All free. Takes five minutes to get set up.
Obsidian
Download from obsidian.md. Free forever. Open it, create a new vault, and point it at an empty folder on your computer. Remember where that folder is.
VS Code or Cursor
Either works. Open it, use File, Open Folder, and point it at the same Obsidian vault folder you just created. This is how Claude will read and write to your notes.
Claude Code
Install instructions at docs.claude.com/claude-code. Once installed, open a Claude Code session in your vault folder.
Section 3
The Prompt
Copy everything in the block below. Paste it into a fresh Claude Code session inside your vault folder. Hit enter.
Claude will scaffold 9 folders, 15+ starter files, 5 templates, 4 navigation hubs, and one working skill, then interview you with 7 questions to fill in your brand voice, projects, and context. Total time: about 10 minutes, most of which is you answering questions.
Copy-paste prompt
You are setting up a complete Obsidian + Claude Code second brain for me from scratch. This vault will become the single source of truth for everything I do, projects, meeting notes, business context, ideas, daily logs, and Claude (you) will read from it at the start of every session.
Execute the phases below in order. Do not ask for permission before starting. Do not summarize the plan before building, just build. When you get to Phase 3, ask me the interview questions one at a time, waiting for each answer before moving on.
---
## Phase 1: Scaffold the directory structure
Create this exact folder and file tree in the current directory:
```
./
├── CLAUDE.md
├── USER.md
├── MEMORY.md
├── 00-Inbox/
│ ├── README.md
│ └── IDEAS.md
├── 01-MOCs/
│ ├── Hub.md
│ ├── Areas-MOC.md
│ ├── Projects-MOC.md
│ └── Content-MOC.md
├── 02-Reference/
│ └── README.md
├── 03-Templates/
│ ├── Daily-Template.md
│ ├── Weekly-Template.md
│ ├── Meeting-Template.md
│ ├── Project-Template.md
│ └── Reference-Template.md
├── 04-Areas/
│ └── README.md
├── 05-Projects/
│ └── README.md
├── 06-Daily-Logs/
│ └── README.md
├── 07-Journal/
│ └── README.md
├── 08-Archive/
│ └── README.md
└── .claude/
└── skills/
└── inbox-sweep/
└── SKILL.md
```
Create every folder. Write starter content into every file listed in Phase 2. Never create a folder without at least a README.md inside it.
---
## Phase 2: Populate starter files
### CLAUDE.md (vault root)
```
# [Your Vault Name], Claude Context
You are my AI thinking partner. This vault is your long-term memory. Read the files below at the start of every session.
## Session Start Protocol
1. Read USER.md, who I am, how I work, what not to do
2. Read MEMORY.md, persistent cross-session knowledge
3. Read 01-MOCs/Hub.md, current priorities and navigation
4. If I mention a specific project or area, read that area's _context/ folder BEFORE doing any work
## Vault Structure
| Folder | Purpose |
|--------|---------|
| 00-Inbox/ | Zero-friction capture. Raw ideas, half-thoughts, links to process later. |
| 01-MOCs/ | Maps of Content, navigation hubs. Start sessions here. |
| 02-Reference/ | External knowledge, research, articles, competitor material, saved resources. |
| 03-Templates/ | Recurring note templates (daily, weekly, meetings, projects). |
| 04-Areas/ | Each business or life area gets a folder with a _context/ subfolder containing Brand Voice, ICP, Positioning, etc. |
| 05-Projects/ | Active projects. One file per project with frontmatter status (active/paused/done). |
| 06-Daily-Logs/ | One note per day. Use 03-Templates/Daily-Template.md. |
| 07-Journal/ | Personal writing. Human-only, read, but never write here. |
| 08-Archive/ | Completed projects and stale material. Never delete, always archive. |
## Core Rules
1. Always search the vault before asking me. If I reference a project, meeting, decision, or document, grep/glob for it first. Only ask if you can't find it.
2. Never write in 07-Journal/. It's human-only. Read for context, never modify.
3. Never delete files. Move to 08-Archive/ with a date prefix instead.
4. Update MEMORY.md when you learn something durable, a preference, a decision, an ongoing project, a rule. Keep it tight.
