# Strategy Engine Setup — Build Your Second Brain > Hi Claude. The person who handed you this file is building their Second Brain — the home base their AI will think from. > > Read this whole file first. Then ask them the questions, show them the plan, and build it. Don't build anything until they say "go." --- ## A quick word before we start (read this part out, plainly) This is the first of three engines we install: - **Strategy Engine (Stage 1)** — your Second Brain. A simple, organised home for your notes, projects, and life — the strategy around how everything is filed. We build this today, once. (That's this file.) - **Self Engine (Stage 2)** — a working model of how you think. We teach the AI who you are. (That's the next file.) - **Systems Engine (Stage 3)** — the AI workflows you layer on top later (content, assets, daily playbooks, and more). Today we're installing the Strategy Engine. Build it once, benefit forever. --- ## How this works (for the person setting up) You don't need to be technical. You don't need to know what a "terminal" is. If you can chat with an AI — which you're doing right now — you can do this. You're talking to me inside **Claude Code** (the Code tab in the Claude desktop app). I can create folders and files for you. The folder you opened me in becomes your Second Brain. I'll ask you five short questions, then build the whole thing in front of you. If a word ever sounds technical, I'll explain it in one plain sentence. And any time you're unsure or something doesn't feel right, just ask me a question — that's what I'm here for. Nothing in this setup is one-way; we can always fix or change it together. --- ## What I'm going to build A **6-folder Second Brain** inside the folder you opened me in, shaped around your real projects and life. Then I'll write a few "instruction manual" files so future-me always knows how your brain works. The 6 folders never change: | # | Folder | What it holds | |---|---|---| | 00 | `00. System/` | The instruction manual (I read these so I know how to help you) | | 01 | `01. Inbox/` | Anything you capture fast. Sort it later. | | 02 | `02. Projects/` | Work with a finish line — a launch, a client, a book | | 03 | `03. Areas/` | Ongoing parts of life — Health, Family, Money, work pillars | | 04 | `04. Resources/` | Reference you'll reuse — articles, templates, swipe files | | 05 | `05. Archive/` | Done and parked. Out of sight, still searchable. | ### A short note on where this shape comes from This is loosely based on **PARA** — Projects, Areas, Resources, Archive — a system created by Tiago Forte. I'll be honest: the exact setup matters more to your AI than it does to you. You don't have to love PARA or even understand it deeply. I use it because the people I learned from use it, it's clean, and the AI works well inside it. If you ever want to change the shape later, you can — nothing here is locked. For now, we model after it because it works. --- ## How to behave during setup (Claude, this is for you) - **One question at a time.** Wait for the answer. Never stack questions. - **Plain English.** No "configure," "directory," "instantiate." Say "folder," "file," "settings." - **Always offer a default.** If they hesitate or say "I don't know," give a starter set they can rename later. - **Confirm before building.** Show the plan. Wait for "yes." - **No surprise files.** Build only what's listed here. Don't add extras without asking. - **Match their language.** If they answer in another language, accept it. - **When unsure, ask — don't guess.** If an answer is unclear or you're not sure where something belongs, ask them a short question to rectify it. Better to check than to build the wrong thing. - **No emojis** in any file or folder you create. --- ## The setup (run in this order) ### Step 1 — Say hello Say something close to: > "Good to meet you. The folder you've opened me in is about to become your Second Brain — the home base I'll think from every time we work together. Before I build anything, I want to understand you and your work a little: five quick questions, then I'll show you the plan and build it while you watch. Shall we?" Wait for "yes" / "ready." If they'd rather do something else first, ask what. ### Step 2 — Quiet safety checks (do these silently, only speak up if something's off) **2a. Check where you are.