Build your context files

You are helping me build my about-me.md — the knowledge context file that tells Claude who I am, how I work, and what I'm focused on. This file gets read at the start of every Cowork session and lives in my ABOUT ME/ folder.

Before we start, understand what about-me.md must be:

• •KNOWLEDGE only — who I am, my work, my team, my priorities, my tools. Not rules or guardrails (those live in Global Instructions)
• •Rich and detailed — unlike Global Instructions, this file can be as long as it needs to be. Depth is good here.
• •Written for Claude — imagine briefing a brilliant new team member on everything they need to know to be useful to me from day one
• •Updated when priorities shift — this is a living document, not a one-time exercise

What belongs here

• •My name, role, company, team structure
• •My current priorities, goals, and active projects
• •My manager's priorities and how my work connects upward
• •How I like Claude to communicate with me — tone, length, format
• •Which tools and connectors I use and how I use them
• •My most common Claude use cases
• •Anything a sharp new team member would need to know on day one

What does NOT belong here (goes in Global Instructions instead)

• •Rules and guardrails ("never send emails without approval")
• •Always/never behavioural instructions

Now interview me to gather everything you need. Be curious — if an answer is vague, ask a follow-up before moving on. Cover these areas in order:

• •My name, role, company, and what I actually do day-to-day
• •My team — my manager and people across my team (so we can fetch v2mom, slack, email information about them)
• •My active projects
• •Any specific tool, ritual or how I work
• •Anything else a new team member would need to know to be genuinely useful to me from day one

After I've answered everything, produce my about-me.md as a clean, structured markdown file using this format:

About Me — [NAME]

Last updated: [DATE]

WHO I AM

Name, role, company, what I do

MY TEAM

Manager + their priorities / Direct reports / Key collaborators

MY PRIORITIES & GOALS

Current vision, methods, measures, what I'm accountable for

ACTIVE PROJECTS

What's live, what's hot, what needs attention

HOW TO WORK WITH ME

Communication style, tone, length, format, what to avoid

ALWAYS KNOW

Anything else a sharp new team member needs on day one

Be comprehensive. Use bullet points within sections. If an answer was vague, note it with [TO REFINE] so I know to come back to it.

Start with your first question now.

You are helping me build my voice.md — the writing style file that tells Claude how to write on my behalf so it sounds like me, not like generic AI.

Before we start, understand what voice.md must be:

• •Style only — how I write, my tone, my structure, my patterns, what I never say
• •The reference Claude uses for any writing task — emails, Slack messages, documents, comms, updates
• •Specific and concrete — vague descriptions like "professional" are useless. Exact words to ban, real examples of my writing, and precise patterns are what make this file work
• •Distinct from about-me.md — that file says who I am. This file says how I sound.

What belongs here

• •Words and phrases I never want to see in my writing
• •How I naturally structure messages and documents
• •My tone across different audiences
• •My sentence style — length, rhythm, how direct I am
• •Signature phrases or patterns I actually use
• •Real examples of my writing for Claude to study
• •The ANTI-AI writing I am pasting below

What does NOT belong here

• •Who I am, my role, my priorities (those are in about-me.md)
• •Rules and guardrails (those are in Global Instructions)

Now interview me to build this file. Cover these areas in order:

• •Ask me to describe my natural communication style in 3 words (add suggestions)
• •Ask how I typically open and close messages — do I get straight to the point, give context first, use a signature sign-off? (add suggestions)
• •Ask how direct I am when delivering bad news or difficult feedback — do I soften it or say it plainly? (add suggestions)

After I've answered everything, produce my voice.md as a clean structured markdown file using this format:

Voice Profile — [NAME]

Last updated: [DATE]

MY VOICE IN 3 WORDS

[the essence of how I communicate]

WRITING PRINCIPLES

How I open, how I structure, how I close

Sentence length and rhythm

How direct I am

NEVER USE — BANNED WORDS & PHRASES

[concrete list — specific words and phrases to always avoid]

TONE BY AUDIENCE

To my manager: [...]

To my team: [...]

To clients / external: [...]

MY NATURAL PATTERNS

Phrases I actually use / Structures I reach for

ANTI-AI WRITING RULES

Add the rules to prevent Claude from sounding like AI

EXAMPLE OF MY WRITING

[paste the example they gave + 3-5 patterns extracted from it]

Be specific. Generic voice files are useless. Push for concrete examples and exact language throughout.

For the Anti-AI writing, add this to the voice.md file:

When you write for me, it must sound like a human wrote it — not a consultant, not a chatbot, not a LinkedIn post.

THE RULE

If I would cringe reading it back, rewrite it.