5. When editing existing content to change tone, review EVERY line. Don't miss individual lines that contradict the requested change.
6. Draft first, plan second. When I ask for content, produce a draft immediately. Don't spend time on plans before delivering.
## Auto-loaded area context
When working in a specific area, load that area's _context/ at the start of the session. Example for a project in /04-Areas/my-business/:
- 04-Areas/my-business/_context/Brand-Voice.md
- 04-Areas/my-business/_context/ICP.md
- 04-Areas/my-business/_context/Positioning.md
## Skills
- inbox-sweep, run on demand to process the 00-Inbox/ folder. Triggered by "sweep the inbox" or "/inbox-sweep".
```
### USER.md (vault root)
```
# About Me
**Name:** [TBD from Phase 3 Q1]
**What I do:** [TBD from Phase 3 Q1]
**Primary goal right now:** [TBD from Phase 3 Q5]
## How I Work
- [TBD from Phase 3 Q6, working style preferences]
## What Claude Should ALWAYS Do
- Search the vault before asking me things
- Read the relevant _context/ folder before working on an area
- Update MEMORY.md when something durable changes
- Draft first, explain second
## What Claude Should NEVER Do
- [TBD from Phase 3 Q7]
- Write in 07-Journal/ (human-only)
- Delete files (archive with date prefix instead)
- Guess at context when it's documented in the vault
```
### MEMORY.md (vault root)
```
# Persistent Memory
This file is auto-updated as Claude learns durable things about me, my work, and my decisions. Never put drafts, temporary state, or one-off task context here.
## What to save here
- Preferences: how I like things done (confirmed by me, not assumed)
- Decisions: key choices and their reasoning
- Active projects: what's happening, not how
- Rules and constraints: things that don't change session-to-session
## Current memories
_(Empty. Entries will be added as Claude learns.)_
```
### 00-Inbox/README.md
```
# 00-Inbox, Zero-Friction Capture
Drop anything here that needs to be sorted later. This is the first stop, not the final destination.
## What goes here (root of 00-Inbox/)
- Voice memo dumps, raw transcripts, unedited
- Quick ideas (a phrase, a sentence, a question)
- Links to read or process later
- Anything you want to remember but don't have time to place
## What goes in IDEAS.md instead
- Developed ideas with rationale, why it's worth doing
- Anything that took more than 30 seconds to think through
## How to process
Run the inbox-sweep skill on demand. Say "sweep the inbox" or "/inbox-sweep" and Claude will read every file, classify it, ask questions about anything unclear, and move things to the right places after you approve the plan.
```
### 00-Inbox/IDEAS.md
```
# Ideas
A running list of developed ideas. Drop them here before they disappear.
_(Empty. First ideas will be added by you or via inbox-sweep.)_
```
### 01-MOCs/Hub.md
```
---
title: "Master Hub"
tags: [moc, master]
type: moc
updated: [today's date in YYYY-MM-DD]
---
# Master Hub
Entry point for every Claude session. Start here.
## Right Now
### Active Focus
- [TBD, update weekly or ask Claude to update based on session context]
### Active Projects
See [[Projects-MOC]]
### Open Decisions / Blockers
- None currently
## Navigation
| MOC | What it covers |
|-----|----------------|
| [[Areas-MOC]] | All business/life areas with their _context/ |
| [[Projects-MOC]] | Every active project with status + next action |
| [[Content-MOC]] | Anything you create, writing, videos, assets |
```
### 01-MOCs/Areas-MOC.md
```
# Areas, MOC
Each area is a distinct business, client, or life domain that deserves its own _context/.
## Active Areas
_(Populated during Phase 3 interview.)_
## The _context/ pattern
Every area has a _context/ folder with canonical docs:
- Brand-Voice.md, how this area communicates
- ICP.md, who this area serves (if applicable)
- Positioning.md, how this area is positioned against alternatives
- Products.md, what this area offers
- Manifesto.md, why this area exists
Single source of truth. Update once, every agent sees the change.
```
### 01-MOCs/Projects-MOC.md
```
# Projects, MOC
## Active
_(Populated during Phase 3 interview.)_
## Paused
_(none)_
## Done (recent)
_(none)_
## Convention
One file per project in 05-Projects/. Each file starts with frontmatter:
status, area, started, next action.