** Run `pwd` (Mac) or `Get-Location` (Windows PowerShell). Show them: > "Quick check — I'm in: `[path]`. This is where I'll build your brain. Right place?" If the path is a big shared folder — exactly `/`, `~/`, `~/Desktop`, `~/Documents` (the top level, not a sub-folder), or `C:\` / `C:\Users\[name]` — push back: > "Heads up — that's a top-level folder, not a dedicated one. If I build here I'll mix your brain into your whole Desktop or Documents. Better to make a fresh folder first. Want me to make one called `my-second-brain` inside your Documents and switch into it?" **2b. Check the folder is empty enough.** List what's there: - Mac: `ls -a` - Windows PowerShell: `Get-ChildItem -Force` Ignore housekeeping files (these don't count as "not empty"): `.DS_Store`, `.localized`, `.git`, `.gitignore`, `.obsidian`, `Thumbs.db`, `desktop.ini`. If there are already PARA-style folders, great — rename/fill the gaps so it ends as the 6 folders above. If there are other files, ask: > "There are already files here: [list]. Did you mean to set up in this folder? If yes, I'll work around them. If no, let's switch to an empty one first." **2c. Check it's not inside a cloud sync folder.** Look at the full path. If it contains any of these (case doesn't matter): `iCloud`, `com~apple~CloudDocs`, `Mobile Documents`, `Dropbox`, `Google Drive`, `GoogleDrive-`, `OneDrive`, `Library/CloudStorage`, `Box Sync`, `Box.com`, `pCloud`, `Proton Drive`, `Sync.com` …then say gently: > "Heads up — this folder lives inside [service]. Those services copy your files in the background, which can fight with how I edit them, and a few weeks in you'll see weird duplicates. Easier to fix now. Want me to move it to a normal local folder in your Documents?" Default to moving. Don't push on if they insist on staying without a clear "continue." (On Windows inside WSL, resolve the Windows path with `wslpath -w "$(pwd)"` and check that too.) If any check stops the flow and they fix it, start again from 2a. ### Step 3 — Their first name > "What's your first name?" If they'd rather not say, use "there." Save as `USER_NAME`. ### Step 4 — What they do > "In one sentence — what do you do? A job, a business, or what you're building." If stuck, default to: "Working on something I'll define later." Save as `USER_ROLE`. ### Step 5 — How they want you to talk > "How do you like me to talk to you — short and direct, friendly and conversational, or detailed and thorough?" Default: short and direct. Save as `USER_STYLE`. ### Step 6 — Their active projects (the stuff with a finish line) > "Now your active projects — anything with a finish line. A launch, a client, a book, a course. Give me 3 to 7. Just describe them, I'll turn them into tidy folder names." If they already organise work inside **Claude Projects or ChatGPT Projects**, mention it: > "If you already keep work inside Claude Projects or ChatGPT Projects, those count too — name them here and we'll mirror them as folders, so your brain and your chat tools line up. Use the same names in both and it stays easy to follow." If they freeze: > "No stress. I'll start you with `Project-A`, `Project-B`, `Project-C`. Rename them when you know." Turn each into a folder-safe name: trim spaces, swap spaces for hyphens, drop emojis and odd characters, keep normal letters (including accents/other languages where the computer allows). Then show your cleanup: > "Quick clean-up — folders work best without spaces or symbols, so I turned those into: [list]. Look right?" Save as `USER_PROJECTS`. ### Step 7 — Their life areas (the stuff that never ends) > "Now your life areas — the ongoing parts that don't finish. Health, Family, Faith, Money, plus a couple of work pillars like Marketing, Sales, or Operations. Give me 3 to 8." If they freeze: > "Default starter: Health, Family, Work, Money, Learning. We can refine later." Same cleanup as Step 6. Confirm the names. Save as `USER_AREAS`. ### Step 8 — Show the plan > "Here's the shape of what I'm about to build — six top folders, plus your projects and areas inside:" ``` {their folder name}/ ├── CLAUDE.md ├── AGENTS.md (same manual, so Codex works too) ├── 00. System/ │ ├── README.md │ ├── TAG-TAXONOMY.md │ ├── VAULT-MAP.md │ ├── LINKING-RULES.md │ └── AUTOMATIONS.md ├── 01. Inbox/ │ ├── welcome.md │ └── agent-drops/ │ └── README.md ├── 02. Projects/ │ ├── {Project1}/ │ ├── {Project2}/ │ └── ... (all from Step 6) ├── 03. Areas/ │ ├── Self/ (the Self Engine fills this later) │ ├── {Area1}/ │ ├── {Area2}/ │ └── ... (all from Step 7) ├── 04. Resources/ │ └── AtomicNotes/ └── 05. Archive/ ``` > "That's the plan. Should I build it?" Wait for "yes." ### Step 9 — Build the folders Create the 6 top folders with their exact names (including the period and space — `00. System`, `01. Inbox`, and so on). - Inside `01. Inbox/` make an `agent-drops/` folder. - Inside `02. Projects/` make one folder per name in `USER_PROJECTS`. - Inside `03. Areas/` make one folder per name in `USER_AREAS`, plus an empty `03. Areas/Self/` folder (the Self Engine setup fills it later). - Inside `04. Resources/` make an empty `AtomicNotes/` folder (imported notes land here later). If making a folder fails, don't quietly retry. Explain in plain English: - Mac "Operation not permitted" → "Your Mac is blocking writes here. Easiest fix: move the folder into Documents and run me again." - Windows "Access is denied" → "Windows is blocking writes here. Move the folder into Documents and run me again." - "No space left" → "The drive is full. Free up some space and run me again." ### Step 10 — Write the instruction-manual files Get today's date first: - Mac: `date +%Y-%m-%d` - Windows PowerShell: `Get-Date -Format yyyy-MM-dd` Save as `TODAY_DATE`. Then create these files, replacing every `{{...}}` with the answers from above. The mapping is: `{{USER_NAME}}` to their name, `{{USER_ROLE}}` to what they do, `{{USER_STYLE}}` to how they want you to talk, `{{LIST_OF_PROJECTS_AS_BULLETS}}` to `USER_PROJECTS` written as a markdown bullet list (one `- name` per line), and `{{LIST_OF_AREAS_AS_BULLETS}}` to `USER_AREAS` as the same kind of bullet list. Don't leave any `{{...}}` token in a finished file. #### 10a. `CLAUDE.md` (at the root of the brain, NOT inside `00. System/`) ```markdown # CLAUDE.md — Operating Manual for {{USER_NAME}}'s Second Brain > Hi Claude. Open this first whenever we start. It's the quickest way for you to know who I am and how I like to work in here. ## Who I am - **Name:** {{USER_NAME}} - **What I do:** {{USER_ROLE}} ## How to talk to me - **Style:** {{USER_STYLE}} - Get to the point. A few clear bullets beat a wall of text. - Push back when I'm off — I'd rather hear it than be flattered. - Skip the warm-up and the wrap-up. Give me the answer. - No emojis unless I ask for them. ## How my brain is organised - `00. System/` — the instruction manual (you read these) - `01. Inbox/` — fast captures, sorted later. Automated captures land in its `agent-drops/` sub-folder. - `02. Projects/` — work with a finish line - `03. Areas/` — ongoing parts of life - `04. Resources/` — reference I reuse - `05. Archive/` — done and parked ## My active projects {{LIST_OF_PROJECTS_AS_BULLETS}} ## My life areas {{LIST_OF_AREAS_AS_BULLETS}} ## Hard rules - When you save to `01. Inbox/`, use the `agent-drops/` sub-folder. The Inbox root is for my own captures. - When I'm working with you live, show me what you're about to create before creating it. - If you're unsure where something belongs, file it anyway and flag it in your log rather than stopping. - Don't invent files I didn't ask for, and never invent facts, quotes, or themes that aren't in my notes. ## My daily and weekly routines These two run on a schedule, and I can also trigger them by hand any time. When they run on a schedule there's no one watching, so don't wait for my approval. Do the full job, then write a short log I can check later, and flag anything you were unsure about instead of stopping. Whenever a step below uses `{date}`, replace it with today's date as YYYY-MM-DD, fetched fresh on each run (Mac: `date +%Y-%m-%d`, Windows PowerShell: `Get-Date -Format yyyy-MM-dd`). On a scheduled run, never delete anything: if something looks like a duplicate or you're unsure, move it to `05. Archive/` instead, so nothing is ever lost. **When I say `sort my inbox`:** 1. List everything in `01. Inbox/` and its `agent-drops/` sub-folder. Skip `welcome.md` and any hidden files. 2. Read each item enough to know what it is. 3. For each one, decide its right home across `02. Projects/`, `03. Areas/`, `04. Resources/`, or `05. Archive/`. Give it the right tags from `00. System/TAG-TAXONOMY.md` and add at least two `[[links]]` to related notes. 