Phygital Brain: Setup

Version 0.1.12. Everything below is for the agent. Follow it step by step, and check each step before moving on.

What you are building

A second brain: a folder of notes that is both the person's knowledge base and your operating system. You show four short forms, then build the folders, the rule files (CLAUDE.md), their first notes and the Brain Keeper. When a new chat opens in this folder later, you read those rules and follow them.

Files you use, all in this skill's folder:

The Brain Keeper lives in the next skill: ../brain-keeper/SKILL.md.

Principles

• One window per step. Each form is one window: the person clicks, types short answers, drops files and presses one button. You get all the answers at once. Then the next form opens. No chat between forms.
• About 10 minutes. No follow-up questions inside a form. Accept whatever they give. Skipped questions stay empty.
• Adapt. Skip what you already know. Skip business questions for people who don't sell anything.
• Real content only. Never write placeholders like [name] or TBD. No information for a section: leave the section out. No information for a file: don't create it.
• Their words. Keep their exact names, numbers and phrases.
• Plain language. Short sentences, everyday words, no jargon, in every question and every file.
• "Your agent", never a product name, in every question and message the person sees.
• Never use em dashes in anything you write. Use periods, commas or colons.
• Work quietly. Don't narrate each file. One short message at the end.

Which agent you are

The brain's rule files are CLAUDE.md. If you are Claude, read them as they are.

If you are another agent and your tool loads a different rule file (for example AGENTS.md in Codex), then at the end of the build, next to each CLAUDE.md, add that file with one line: Read CLAUDE.md in this folder and follow it. One set of rules, never two copies. Skip this if your tool already reads CLAUDE.md.

Step 0: Checks

1. Confirm you can write to the selected folder. No folder selected: ask them to pick their brain folder (the one inside their cloud drive, see the Getting started guide) and send the message again.

2. If a CLAUDE.md exists in the folder root, read its first line:

   • <!-- Phygital Brain, version X, language: xx -->: a Phygital Brain is already here. Ask one question (header Setup): Upgrade my brain (only if X is older than this skill's version, goes to Step U) · Update my answers (keep everything, redo the forms, refresh the About files) · Cancel.
   • Any other CLAUDE.md: this folder already has rules from something else. Don't build here. Ask them to pick an empty folder for their brain.

   Never delete or overwrite existing notes.

3. If the folder has other files, don't touch them. Build around them and mention them at the end.
4. The cloud drive comes in Step 1.6, after the language question, so it's asked in the person's language. Add-ons and Phygital Memory come last, in Steps 4.1 and 4.2.
5. Form window. Check whether you can show an HTML form window (in Claude, the show_widget tool). Remember the answer. If you can't, use forms/questions.md in Step 2: click questions where you have them (AskUserQuestion, up to 4 questions per call), otherwise numbered options in chat. On a phone, never use form windows, even if you have the tool: they don't work well there. Use click questions for every step, from the language question on, with the texts and options from the forms (for a ready language, from i18n/{code}/forms/).

Step 1: Language

With a form window, show forms/language.html (made by scripts/i18n.py build), adapted as its first comment says. It shows the main languages as cards; every other ready language below is typed in Other and counts the same as a card. Its Chinese card (zh) leads to the Simplified or Traditional question below. Without one, or on a phone, ask one question with click options, in the language of their first message (English if unclear): "Which language should your brain speak?" The ready languages, each with its code:

English (en) · 简体中文 (zh-CN) · 繁體中文 (zh-TW) · हिन्दी (hi) · Español (es) · العربية (ar) · Français (fr) · বাংলা (bn) · Português (Brasil) (pt-BR) · Português (Portugal) (pt-PT) · Русский (ru) · اردو (ur) · Bahasa Indonesia (id) · Deutsch (de) · 日本語 (ja) · Tiếng Việt (vi) · Türkçe (tr) · 한국어 (ko) · Italiano (it) · Polski (pl) · Nederlands (nl) · فارسی (fa) · ไทย (th) · Українська (uk) · Kiswahili (sw) · தமிழ் (ta) · తెలుగు (te) · मराठी (mr) · ગુજરાતી (gu) · ਪੰਜਾਬੀ (pa) · Bahasa Melayu (ms) · Filipino (fil) · עברית (he) · Ελληνικά (el) · Čeština (cs) · Română (ro) · Magyar (hu) · Svenska (sv) · Dansk (da) · Norsk (bokmål) (nb) · Suomi (fi) · Other