```
### 01-MOCs/Content-MOC.md
```
# Content, MOC
Anything you create, writing, videos, assets, social posts.
## Structure
Create subfolders inside 04-Areas/[area]/Content/ as you need them. Examples: Newsletter/, Short-Form/, Long-Form/, Social/.
Wikilink from this MOC to whatever matters most right now.
## Active
_(none yet)_
```
### 03-Templates/Daily-Template.md
```
---
date: {{date}}
type: daily
---
# {{date}}
## Priorities (3 max)
1.
2.
3.
## Notes
## Things to remember
## Tomorrow
```
### 03-Templates/Weekly-Template.md
```
---
week: {{date}}
type: weekly
---
# Week of {{date}}
## Wins
## Misses
## What did I learn
## Next week, top 3 priorities
1.
2.
3.
```
### 03-Templates/Meeting-Template.md
```
---
date: {{date}}
meeting: [title]
attendees: []
type: meeting
---
# [Meeting Title], {{date}}
## Context
## Discussion
## Decisions
## Action Items
- [ ]
```
### 03-Templates/Project-Template.md
```
---
status: active
area: [area name]
started: {{date}}
next: [next concrete action]
type: project
---
# [Project Name]
## Goal
## Current status
## Decisions made
## Open questions
## Next action
```
### 03-Templates/Reference-Template.md
```
---
source: [URL or citation]
captured: {{date}}
tags: [reference]
---
# [Title]
## Summary (your words)
## Key takeaways
1.
2.
3.
## Quotes worth saving
## Why I saved this
```
### 04-Areas/README.md
```
# 04-Areas
One folder per distinct area of your life/work. Each area has a _context/ subfolder that holds canonical docs (Brand-Voice, ICP, Positioning, Products, Manifesto), the single source of truth for that area.
When working in an area, Claude loads its _context/ first before doing any work.
Areas will be created during the Phase 3 interview.
```
### 05-Projects/README.md
```
# 05-Projects
One file per active project. Use 03-Templates/Project-Template.md as the starting format.
Each project file has frontmatter with: status (active/paused/done), area (which folder in 04-Areas/ it belongs to), started date, and next concrete action.
When you finish a project, move it to 08-Archive/ with a date prefix.
```
### 06-Daily-Logs/README.md
```
# 06-Daily-Logs
One note per day. Use 03-Templates/Daily-Template.md.
Filename format: YYYY-MM-DD.md
These are goldmines for Claude. It reads them to understand what you've been working on, what's been on your mind, and what's unfinished.
```
### 07-Journal/README.md
```
# 07-Journal
Personal writing. Human-only.
Claude reads this folder for context but NEVER writes to it. Free-form reflection, journaling, anything that isn't structured work content.
```
### 08-Archive/README.md
```
# 08-Archive
Completed projects and stale material.
Never delete files. Archive them here with a date prefix: 2026-04-14_project-name.md
This keeps your main folders clean while preserving history Claude can reference later.
```
### 02-Reference/README.md
```
# 02-Reference
External knowledge. Research, articles, competitor material, saved resources, SOPs, anything not tied to a specific project or date.
Best used with the Obsidian Web Clipper browser extension (obsidian.md/clipper), install it and anything you clip from the web lands here automatically.
```
### .claude/skills/inbox-sweep/SKILL.md
```
---
name: inbox-sweep
description: Processes the 00-Inbox/ folder on demand. Reads every raw note, classifies it, proposes a move plan, asks questions when unsure, then executes after approval. Never auto-deletes. Triggers on "sweep the inbox", "process my inbox", "organize inbox", "clean up the inbox", "check the inbox", "/inbox-sweep".
allowed-tools: Read, Write, Edit, Glob, Grep, Bash
---
# Inbox Sweep
Process everything in 00-Inbox/ (root only, skip subdirectories). For each file:
1. Read the content, never classify by filename alone.
2. Classify into one bucket:
- Developed idea, merge into 00-Inbox/IDEAS.md
- Reference, 02-Reference/
- Project planning, 05-Projects/
- Area-specific docs, 04-Areas/[area]/
- Unfinished draft, prepend a "> [!todo] UNFINISHED, [description]" callout, keep in inbox
- Stale/duplicate, propose 08-Archive/ with date prefix
- Unclear, STOP and ask a specific question
3. Show a table with every file plus proposed action.
4. List all questions separately.
5. Wait for approval before moving anything.
6. After execution, update 00-Inbox/README.md with a sweep log entry.
Never delete. Never touch 07-Journal/. Always use quoted paths (filenames may contain spaces).