4. If an item is rich enough to stand on its own, distill it into a clean note in my words, not a raw dump. 5. Move each item to its home and change its status tag from `#status/inbox` to `#status/organized`. Log every move as a one-line summary to `01. Inbox/agent-drops/sort-log-{date}.md`, with anything you weren't sure about listed at the bottom. **When I say `review my week`:** 1. Look at everything that landed or changed across the brain in the last 7 days. 2. Pull out the recurring themes, patterns, and lessons. 3. Turn each clear cluster into a short reference article in `04. Resources/`: a clean, titled page in my voice that I can search and re-read. Link each article back to the notes it came from. 4. Tidy the obvious mess: notes still tagged `#status/inbox`, missing links, near-duplicates. 5. Write a one-page summary of what you found and what you wrote to `01. Inbox/agent-drops/weekly-review-{date}.md`. Only write what's genuinely in my notes. If a theme isn't really there, leave it out. ## Files worth reading - `00. System/TAG-TAXONOMY.md` — the tags this brain uses - `00. System/VAULT-MAP.md` — what each folder is for - `00. System/LINKING-RULES.md` — how to connect notes - `00. System/AUTOMATIONS.md` — the daily and weekly helpers that tend this brain ## Privacy Nothing here shares anywhere automatically. Everything stays on my computer. --- This file grows as my patterns get clearer. Edit it any time. ``` #### 10a-2. `AGENTS.md` (at the root, next to `CLAUDE.md`) Some AI tools read a file called `CLAUDE.md`, and others (like OpenAI's Codex) read one called `AGENTS.md` instead. To make this brain work in either tool, write a second manual at the root called `AGENTS.md`. Keep `CLAUDE.md` as the single source of truth, and make `AGENTS.md` point to it. Copy the two routines in directly though, so a scheduled Codex run still does the full job even if it doesn't open the other file first. ```markdown # AGENTS.md — Operating Manual for {{USER_NAME}}'s Second Brain > This is the same second brain. If you're Codex, or any tool that reads AGENTS.md, start here. My full manual is `CLAUDE.md` in this same folder. Read it first and follow all of it: who I am, how to talk to me, how the brain is organised, my projects and areas, and my hard rules. Treat `CLAUDE.md` as the source of truth. If this file and that one ever disagree, `CLAUDE.md` wins. ## My daily and weekly routines These two run on a schedule, and I can also trigger them by hand any time. When they run on a schedule there's no one watching, so don't wait for my approval. Do the full job, then write a short log I can check later, and flag anything you were unsure about instead of stopping. Whenever a step below uses `{date}`, replace it with today's date as YYYY-MM-DD, fetched fresh on each run (Mac: `date +%Y-%m-%d`, Windows PowerShell: `Get-Date -Format yyyy-MM-dd`). On a scheduled run, never delete anything: if something looks like a duplicate or you're unsure, move it to `05. Archive/` instead, so nothing is ever lost. **When I say `sort my inbox`:** 1. List everything in `01. Inbox/` and its `agent-drops/` sub-folder. Skip `welcome.md` and any hidden files. 2. Read each item enough to know what it is. 3. For each one, decide its right home across `02. Projects/`, `03. Areas/`, `04. Resources/`, or `05. Archive/`. Give it the right tags from `00. System/TAG-TAXONOMY.md` and add at least two `[[links]]` to related notes. 4. If an item is rich enough to stand on its own, distill it into a clean note in my words, not a raw dump. 5. Move each item to its home and change its status tag from `#status/inbox` to `#status/organized`. Log every move as a one-line summary to `01. Inbox/agent-drops/sort-log-{date}.md`, with anything you weren't sure about listed at the bottom. **When I say `review my week`:** 1. Look at everything that landed or changed across the brain in the last 7 days. 2. Pull out the recurring themes, patterns, and lessons. 3. Turn each clear cluster into a short reference article in `04. Resources/`: a clean, titled page in my voice that I can search and re-read. Link each article back to the notes it came from. 4. Tidy the obvious mess: notes still tagged `#status/inbox`, missing links, near-duplicates. 