If the click tool takes only 4 options, offer the 3 most likely from their first message plus Other. If their first message is in Portuguese, two of those are Português (Brasil) and Português (Portugal); if it's in Chinese, 简体中文 and 繁體中文. A ready language typed in counts the same as a click, whether they type its own name, its English name (German, Japanese) or its code (de, ja).

Two kinds of Portuguese, two kinds of Chinese. Never guess between them. If they pick or type just "Portuguese" or "Português" (no country), ask one click question: Português do Brasil · Português de Portugal. If they pick or type just "Chinese" or "中文", ask: 简体中文 · 繁體中文. A country or region they name decides it without asking: Brazil gives pt-BR; Portugal, Angola, Mozambique or another African Portuguese-speaking country gives pt-PT; mainland China, Singapore or Malaysia gives zh-CN; Taiwan, Hong Kong or Macau gives zh-TW.

A ready language (a folder i18n/{code}/ exists in this skill)

The brain is built fully in that language, once, from files that are already translated. Don't translate anything yourself.

• Forms: use i18n/{code}/forms/ instead of forms/.
• Templates: use i18n/{code}/templates/ instead of templates/, and pass i18n/{code}/keeper/ to the Brain Keeper (3.8) as its templates.
• Messages: say the setup's messages from i18n/{code}/strings.json (messages), and use its features names for the Brain Keeper, morning brief and the rest.
• Folder and note names: i18n/{code}/strings.json (paths) maps every English name to its name in that language. Wherever this file names a brain path in English (About/me.md, Skills/brain-keeper/, [[goals]]), use the mapped names. The templates in i18n/{code}/ already use them.
• Folder map: at the end of the root CLAUDE.md, add a ## Folder map section: a two-column table, English name and the brain's name, for every entry in paths, with one line above it: "For Phygital Brain upgrades and skills only. Daily work uses the names in this brain." This is the only place the English names appear in the brain.

Any other language

The brain is still built fully in their language, folder and note names included. You translate it yourself, while you build it, following translating.md in this skill's folder: read it now. Don't ask anything about the translation; choose every word yourself.

• Forms, templates and messages: use the English forms/ and templates/, and translate each one as translating.md says.
• Brain Keeper: pass the English ../brain-keeper/templates/ to the Brain Keeper (3.8), and tell it to translate them the same way, with your names.
• Folder map: add it at the end of the root CLAUDE.md, as translating.md says.
• In the finish message (Step 4), add one line: "Your agent translated your brain during the setup. To make a checked set that others can use too, say 'translate Phygital Brain into my language'."

In every language

These stay in English, so the rules and the Brain Keeper work the same everywhere:

• The file names CLAUDE.md, README.md and .claudeignore, and the .backups/ and Skills/no-ai-slop/ folders.
• Info block (frontmatter) keys and values: type: daily, status: active, day: Monday, every tag. The one exception is description:, a sentence in their language.
• Every data-name and data-value in the forms, so the answers come back the same in every language.

Brazilian and European Portuguese (pt-BR, pt-PT) are different sets, and so are Simplified and Traditional Chinese (zh-CN, zh-TW): use the one they chose in Step 1, everywhere. Names they give (their projects, people, companies) stay exactly as they wrote them.

Step 1.5: Welcome

Right after the language, show forms/form-0-welcome.html (for a ready language, i18n/{code}/forms/form-0-welcome.html; for any other language, translate it as translating.md says, but never the lettering). It shows the name, what Phygital Brain is in a few lines, and one button: Continue. Wait for it.

Without a form window, or on a phone, send the same texts as one chat message, with the lettering in a code block exactly as it is in the form, and one click option: the form's Continue.