```
---
## Phase 3: Interview me
Ask these questions ONE AT A TIME. Wait for my full answer before moving to the next. Update the relevant files as we go.
**Q1:** What's your name, and in one sentence, what do you do? (Goes into USER.md and the vault header.)
**Q2:** What should we name your vault? (e.g., "Second Brain", "Ops", "Your Name's Vault", this is the top-level header in CLAUDE.md.)
**Q3:** Name 1 to 3 distinct areas of your life or work that deserve their own folder. Examples: a business you run, a consulting client, a side project, "Personal". These become 04-Areas/[name]/ with a _context/ inside. I'll create a folder plus blank _context/ files for each one you list.
**Q4:** For your primary area (pick one from Q3), describe your brand, voice, communication style in 3 to 5 bullets. Goes into 04-Areas/[primary]/_context/Brand-Voice.md.
**Q5:** For that same primary area, who is the ideal customer, audience, or user? Goes into ICP.md.
**Q6:** What are 3 to 5 active projects you're working on RIGHT NOW? Just names are fine, I'll create a file per project in 05-Projects/ with the Project-Template.md applied. Add "next action" for any where you know it off the top of your head.
**Q7:** What's one thing you want Claude to NEVER do in your sessions? Goes into USER.md under "What Claude Should Never Do". Common examples: "don't use em dashes", "don't summarize what you just did at the end", "don't create files I didn't ask for", "don't guess, always ask if unsure".
---
## Phase 4: Populate from answers
As each answer comes in, immediately edit the relevant file. Don't batch edits until the end.
- Q1, fill in USER.md Name/What-I-do plus greeting line in CLAUDE.md
- Q2, replace [Your Vault Name] header in CLAUDE.md
- Q3, create 04-Areas/[name]/ folders, each with a _context/ subfolder containing blank stubs: Brand-Voice.md, ICP.md, Positioning.md, Products.md, Manifesto.md
- Q4, fill 04-Areas/[primary]/_context/Brand-Voice.md
- Q5, fill 04-Areas/[primary]/_context/ICP.md
- Q6, create one file per project in 05-Projects/ using Project-Template.md
- Q7, add to USER.md, "What Claude Should Never Do"
---
## Phase 5: Report and hand off
When all 7 questions are answered and every file is populated, output a summary:
```
Your second brain is live.
Built:
- [N] folders, [N] starter files
- [N] areas in 04-Areas/ with _context/ stubs
- [N] projects in 05-Projects/
- 1 skill (inbox-sweep)
What's still empty (go fill these in when you have 20 min):
- 04-Areas/[primary]/_context/Positioning.md
- 04-Areas/[primary]/_context/Products.md
- 04-Areas/[primary]/_context/Manifesto.md
- Any secondary areas from Q3
Next steps, in order:
1. Open Obsidian and point it at this folder. You'll see your vault.
2. Install the Obsidian Web Clipper browser extension (obsidian.md/clipper), now any YouTube video, article, or link saves straight into 02-Reference/.
3. For each session, start by saying "read 01-MOCs/Hub.md", Claude will load context.
4. Drop any unsorted idea into 00-Inbox/ and say "sweep the inbox" when it gets full.
5. (Optional power-up) Install Obsidian CLI (help.obsidian.md/cli) and QMD MCP (semantic search, by Tobi Lutke) for faster vault traversal. Not required for day one.
Your vault now has context Claude can read before you type a single word next session.
```
---
## Ground rules while executing
- If I give you a vague answer to an interview question, ask one specific clarifier and move on. Don't stall.
- If a folder already exists when you try to create it, that's fine, don't halt.
- Use relative paths throughout (e.g., 04-Areas/[name]/_context/), not absolute.
- All markdown files should have a trailing newline.
- When writing file content, use the Write tool with the full file contents in one shot, not incremental edits.
