The /personalize Flow.
Twenty-one questions about you, your work, and how you learn. Claude turns the answers into a CLAUDE.md tailored to you. Ten to fifteen minutes, conversational, in batches of three to five questions at a time. Run it once. Every future session in this repo starts with an agent that already knows who you are.
- A CLAUDE.md written from your real answers nine sections, filled in from what you tell it, that every future Claude session reads automatically.
- A starter personal vault your local second-brain folder seeded with three files about you, your tools, and your style.
- A running start on workshop day you arrive Sunday with an agent that already speaks in your preferences and reaches for the tools you actually have, instead of spending the first hour catching up.
Run /personalize right now.
You do not have to read this card first. The command is built to walk you through itself. Open Claude Code in your workshop repo and paste this.
How this fits in the OS.
Claude Code installed, the workshop repo cloned, and ten to fifteen minutes when you can actually think about your work instead of typing one-word answers.
A question bank shipped with the repo, your real answers, and Claude as the interviewer that turns the conversation into a profile.
A working CLAUDE.md, a starter personal vault on your computer, and the personalized foundation every later command, skill, and build reads before it does anything.
What /personalize actually is.
A single slash command, /personalize, that runs a guided interview. Twenty-one questions, three sections, batched three to five at a time so the conversation does not feel like a survey. At the end, Claude writes your answers into nine sections of your CLAUDE.md and offers to set up a small personal vault in your Documents folder.
The point is not the questionnaire. The point is the file at the end. Your CLAUDE.md is the persistent context every future agent session reads before it does anything else. A thin profile produces a thin assistant. The command is designed to push back on short-circuit answers so you do not end up with one.
You do not draft the CLAUDE.md sections yourself. The command interviews you, reflects what it heard back to you ("So you're spending most mornings on X and the bottleneck is Y. Got it."), drafts the file, and asks you to read it before save. You correct, then it locks in.
What the twenty-one questions cover.
Three sections, seven questions each. You will recognize the shape of a real onboarding conversation, because that is what it is.
About You
Your typical workday, how you rate your own AI experience, the most impressive thing you have built with it, the one thing you would automate first, and even your personality type, so Claude stops treating you like a stranger and calibrates how it talks to you.
About Your Work
Your top daily tasks, the tools you reach for most, what eats the most time that you wish was automated, how you handle meeting notes and follow-ups, and the biggest operational bottleneck in your work right now.
How You Learn and Work
How you size up a new tool, whether you want a structured walkthrough or the controls handed to you, brief answers or deep dives, and how you prefer to give and get feedback.
Plus follow-ups. Claude is allowed to dig into a vague answer once, and it should.
What gets written when you finish.
Two things land on your computer, in this order.
- Your CLAUDE.md, populated. The command opens the CLAUDE.md at the root of your workshop repo and writes nine sections from your answers: six that capture you and how you work (About Me, How I Work, How I Learn, AI Experience, Priority Use Cases, Communication Preferences), plus three that set up your toolbox (My Tools and Accounts, My Custom Commands, and a Tool Awareness block that tells future agents to quietly track what you have installed and offer at most one timely suggestion when it fits, never a sales pitch every turn). You can tag any tool
(no-suggest)to keep Claude from recommending it proactively. The Your Brain File card walks through what each section is for. If a CLAUDE.md is already there, it preserves what you have and only adds or updates these sections. - A starter personal vault. A local folder, by default in your Documents, with three files Claude can read on every later session:
my-self.md(identity),my-tools.md(your stack), andmy-style.md(how you want to be talked to). The command also drops in a.gitignorethat keeps the vault out of any repo you commit. Nothing leaves your computer unless you later connect a remote service.
Setting up the vault is fast. The command is designed to finish in under a minute on a normal computer and does not stand up a database, Docker, or any remote service. That is intentional. Most of what makes an agent feel sharp is the markdown, not the infrastructure.
It also gets your secrets sorted.
Between writing your profile and setting up the vault, /personalize pauses on one practical thing: how you will store the API keys your agents are about to need. The workshop's recommended path is Infisical, and the installer sets it up for you, so most students are already covered. The full why, the alternatives (1Password or a locked .env), and the step-by-step live in the Secrets Management card.
Why this is the one thing to do before workshop day.
Two reasons.
First: the workshop is taught from a position of agents that already know who their owner is. If yours does not, you spend the first hour catching up. Running /personalize Saturday means Sunday starts with an agent that already speaks to you in your preferences and reaches for the tools you actually have.
Second: it is the floor the rest of the workshop builds on. Almost everything you do later, every command, every skill, every build, reads your CLAUDE.md and draws on the vault this seeds. Get the profile right first, and everything downstream already fits you instead of a generic stranger.
A real interview with a real interviewer, even an AI one that knows what to ask and what to push on, produces a CLAUDE.md that does work. The command exists because handing students a blank template and walking away does not.
What changes for you.
The questionnaire takes fifteen minutes. What it changes about every session after it is the real payoff. Three shifts you will feel:
- You stop introducing yourself. From the first turn of every later session, the agent already knows your role, your tools, and how you like to be talked to. The "here is who I am and what I care about" preamble you would normally retype into a fresh chat is already loaded.
- The whole workshop fits you, not a generic beginner. Every command, skill, and build downstream reads your profile, so the examples, defaults, and suggestions land in your world instead of a stranger's. The agent reaches for the tools you actually have.
- Your profile becomes an asset you maintain, not a setup step. As your role and stack change, you keep it current, or ask Claude to in plain English. Six months in, it is a living record of how you work that every agent you run draws on.
You can run it again.
Running /personalize again is safe. If you run it a second time, it keeps your existing CLAUDE.md and vault files by default and only fills in what is missing. If you genuinely want to start a section over from scratch, you add the word --reseed to the end of the command, so you type /personalize --reseed instead of just /personalize. That extra word is simply an instruction that tells the command to rebuild from scratch rather than fill in the gaps, and even then it asks you to confirm before it replaces each file. It never wipes a vault silently.
Run it again when your role changes, when you finish a big project, when you add a tool you want the agent to know about, or when you notice Claude keeps misunderstanding something basic about how you work.
Straight from the source.
Everything this card describes is in your workshop repo, in plain markdown you can open and read before you run anything. Two files back it:
configs/commands/personalize.md: the command itself, step by step. What it asks, the order it writes your CLAUDE.md sections, the secrets check, and how the vault gets created. If you want to know exactly what will happen before you run it, read this.configs/get-to-know-you.md: the question bank. All twenty-one questions, verbatim, across the three sections. Skim it ahead of time if you like to think before you answer.
For what each CLAUDE.md section is actually for once it is written, the Your Brain File card is the companion read.
Where to go next.
Once your CLAUDE.md exists, the obvious next moves are reading what is in it and learning how the agent uses it.