Step 1.6: Cloud drive

The cloud drive is part of the brain unless the person says they don't want it: it backs the brain up and lets them use it from their phone. On a phone (docs/phone-only.md), use On a phone at the end of this step instead.

Each name below is a message: for a ready language, say it from strings.json (messages); otherwise say the English text given here, translated into their language. For each window, without a form window use a click question with the same texts and options.

1. Is the folder inside a cloud drive? Look at the folder's path. Common signs: on a Mac, ~/Library/CloudStorage/ (Google Drive, Dropbox, OneDrive), ~/Library/Mobile Documents/ (iCloud), a Nextcloud or Dropbox folder in the home folder; on Windows, a G: drive (My Drive), or a Dropbox, OneDrive or Nextcloud folder. If you can't tell, ask: "Is this folder synced to a cloud drive?" If not, say drive_why ("A cloud drive keeps your brain backed up and lets you use it from your phone.") and ask: drive_move ("Help me move it to a cloud drive") (ask which drive they use, walk them through Getting started step 2: install the drive's desktop app, make the folder inside it, then pick that folder for a new chat and say "Set up my second brain" again; then stop) · drive_skip ("Continue without a cloud drive") (the brain stays on this computer only; add "No cloud drive: this brain is on one computer only." to About/tools.md later; skip 2 and 3).
2. Is the drive connected to your agent? If you have tools for their cloud drive in this chat (a connector or MCP server that can read, edit and create files), go to 3. If not, show forms/form-0-cloud.html:
   • Set it up later (answer later): say drive_later_note ("No problem. Your brain works on this computer. Say "connect my cloud drive" anytime to add it."), remember it for the finish message, and go on to Step 2.

   • Set it up now (answer now): show forms/form-0-drive.html (Google Drive starts selected; "Set it up later" there counts as later). Without a form window, ask two click questions: Google Drive · drive_other ("Other cloud drive"), then for Other: Dropbox · OneDrive · Nextcloud · iCloud, with the box for any other drive. Then:

     • Google Drive: say drive_connect_google ("Connect Google Drive: in Claude, open Settings, then Connectors, then Google Drive, and click Connect. Sign in and allow what it asks, then come back here.").
     • iCloud: say drive_icloud ("iCloud has no connector for most agents, so your agent can't reach your brain from your phone. It still syncs to your other Apple devices."), then go on as for later.
     • Any other drive: say drive_connect_other ("Connect {drive}: in Claude, open Settings, then Connectors, and look for {drive}. If it isn't there, add its MCP server. Then come back here."), with {drive} as its name.

     Then one click option: drive_connected_button ("I've connected it"). When they click it and the drive's tools still aren't in this chat, ask them to turn the connector on for this chat (in Claude: the + button, then Connectors), or to start a new chat on the same folder and say "Set up my second brain" again.

3. Does the drive see this folder? Say drive_checking ("Checking that your agent can see this folder in {drive}…"). Save a small file called phygital-brain-check.md in the brain folder with your own file tools. Then look for it by name through the drive's tools.
   • Found: delete it, say drive_ok ("Connected. Your brain will be on your phone too."), and go on.
   • Not found: the drive may still be syncing. Wait about 30 seconds and look again, up to three times. Still not found: delete the file, say drive_failed ("Your agent can't find this folder in {drive} yet. Check that the folder is inside your {drive} folder and that the connector is on. Syncing can take a minute.") and ask try_again ("Try again") (do 3 again) · the cloud form's "Set it up later" (say drive_later_note, remember it for the finish message).

On a phone. The drive isn't optional there: it's the only place the brain can be saved. If you have tools for a cloud drive in this chat, do the folder check in docs/phone-only.md (step 2: make the My Brain folder and test it), then go on to Step 2. If not:

1. Say phone_drive_needed ("Your brain needs a cloud drive to live in, so you can reach it from your phone. Let's connect one.") and ask two click questions: Google Drive · drive_other, then for Other: Dropbox · OneDrive · Nextcloud · iCloud, with the box for any other drive. No "later" here.
2. Google Drive: say drive_connect_google (tap instead of click), then phone_google_account ("No Google account yet? Make a free one at accounts.google.com. It comes with 15 GB of Google Drive.")
3. iCloud: say phone_icloud ("iCloud can't be used from your agent on a phone yet. Google Drive is the easiest choice: it's free and works right away.") and ask the first question again.
4. Any other drive: say drive_connect_other, with {drive} as its name.
5. Then one click option: drive_connected_button, or phone_drive_stop ("No problem. When your drive is connected, come back and say "Build my second brain on my phone".") if they'd rather stop. When they click it and the drive's tools still aren't in this chat, ask them to turn the connector on for this chat (the + button, then Connectors) and try again.
6. Then do the folder check in docs/phone-only.md (step 2), and go on to Step 2.

Step 2: The forms

The welcome already said how long it takes, so go straight to the first form.

For each form (form-1 to form-4): read the file, take out every <!-- ... --> comment and the blocks the comments tell you to remove, translate if needed, and show it. Wait for the answers before the next form.

Form 1: About you (forms/form-1-about-you.html)

Show as it is.

Switches, from the answers:

• BUSINESS = on if selling is Yes, services or clients, Yes, products or Not yet, planning to. Otherwise off.
• BRIEF = on if help includes Organise projects or Goals and decisions. Otherwise off.
• WRITE = on if help includes Write for me.

If they gave a link, read it (web fetch) and pull facts from it: services, clients, bio, tone. Don't ask them to confirm each fact.

Time zone: from the city (Lisbon gives Europe/Lisbon, São Paulo gives America/Sao_Paulo). No city: the computer's time zone.

Form 2: Your work (forms/form-2-your-work.html)

Before showing it:

1. Job questions. Replace the JOB QUESTIONS block with 3 questions written for what they do (from name and you in form 1, and the link if you read one). Follow the example in the block: one choice question about the kind of work, one text question about who they work with or for, one choice question about how they charge or measure their work. Every choice question gets an Other option. Header: "your work as {what they do}". Use data-name values job_type, job_clients, job_pricing. If they picked Personal projects, ask about the projects instead (what kind, who's involved, how much time they have).
2. Keep the BUSINESS ON block if BUSINESS = on, otherwise the BUSINESS OFF block. Remove the other.

Form 3: Your days and people (forms/form-3-days-and-people.html)

Before showing it: remove the Clients option if BUSINESS = off. Remove the VOICE block unless BUSINESS = on and WRITE = on. Remove the BRIEF block if BRIEF = off.

Form 4: Files (forms/form-4-files.html)

Show as it is. Read every file they drop and every link they paste.

If brief_time is No brief, set BRIEF = off.

Check before building: you have their name, BUSINESS, BRIEF, the language and the time zone. If the name is missing, ask for it once in chat.

Step 3: Build

3.1 Folders

Always create:

About/
Daily/
Projects/
Memory/decisions/
Memory/research/
Memory/finances/
Memory/archive/
Library/
Skills/brain-keeper/
Skills/no-ai-slop/
Skills/import/
Skills/meeting-prep/
Skills/thinking/
Skills/connected-apps/
Relationships/

If BRIEF = on, also Skills/morning-brief/. The Brain Keeper's files go in Skills/brain-keeper/ in 3.8. Memory/finances/ is private: it's listed in .claudeignore.

Relationships subfolders only for what they picked in form 3 (people):

Just me or nothing: no subfolders yet. A key person who fits none of the picked types: create the matching folder anyway.

Never create empty extra folders "just in case". Project subfolders, Memory/applications/ and Memory/competitors/ appear later, when content needs them.

3.2 Rule files

Some templates start with an ADAPTING comment for you: follow it, then leave it out of the file. In every template:

• Replace {Name} with their first name, {Time zone} with their time zone, {Language} with the language name and {lang} with its code (en, pt-BR, es...).
• Remove table rows and rules for files or folders that weren't created, unless the row says "created on first use".
• Keep everything under "Always stays in English" (Step 1) in English. Write the rest in their language.

Check: every file in the table exists, and none of {Name}, {Time zone}, {Language} or {lang} is left. Path patterns in routing tables (Projects/{name}/, {title}, {topic}, {skill-name}, {client-name}) stay as they are.