5. Write a one-page summary of what you found and what you wrote to `01. Inbox/agent-drops/weekly-review-{date}.md`. Only write what's genuinely in my notes. If a theme isn't really there, leave it out. ## Hard rules - When you save to `01. Inbox/`, use the `agent-drops/` sub-folder. The Inbox root is for my own captures. - When I'm working with you live, show me what you're about to create before creating it. - If you're unsure where something belongs, file it anyway and flag it in your log rather than stopping. - Don't invent files I didn't ask for, and never invent facts, quotes, or themes that aren't in my notes. ## Running me on a schedule with Codex When you set up a scheduled task with Codex, run it non-interactively and fully on its own, for example: `codex exec --sandbox workspace-write --ask-for-approval never "sort my inbox"` That lets it sort and write files without stopping to ask. Use the same shape with `review my week` for the weekly one. --- The full picture lives in `CLAUDE.md`. Read it. ``` #### 10b. `00. System/README.md` ```markdown # 00. System — Start Here Think of this folder as the brain's owner's manual. These are the files Claude opens first so it knows how everything works. ## What's in here - `CLAUDE.md` — at the root (one level up). The main manual. - `README.md` — this file. - `TAG-TAXONOMY.md` — the tags this brain uses. - `VAULT-MAP.md` — what each folder is for, and what not to put in it. - `LINKING-RULES.md` — how to connect notes so the brain stays searchable. - `AUTOMATIONS.md` — recurring helpers (not running yet — we switch these on later). ## Don't dump notes in here System files only. Captures go in `01. Inbox/`. ``` #### 10c. `00. System/TAG-TAXONOMY.md` ```markdown # Tags A small starter set. Add more later — keep them few and consistent. ## Status — where is this note in its life? - `#status/inbox` — just captured, not sorted yet - `#status/organized` — filed in the right folder - `#status/distilled` — refined into something genuinely useful ## Type — what kind of note? - `#type/note` — general - `#type/idea` — half-formed thought - `#type/decision` — a decision and the reasoning behind it - `#type/meeting` — meeting or call notes - `#type/output` — something I shipped (post, email, doc) - `#type/atomic` — one self-contained fact per note - `#type/profile` — a note about me (my values, voice, story) ## Source — where did it come from? - `#source/web` — read it online - `#source/conversation` — heard it from someone - `#source/book` — from a book - `#source/agent` — produced by an AI helper - `#source/chat-export` — pulled from my imported chat history - `#source/user-statement` — I said this in a chat (my words) - `#source/model-statement` — the AI said this (NOT my words) - `#source/user-belief` — a confirmed belief of mine - `#source/calibration` — from a Self Engine question session (my own answers) That's enough for now. ``` #### 10d. `00. System/VAULT-MAP.md` ```markdown # Where Things Live A quick guide to which folder gets what — and the mistake to avoid for each. ## 00. System The files that explain how the brain runs. Keep it to that — no notes or working files in here. ## 01. Inbox Where anything lands the moment I capture it, from anywhere. It's a staging area, not a home — clear it out regularly. Its `agent-drops/` sub-folder catches anything an automated helper saves, so I rarely open it by hand. ## 02. Projects Anything with a finish line lives here. If it never really ends, it's not a project — it belongs in Areas. ## 03. Areas The ongoing parts of my life and work — Health, Family, Money, and my work pillars. If it has a deadline, it's a project, not an area. ## 04. Resources Reference I'll come back to and reuse. My own finished work doesn't go here — that lives with its Project or Area. ## 05. Archive Done projects and retired areas, parked but still searchable. This isn't a bin, though — if something's genuinely useless, just delete it. ``` #### 10e. `01. Inbox/agent-drops/README.md` ```markdown # agent-drops/ Automated captures land here. When you build AI helpers later (capture bots, schedulers), point them at this folder. You don't write here by hand. Treat it as a holding area — a regular sort moves things out into Projects, Areas, or Resources. ``` #### 10f. `00. System/LINKING-RULES.md` ```markdown # Linking Rules Tags tell you what a note is. Links tell you what it relates to. You want both. ## The minimum Once a note is filed, connect it to at least 2 others. A note with no links is an island — easy to store, hard to ever think with again. ## What to link - People → their note in `03. Areas/` - Projects → the project folder in `02. Projects/` - Ideas you've defined before → link instead of redefining - Sources you cite → the source note in `04. Resources/` ## How to write a link Double brackets: `[[note-name]]` or `[[note-name|words you want shown]]`. If the note doesn't exist yet, the link still works — your viewer will offer to create it when clicked. ## What not to link - Don't link common words for the sake of it. - Don't link inside the top settings block (frontmatter) — keep links in the body. ``` #### 10g. `00. System/AUTOMATIONS.md` ```markdown # Automations — The Helpers That Tend This Brain This brain is built to be tended by simple recurring helpers. The two main ones are defined in full in the root `CLAUDE.md`, under "My daily and weekly routines." That's the source of truth for exactly what each one does. This file is the quick map. --- ### 1. Daily Inbox Sort — ACTIVE (command: `sort my inbox`) - What: routes everything in `01. Inbox/` to the right folder, tags it, links it, distills the rich ones, and logs every move. - When: daily, on a schedule, or by hand any time. - Full spec: see `CLAUDE.md` → "My daily and weekly routines." ### 2. Weekly Review — ACTIVE (command: `review my week`) - What: reads everything new from the last 7 days, pulls the recurring themes into clean reference articles in `04. Resources/`, tidies loose ends, and writes a summary. - When: weekly, on a schedule, or by hand any time. - Full spec: see `CLAUDE.md` → "My daily and weekly routines." ### 3. Chat-History Import — ON DEMAND (command: `add this to my brain`) - What: pulls past AI chats, notes, and documents into small notes and refreshes my Self Engine. This is what the Self Engine setup runs. --- ## How these run - Both routines are built to run on their own on a schedule, without stopping to ask permission. They do the work and leave a log in `01. Inbox/agent-drops/` so I can check what happened. - I can also just say `sort my inbox` or `review my week` any time to run one on the spot. - Always follow the full spec in `CLAUDE.md`, not a shortcut version. ``` ### Step 11 — Drop a first note Create `01. Inbox/welcome.md`: ```markdown --- tags: [status/inbox, type/note] created: {{TODAY_DATE}} --- # Welcome to your Second Brain This is your first note. Captured straight into the Inbox — that's the rhythm. Try this next time you open Claude here: > "Read CLAUDE.md and tell me what you understand about my brain." When that works, drop a real note in `01. Inbox/` and start filling it. ``` ### Step 12 — Show the result > "Done. Here's what I built: > > - 6 top folders (`00. System` through `05. Archive`) > - {N} project folders: {names} > - {M} area folders (plus a `Self/` spot the Self Engine fills later): {names} > - An `AtomicNotes/` spot in Resources for imported notes > - Your instruction-manual files in `00. System/`, plus `CLAUDE.md` at the root > - One test note in your Inbox > > Try this: ask me 'what do you see in my brain?' — I'll read CLAUDE.md and confirm. > > Next: your **Self Engine** — teaching me who you are. That's the next file." --- ## If something's already here (re-run) If, when you start, any of these exist — root `CLAUDE.md`, any of the 6 folders, or `00. System/INSTALL-PROGRESS.md` — tell them first: > "Looks like a setup was started here before — I can see [list]. Two options: > - (a) Start fresh. I'll back up the old files to `00. System/old-install-{{TODAY_DATE}}/`, then build clean. > - (b) Fill in the gaps. I'll add only what's missing and leave your files alone. > > Which one?" If `INSTALL-PROGRESS.md` exists, prefer (b) and pick up where they left off. If they stop halfway, save what you have to `00. System/INSTALL-PROGRESS.md` with the date and their answers, and tell them: "When you're ready, hand me this file again — I'll resume." --- ## What NOT to do - This is folders and text files, nothing else — don't install or download any software. - Don't touch logins, connect accounts, or change any computer settings. - Build only the folders they named. Resist adding your own clever extras (`Templates/`, `Daily/`, and the like). - Leave Project and Area folders empty for them to fill — no placeholder or tutorial content inside. - Keep file and folder names clean. The numbered prefix is the only decoration — no emojis. --- That's the Strategy Engine. When they're ready, start at Step 1.