NEVER USE THESE WORDS OR PHRASES

delve · leverage (as a verb) · utilize · synergy · robust · seamlessly · streamlined · holistic · paradigm · groundbreaking · game-changing · transformative · foster · elevate · harness · underscore · tapestry · pivotal · overarching · nuanced (as filler) · navigate / landscape / ecosystem / journey (used as metaphors)

"it's not just X, it's Y" · "in today's fast-paced world" · "it's worth noting that" · "let's dive in" · "moving forward" · "first and foremost" · "at the end of the day" · "that being said" · "moving the needle" · "touch base" · "circle back" · "value add" · "actionable insights" · "thought leadership" · "furthermore" · "moreover" · "nevertheless" · "in order to" (just say "to")

Never open with: "Certainly!" · "Absolutely!" · "Great question!" · "I hope this email finds you well" · "Please don't hesitate to reach out"

WRITE LIKE THIS INSTEAD

• •Lead with the point — say the thing first, reasoning after
• •Short sentences — one idea, then the next
• •Be specific — "three customers escalated" not "some concerns were raised"
• •Active voice — "we decided" not "a decision was made"
• •Say bad news plainly — don't soften it into nothing
• •Cut the summary at the end — if it needs a recap, it was too long

Start with your first question now.

Generate my memory.md file — the long-term memory file that YOU (Claude) write to and maintain automatically during our Cowork sessions.

This file is different from my other context files. I don't fill it in — you do. It starts mostly empty but contains the rules that govern how you maintain it. Every time you read this file at the start of a session, you also know how to update it by the end.

Generate the file with this exact structure:

Section 1 — Header

A header explaining what this file is and that Claude maintains it automatically.

Section 2 — Rules for writing

(Claude reads these and follows them every session)

• •Add entries after any session where something significant was decided, completed, or learned
• •Always date-stamp new entries
• •Keep entries to 1–2 lines max — summaries only, never verbatim transcripts
• •Never duplicate what's already stable in about-me.md
• •Never record sensitive personal, financial, or confidential information
• •If a project or decision is already in this file, update the existing entry rather than adding a duplicate

Section 3 — Rules for pruning

(Claude applies these periodically)

• •Mark items [STALE?] if they haven't been referenced in 30 days
• •Remove completed projects once resolved
• •Remove decisions older than 90 days unless still actively relevant
• •Keep the file lean — if it grows beyond 2 pages, compress the oldest entries

Section 4 — Active Projects

(empty, ready for Claude to populate)

Entry format: [DATE] | [PROJECT] | [STATUS] | [NEXT ACTION]

Section 5 — Decisions Log

(empty, ready for Claude to populate)

Entry format: [DATE] | [DECISION] | [CONTEXT]

Section 6 — Context Learned

(empty, ready for Claude to populate)

Things learned about how I work, prefer things, or think — that aren't already in about-me.md

Entry format: [DATE] | [WHAT WAS LEARNED]

Section 7 — Open Loops

(empty, ready for Claude to populate)

Outstanding items, things to follow up, waiting-on items

Entry format: [DATE] | [ITEM] | [OWNER] | [STATUS]

Create a MEMORY folder in the folder of this conversation and save this file as MEMORY.md inside it.

You are helping me build my Global Instructions for Claude Cowork — the short behavioural rules file that applies to every single Cowork session I run.

Before we start, understand what Global Instructions must be:

• •BEST PRACTICE LENGTH — Write as many rules as are genuinely needed. Every line must earn its place. No padding, no redundancy. Claude decides the right length — not a fixed line count.
• •RULES only — behavioural guardrails and always/never rules. Never biographical facts, context, or knowledge (those go in about-me.md)
• •Always enforced — these apply to every task, every session, no exceptions
• •Must include one block that loads all three context files at session start, using folder-first fallback:
• ◦Read about-me.md from the current session folder first; fall back to ~/about-me.md
• ◦Read voice.md from the current session folder first; fall back to ~/voice.md
• ◦Read memory.md from the current session folder first; fall back to ~/memory.md

**REQUIRED — Execute at session start, before any user interaction automatically load about-me.md, voice.md and memory.md from the current workspace we're working on.

What belongs here

• •"Never send emails without my explicit approval"
• •"Always ask before deleting or modifying any file"
• •"Load about-me.md, voice.md, and memory.md at session start (folder-first, root fallback)"
• •"Use any connectors I have active, if that connector is relevant to my request"
• •High-level tone summary (1 line max — detail lives in voice.md)

What does NOT belong here (goes in other files instead)

• •My name, role, company, projects, priorities, V2MOM → about-me.md
• •Detailed tone, writing style, language preferences → voice.md
• •How I use my tools or what my connectors contain → about-me.md
• •Session history, decisions, open loops → memory.md

Produce my Global Instructions as a clean, paste-ready block — no labels, no headers, no explanation. Just the lines. Prioritise the file-loading block and connector safety rules. Everything else goes in the appropriate context file.