3.3 About files (only when there's real information)

Use templates/about-file.md for the shape: current truth at the top, dated evidence below, newest first.

In me.md, also add a How I want my agent to work section:

• Their style choice, for example "Short and plain: a few short sentences, everyday words, no jargon."
• How they take things in, for example "Short lists and examples."
• Language: {Language}.
• Time zone: {Time zone}.
• Work days: {days}. and when they focus best.

If a file gave their full date of birth, it goes only in me.md (root rule 25).

Anything from form 4's anything_else box goes where it fits by the table above (a preference in me.md, a goal in goals.md, a project in 3.4). Old notes from form 4 are not filed during the build: the finish message offers to bring them in.

3.4 Projects

For each project from form 2 (projects), create Projects/{Project-Name}/README.md from templates/project-readme.md. Their project name, with hyphens instead of spaces.

3.5 People

Key people from form 3 (key_people), and real people or companies named in answers, files or links. One note each, in the right Relationships subfolder, with the info block from templates/relationships-CLAUDE.md and the profile structure from templates/relationships-subfolder-CLAUDE.md. Search first so you never make two notes for the same person. Never invent people.

3.6 Their own rules

If they named work they say no to (form 1, values), add it as the first rule under {Name}'s Rules in the root CLAUDE.md, in their words: "33. Work {Name} says no to: ..." (31 if BUSINESS = off).

3.7 First daily note

Daily/YYYY-MM-DD.md (today, in their time zone) from templates/daily-note.md. Don't create Daily/tasks.md yet: it's created when the first task comes up.

3.7b Connected apps

Look at which connectors you have in this chat for the apps {Name} named in form 3 (comms, notes_now): email (for example Gmail or Outlook), calendar (Google Calendar or Outlook), and their cloud drive. Then ask one question (click options, multi): "Which of these should your brain read from?" with one option per connected app, and None for now. Write the ones picked into Skills/connected-apps/connected-apps.md under Connected, read only, in the shape its ADAPTING note shows. An app {Name} uses but hasn't connected: it's offered in Step 4.1, if it's in ../suggested-apps/README.md.

3.8 Brain Keeper

Set it up now, so nobody has to come back for it. Follow ../brain-keeper/SKILL.md from its Step 2, without its questions:

• Run time: one hour before the morning brief time from form 3. BRIEF = off: 8:00 on their work days.
• Fix mode: safe fixes.
• Where it runs: on this computer, as a scheduled task. If your app can create cloud routines (in Claude: routines) and the brain is in a cloud drive with a cloud connector, ask one question: On this computer · In the cloud, even with my computer off. For the cloud, follow ../brain-keeper/SKILL.md Step 3, cloud.

If BRIEF = on, also schedule the morning brief (the Schedule part of Skills/morning-brief/morning-brief.md) at that time, on their work days, without asking again.

The Brain Keeper's check-up (check-up.md) has no schedule: it runs when they ask.

3.9 Check your work

Fix anything that fails, then check it again.

• No file has leftover placeholders ([name], TBD, {Name}, {Time zone}, {Language}, {lang}, {Fix mode}, {keeper time}, {brief time}, {n}) or em dashes. Path patterns in routing tables stay.
• Every routing table lists only folders and files that exist, except rows marked "created on first use".
• Every note links to at least one other note.
• No files in the folder root except CLAUDE.md, .claudeignore, .backups/ (and the other agent's rule file, if you added one).
• The root CLAUDE.md starts with the version line and has the Language section.
• Every note has a complete info block, keys and values in English.
• Rule numbers in the root CLAUDE.md run with no gaps, and {Name}'s Rules is the last section.
• No [[link]] to a file was translated.
• The Brain Keeper checks in ../brain-keeper/SKILL.md Step 2 and Step 3 pass.

