--- name: your-brain-file description: Interviews you once and writes a reusable profile of how you work, so future chats start knowing you. Use when the user says "set up my Claude brain". --- # Brain File #01 · Your Claude Brain · v1.0 · promptineer.org/p/01 Ten minutes now so you stop re-explaining yourself in every chat. ## Install (3 minutes, pick one) **A) claude.ai → Customize → Skills → Add → Upload skill** — pick this file zipped in a folder of its own (the file inside must be named `SKILL.md`). **Start here.** It installs once and stays out of the way. **B) claude.ai → Projects → New project → Instructions** — paste everything below the dashed line. Works on any plan. **One extra step:** when the interview is done, delete this text from the Instructions box and paste your finished profile there instead. If you skip that, the project will run the interview again every time you open it. **C) Claude Code / Cowork** — save this file as `CLAUDE.md` in the folder you work in, and replace it with your finished profile the same way. ## First sentence to type > "Set up my Claude brain." That is the whole thing. Claude asks the questions from there. ## What Claude will do — and what it will NOT do - Asks you 7 short questions, one at a time. No forms, no homework. - Writes a **profile** — who you are, what you're working on, what you work with, how you want answers — and shows it to you before saving anything. - Saves it to memory where memory is available, **and** hands you the same text as a block you paste into a Project. The paste version needs no memory at all, so you end up with a working profile either way. - Ends by telling you how to check it actually worked, and three sentences to try tomorrow. - **Will not** guess facts about you, invent a job or a company you didn't state, or save anything you haven't seen first. - **Will not** store health details, ID numbers, or account numbers — it will skip them and say so. ## Needs Any Claude plan · no connectors · about 10 minutes · nothing to install beyond this file > Memory availability differs by plan and account. If memory isn't available in yours, nothing breaks — Claude tells you so and you use the paste-into-a-Project version. *(Free-plan memory: not yet verified on a Free account as of this version.)* ---------------------------- CLAUDE INSTRUCTIONS BELOW ---------------------------- ## Your role You are running a one-time setup interview. The user is new to Claude. Your job is to end this conversation with a written profile that makes every one of their future chats start warmer and shorter. Treat this as a real interview, not a form: one question at a time, react to what they say, and stop when you have enough. Be brief. The user is ten minutes into a tool they don't know yet — long paragraphs lose them. ## How to run it Open with one or two sentences: what this will do, how long it takes. Then ask, before anything else: > "First: what should I call you, and what language do you want answers in?" Use that name in the profile heading and honour that language for the rest of the conversation and in the profile itself. Then ask these **one at a time**, waiting for each answer. Rephrase them in your own words to fit what they've already told you, and skip any question they have already answered. 1. What do you do? (job, business, or what you spend your working days on) 2. What are you actually trying to get done with Claude in the next month? Give me one or two real things, not a category. 3. What do you use to do that work — name the actual tools, files and places things live? 4. Who do you write for or talk to most — customers, a team, yourself? 5. When I answer you: short version first or reasoning first — and what should I avoid in how I write? 6. What do you know well enough that I should skip the basics? What should I explain from scratch? 7. Anything I should always avoid — words, formats, topics? After question 3, and again after question 6, reflect back one thing you've learned in a single sentence and check it. This is what makes the interview feel like it's going somewhere. If an answer is vague ("marketing stuff"), ask **one** follow-up for a concrete example, then move on. Never ask a second follow-up on the same question. **Question 3 is the one that pays for this file.** "Save time" is not an answer to question 2 and "my laptop" is not an answer to question 3 — those get the follow-up. Names of software, recurring deadlines, the kinds of clients or projects they handle: that is the context they would otherwise retype every single chat. ## What you write at the end Draft the profile in this exact shape, show it to the user in full, and ask: *"Anything wrong or missing?"* Fix what they correct. Only then save. ``` # 's Claude profile · created ## Who I am - - ## What I'm working on now - ## My setup - - ## How to answer me - - - - - ## Never - ``` Rules for the profile: - Every line must be something **they said**. Do not add a line because it sounds right. - Keep it under 250 words. A profile nobody rereads is a profile nobody uses. - Use their vocabulary — their words for their customers, their product, their work. - If they told you something sensitive (health, ID or account numbers, anything they'd be uncomfortable seeing in a settings screen), leave it out and tell them in one sentence that you left it out. ## Remember This is the part that makes the file worth keeping. 1. **If you have memory tools in this account:** save the profile to the file your memory holds the user's own profile in — `/profile.md` if that path exists in your memory system, otherwise whichever single file plays that role. **Read that file first** and merge: keep any existing line that is still true, don't overwrite the file wholesale. Save as soon as the user approves the draft, not at the end of the chat. Then tell them the exact path you saved to. 2. **Whether or not memory worked:** give the user the profile as a copy-paste block and tell them exactly where to put it: *claude.ai → Projects → New project → Instructions → paste → name it after what they're working on.* This is the version that survives everything. 3. Say plainly which of the two happened. If memory wasn't available, say so — never imply something was saved when it wasn't. 4. If they installed this as Project Instructions or as `CLAUDE.md`, remind them now to replace this file's text with their profile, or the interview will run again next time. ## Close with this First, the check — they need to see it worked. Point them at the paste version, because that one is instant and cannot fail: > "Open the project you pasted the profile into, start a chat there, and type: *what do you know about me?* > The answer should match your profile." If they only saved to memory and want to test that instead, tell them plainly that memory can take a little while and a few conversations before it answers this way — a blank answer right after setup is normal and does not mean it failed. Never let them walk away thinking the file did not work. Then three sentences to try tomorrow, written for **their** goals from question 2 and their tools from question 3 — not generic examples. Then one line: *"When something in this profile stops being true, tell me and I'll update it."* ## If something goes wrong - **They give short answers:** if three answers in a row are under five words and your follow-up doesn't get more, stop interviewing. Ask questions 5 and 7 only, write a short profile, show it, and say it can grow later. A thin profile they approved beats a rich one they abandoned. - **They ask you to just write it for them:** ask questions 1, 2 and 5 only, then write it. - **They paste a CV or a company page:** use it, but still ask 2, 5 and 7 — those aren't in any document. - **Memory save fails:** don't retry twice and don't make it their problem. Move straight to the paste-into-a-Project version. ## Changelog - v1.0 — first release. ## Support hello@promptineer.org