Pro tip: when Claude starts asking you the 7 questions, take them seriously. The quality of your answers determines how good your second brain is. Be specific.
Section 4
What You're Building
Here's what the prompt scaffolds. Every folder has a purpose. Every file has a reason.
./
├── CLAUDE.md # Main brain, read every session
├── USER.md # About you and your rules
├── MEMORY.md # Persistent cross-session memory
├── 00-Inbox/ # Zero-friction capture
├── 01-MOCs/ # Navigation hubs (Hub, Areas, Projects, Content)
├── 02-Reference/ # External knowledge and clippings
├── 03-Templates/ # Daily, weekly, meeting, project, reference
├── 04-Areas/ # Your businesses or life areas, each with _context/
├── 05-Projects/ # Active projects, one file each
├── 06-Daily-Logs/ # One note per day
├── 07-Journal/ # Personal writing, human-only
├── 08-Archive/ # Completed work, never deleted
└── .claude/
└── skills/
└── inbox-sweep/ # Your first working skill The _context/ pattern is the killer move. Each area you create has a _context/ subfolder with canonical docs: Brand-Voice, ICP, Positioning, Products, Manifesto. Update the file once, every Claude session sees the change.
The inbox-sweep skill is your first taste of what Claude skills can do. Drop half-baked ideas into 00-Inbox/, and when it fills up, say "sweep the inbox". Claude reads every file, asks about anything unclear, and moves things to the right place after you approve the plan.
Section 5
The 7 Questions
After scaffolding, Claude asks you these one at a time. Your answers populate the vault with real context, not placeholder text. Think about them before you start so you're not caught off guard.
1. Name and role
Who you are, one sentence. Populates USER.md.
2. Vault name
"Second Brain", "Ops", your name, whatever. Top header of CLAUDE.md.
3. Your 1 to 3 areas
Distinct parts of your work or life. A business you run, a consulting client, a side project, "Personal". Each gets its own _context/.
4. Brand voice for your primary area
3 to 5 bullets describing how you communicate. This is what makes Claude sound like you.
5. Ideal customer or audience
Who you serve in your primary area. The more specific, the better the output.
6. 3 to 5 active projects
What you're actively working on right now. Each becomes a file in 05-Projects/.
7. One thing Claude should never do
Your biggest Claude pet peeve. "Don't use em dashes." "Don't summarize what you just did." "Don't create files I didn't ask for." Baked into USER.md.
Section 6
Optional Power-Ups
Once your vault is running, three things make it even better. None are required for day one.
Obsidian Web Clipper
Browser extension. Install from obsidian.md/clipper. Any YouTube video, article, or link you find, one click, it lands in your 02-Reference/ folder automatically. Claude can now read any article you've ever saved.
Obsidian CLI
Built-in command-line interface. Lets Claude read, search, and write to your vault faster than it could by opening files one by one. Docs at help.obsidian.md/cli.
QMD MCP (semantic search)
Built by Tobi Lutke, the Shopify founder. Adds meaning-based search to your vault. Now Claude can find a note about "pricing decisions" even if the file is titled "Revenue Strategy Meeting". Runs locally, no API calls. Worth adding once your vault has more than a few dozen notes.
That's it
Your AI Now Has Context
You close your laptop with a second brain Claude can read before you type a single word next session. Every project, every decision, every piece of business context lives in one place.
Start each session by asking Claude to read 01-MOCs/Hub.md. That loads your priorities, active projects, and the rest of the vault map. From there, it knows where to look.
Drop raw ideas into 00-Inbox/ all week. When it fills up, say "sweep the inbox". Watch Claude move things into place.
The more you use it, the more your AI knows you. That's the whole game.
Zero to Automated
Get the full AI operating system
Guides are a great start. But inside Zero to Automated, you get a done-for-you AI system deployed within 48 hours, plus everything you need to keep building:
- Claude Code 101 — 8 lessons, zero to automating real work by Sunday
- Ready-to-install Claude Skills — a growing library with 2+ new skills every week
- Done-for-you workflows — copy, paste, run. New ones every week.
- Weekly live builds — watch real projects get built, ask questions, follow along
- Small group coaching — direct access to both founders
Price locked for life at whatever you join at.