Step 4: Finish

A short, warm welcome, not a report of what was built. Say each message below from strings.json (messages) for a ready language, in their language for any other, as one message:

1. finish_welcome ("Welcome to your second brain, {Name}.")
2. finish_assistant ("It's ready, and from now on I'm your personal assistant. I know your work, your goals and the people around you, and I'll learn more each time we talk. We'll grow together and get better at this.")
3. finish_ask ("Ask me anything, anytime: how to use your brain, what it can do, or an idea you'd like to work on.")
4. finish_start ("To start, talk to me the way you'd talk to a friend: "I had a call with Marta, here's what we agreed." I'll put it in the right place.")
5. The cloud drive: if they chose "Set it up later" in Step 1.6, finish_cloud_later ("For now, your brain is only on this computer. When you're ready, say "connect my cloud drive" and I'll help you link Google Drive or another drive. Your brain is then backed up, and you can use it from your phone too.") If the brain is in a connected drive, finish_cloud_on ("Your brain is in your cloud drive, so it's backed up and on your phone too.") With no drive at all, leave this out.
6. finish_keeper_short ("Every work day at {keeper time}, the Brain Keeper tidies your notes for you.") Only if BRIEF = on, add finish_brief ("Your morning brief comes at {brief time} on work days. Say "morning brief" anytime.")
7. Only for a language with no ready set: the line from Step 1 about the agent translating the brain.
8. Only if they gave old notes in form 4: finish_old_notes ("Want me to bring in your old notes now?") If yes, follow Skills/import/import.md.

Don't list the folders and files, don't explain the rules, and don't give the instructions text (it's only for the phone-only setup, see below): they can ask.

Step 4.1: Add-ons for their work

Right after the welcome, pick add-ons for them from two lists: the apps in ../suggested-apps/README.md and the skills in ../suggested-skills/README.md. Keep only those whose "Good for" clearly fits what they told you in the forms (their work, who they sell to, where they post, which apps they use, their projects), at most 5 in all. Leave out:

• apps that already show up as connectors in this chat (Step 3.7b already asked about those). For a set of apps, like Google Workspace (Gmail, Calendar, Drive), offer only the parts not connected yet and name them ("Google Calendar"),
• skills you already have,
• the ones your agent can't run (for example GEO-SEO needs Claude Code), and on a phone the ones that need a computer.

If none is left, skip this list and go to Step 4.2.

Otherwise, show one list, not a question: the title addons_title ("Add-ons for your work"), then each add-on as a choice they can tick, several at once. For each one, in their language: its name, what it is (addon_type_app ("app to connect") or addon_type_skill ("skill")), and one or two sentences on what it does and what it would do for their own work, named from their answers. Be specific. Not "for your reels", but "turns a script into a short video in your brand colours, so you can make the weekly reels for your bakery without a video editor." Buttons: addons_add_button ("Add selected") · not_now ("Not now"). Don't install or connect anything yet.

For Not now, or if they tick none, say addons_later ("You can add these anytime. Just ask "what can I add to my brain?"").

Step 4.2: Phygital Memory

Then, one more click question (skip it on a phone): phygital_memory_offer ("Want to browse your notes like an app? Phygital Memory shows your brain as a notes app on your computer.") phygital_memory_install_button ("Install Phygital Memory") · not_now ("Not now").

Step 4.3: Install what they chose

Only now, after both answers:

1. Each add-on they ticked, one at a time: follow its page in ../suggested-apps/ or ../suggested-skills/. Check what it needs, skip the install if you already have the skill, ask before installing anything, and when two picks share one install (Skill Creator and Canva's Canvas Design both come with Anthropic's example skills), run it once. Then write Skills/{skill-name}/notes.md from the page's example and add its row in Skills/CLAUDE.md. An app whose page has no notes example (Google Workspace) goes in Skills/connected-apps/connected-apps.md instead, as its page says.
2. For Install Phygital Memory, give the link https://github.com/Phygital-Brain/Phygital-Memory/releases/latest and say to open this folder in it once it's installed (Add memory › From this computer).

Then stop.

The instructions text

Only for the phone-only setup (docs/phone-only.md), where there's no folder to open a chat on. Write it in their language, 3 to 5 short sentences, from their answers. No names, no private details, nothing about this folder. Build it like this:

1. How to talk, from style: for example "Use simple, plain words and no jargon. Keep replies short, no walls of text."
2. How they take things in, from learning, if answered: for example "Use short lists and examples."
3. Always: "Work step by step and check each step before moving on. Do only what I asked."
4. Only if their language isn't English: "Always answer in {Language}."

Example for "Short and plain" and "Short lists":

Use simple, plain words and no jargon. Keep replies short, no walls of text. Use short lists and examples. Work step by step and check each step before moving on. Do only what I asked.

Step U: Upgrade an existing brain

When Step 0 found an older Phygital Brain and they picked Upgrade my brain.

1. Read first, change nothing. Read every CLAUDE.md, .claudeignore, About/me.md, and the files in Skills/brain-keeper/. Note their name, their language (the Language section, or the language their notes are in) and their time zone (from me.md or their city; ask once if neither exists).
2. Count their own rules under {Name}'s Rules in the root CLAUDE.md (in a translated brain, the same heading in its language: the last section of the file). Write the number down. You check it again at the end.
3. Make the list. Compare their files with templates/. List what's new for them in plain words, one line each, grouped by file. Use the CHANGELOG entries newer than their version as a guide.
4. Ask (header Upgrade): Upgrade everything · Show me each change · Cancel.
5. Apply, keeping what's theirs:
   • Add missing sections, routing rows, rules and files, including Skills/no-ai-slop/ (copied unchanged from bundled/no-ai-slop/), Skills/import/, Skills/meeting-prep/, Skills/thinking/ and Skills/connected-apps/ (ask 3.7b's question), and .backups/ in .claudeignore.
   • Make a backup of every file you'll change before you change it (Backups and Undo).
   • Replace the numbered base rules with the new list (1 to 32, or 1 to 30 when they don't sell). Their own rules stay under {Name}'s Rules, word for word and in the same order, renumbered to follow on. If one of their rules says the same as a new base rule, keep theirs and mention it.
   • Keep their own routing rows, folders and anything else they wrote that isn't in the templates.
   • Never rewrite notes to new info blocks. Old notes get them the next time they're edited.
   • Replace the version line at the top of the root CLAUDE.md with this skill's version: <!-- Phygital Brain, version X, language: {lang} -->.
   • Folder names. If the root CLAUDE.md has a ## Folder map, use it for every path in the new templates, and the matching i18n/{code}/ set if one exists. Add map rows for any new folders or notes in this version: from i18n/{code}/strings.json, or, with no ready set, choose the names yourself following translating.md. With no ready set, translate the new parts the same way.
   • Brain Keeper. Update it (follow ../brain-keeper/SKILL.md, its Step 0.4 keeps queue.md and the reports). If it isn't installed, install it from its Step 1.
6. Check: the same number of their own rules as in step 2, word for word; every check in 3.9 that applies; nothing deleted.
7. Finish: one line in today's daily note, then 2 or 3 lines on what changed.

Step T: Set up a team

When {Name} says "set up my team". The guide for people is https://github.com/Phygital-Brain/phygital-brain/blob/main/docs/phygital-memory.md#share-with-a-team.

1. Check the folder is shared. Ask whether the brain is already shared through Phygital Memory (or another sync service the team uses). If not, point them to the guide and stop. Don't move or copy the folder yourself.
2. Ask who's on the team (one chat message): each person's name, role, and which folders they may change. The person asking is the owner unless they say otherwise.
3. Back up the root CLAUDE.md (Backups and Undo).
4. Replace the short Team section of the root CLAUDE.md with the one from templates/team-section.md, filled in. Ask first (root rule 9): show the section, then write it after the yes.
5. For each person: a note in Relationships/Team/ (search first, create the folder and its CLAUDE.md if needed), and their daily notes folder Daily/{person}/ (created when their first note is written).
6. Brain Keeper: ask whose agent, or which cloud routine, runs it. Only one may. Remove any other schedule. In Skills/brain-keeper/daily.md, note under Settings that daily notes are per person, in Daily/{person}/.
7. Finish: one short message: who's in, what each can change, and that permissions themselves are set when sharing in Phygital Memory.