Give me 2 hours and you'll have your own leadership OS running in Claude code

Your calendar is packed, your company is rushing to use AI, your teams are on their way, but what about you? Sure you're prompting claude and chatgpt, might have connected some tools, but you know you should be learning and using this stuff at a much deeper level.

When you've finally carved out time for longer focused session, you're faced with the next obstacle: What should you do? Listen to the latest episode of the How I AI podcast? Define your learning goals so you can figure out something to build? Or push it to next week and instead tick off an urgent task. The stress compounds.

Let's cut through it: If you've already installed Claude Code, give me 2 hours and you'll kickstart your own Leadership OS. You'll set up the foundation pieces, understand how context works with AI (which makes you better at everything else), and start building the support system no one else is going to build for you.

Why leaders need a Leadership OS in Claude Code

The value of Claude Code is repetition.

You set up workflows once, they get better every time you use them, and over time you automate the thinking patterns and tasks that are relevant to your work. Also, no one else will build this for you (...yet). Your company might roll out AI tools, your team might share prompts, but the specific system that supports your specific priorities, your stakeholders, your decision patterns is yours to create.

And perhaps most importantly: you'll learn hard skills (context management, prompt design, automation) that will keep you relevant as a leader while most people are still asking "how should I use AI?"

What you're building today

By the end of this 2-hour session, you'll have three things set up and running:

CLAUDE.md — Your global preferences file that makes Claude work the way you think, saves you from repeating yourself, and teaches you how to be specific about context (this skill transfers to every AI tool you'll ever use).

business-profile.md — A strategic reference file Claude reads before helping with decisions, so it can give you grounded feedback on goals and priorities instead of generic advice.

/weekly-review — A repeatable command you'll run every week that reflects on what moved, plans what's next, reads your calendar, and writes a weekly note. You'll notice the difference between "using AI sometimes" and "having a thinking partner" pretty quickly once you start using it.

These three pieces are your foundation. Everything else you build later (connecting to Slack, automating business reviews, building custom workflows) builds from here.

Subscribe to follow this series.👇🏻

Step 1: Start Claude Code and create your workspace

If you haven't already, download Claude Code. Here's the official guide https://code.claude.com/docs/en/quickstart. Keep a chat with Claude open for any questions you have along the way and you will set it up in no time.

Then create a folder specifically for this work, call it "leadership-os". You want control over which files Claude Code can see, so don't just open it in your Desktop or Documents folder.

Once you've created the folder, navigate there and open the terminal (On Mac, right-click the folder and select 'New Terminal at Folder').
In the terminal type claude and press enter (like this)

leadership-os/ claude

When you first open Claude, you'll see some initial messages. For anything you have questions about, ask Claude Code or in the chat you're used to. The interface is simple: you type, Claude responds, and you can ask it to create or edit files in your workspace.

Step 2: Create CLAUDE.md so Claude remembers how you work

What this does: CLAUDE.md lives at ~/.claude/CLAUDE.md and loads every time you open Claude Code. Think of it as your standing instructions, the preferences you want Claude to remember in every session so you don't waste time re-explaining how you like to work.

Writing this file helps you practice being specific about context, which is the key skill to working well with AI. The more precise you are about what you want (response length, tone, when to push back on your ideas), the better Claude performs, and this same skill makes you better at using every AI tool because you create a sense of what needs to be specified versus what can be left implicit.

The benefit: You save 5-10 minutes every session by not repeating yourself, and Claude gives you responses that match how you actually think instead of generic AI output.

Paste this prompt into Claude Code:

I'm setting up Claude Code for the first time. I want to create my global preferences file — ~/.claude/CLAUDE.md — so Claude works the way I think from day one.

Ask me the questions below one at a time. Wait for my answer before moving to the next.

1. What's your name?
2. How long should Claude's responses be by default? (Short and direct / thorough with context)
3. What format do you prefer? (Bullet points / prose / depends on context)
4. What's your tone preference? (Professional and direct / conversational / energetic with emojis)
5. When you disagree with my direction — should you push back and make the case, or flag it once and move on?

Once I've answered all 5, write ~/.claude/CLAUDE.md with the following structure:

# [Name]'s Global Claude Code Preferences

## How [Name] and Claude work together

- Always use [Name]'s name (never "I", "you", or "my") when writing instructions — this avoids ambiguity since the file is read by Claude, not written for humans
- Ask one clarifying question at a time — never stack multiple questions
- Discuss approach before acting. Get alignment, then move.

## Communication preferences

(fill in from answers)

## Trust and expertise

- Trust [Name]'s domain judgment. Ask for clarification rather than assume something is wrong.
- Offer alternatives rather than insist on a particular approach.

## Reference context files

- Business profile: context/business-profile.md

Show me the finished file when done. Ask question 1 now.

Claude will interview you, then write the file. Once it's done, you'll have a preferences file that loads automatically every time you open Claude Code. And of course this is a file that you will grow and tweak over time, but this is a good starting point.

Step 3: Create business-profile.md so Claude understands your work

What this does: This file gives Claude strategic context about your role, priorities, stakeholders, and constraints. It's a reference document Claude reads before giving you feedback or helping you think through decisions.

Without this, Claude gives generic leadership advice. With it, Claude can be more pointed in your interactions: "this goal conflicts with your Q1 priority" or "have you talked to Sarah about this, since it touches her team's roadmap?" The difference is strategic thinking versus surface-level tips.

The benefit: You get a thinking partner who knows what you're accountable for, what's blocking you, and who matters in your landscape. This turns weekly planning from "write down some goals" into "get honest feedback on whether your plan actually moves your priorities."

Paste this prompt into Claude Code:

Next I want to create my professional context file — context/business-profile.md — so you can be a genuine thinking partner, not just a task executor.

Ask me the questions below one at a time. Wait for my answer before moving to the next. Receive my answers neutrally — don't reframe, summarize or celebrate them. Just confirm you heard it and ask the next question.

1. What is your role and what are you measured on?
2. What are your top 3 priorities right now? For each: what does success look like, and what's currently standing in the way?
3. Who are the 3 key people you need to work with or influence right now? For each: their name, their role, and what their current goals are.
4. What else shapes how you operate — company stage, constraints, anything you're working on in yourself as a leader?

Once I've answered, create context/business-profile.md with this structure:

# Business Profile — [Name]

## Role and accountability
What [Name] owns and what they are measured on.

## Priorities
Each priority as a single entry: what it is, what success looks like, and what's currently blocking it.

## Landscape
Each key person: name, role, their goals, and why they matter to [Name]'s priorities.

## Context
Anything that shapes how [Name] operates — company stage, constraints, development edges.

Rules for writing this file:
- This is a strategic reference. Write it so Claude can give sharp, grounded feedback on weekly goals and decisions — not a summary of facts.
- Each priority gets one name. Use it consistently throughout the file.
- Write in third person: "[Name] owns..." not "I own..."
- Only include what's professionally relevant.

Show me the finished file when done. Ask question 1 now.

Answer Claude's questions honestly. The more specific you are about what's blocking your priorities and who matters in your landscape, the better Claude can help you think strategically.

Pro tip: Claude Code is rolling out voice mode during March, so use that or Wispr Flow if you are tired of typing.

Step 4: Set up your /weekly-review command

What this does: A slash command is a shortcut you can run anytime by typing /weekly-review in Claude Code. The one we are creating today leads you through a reflection on last week, asks questions about next week, reads your calendar, and finishes it all off by writing a weekly note with your plan.

I love a good weekly review! But sometimes I end up deprioritising them and only have time to look at the tactical stuff OR I end up rabbit holing on my reflections and when it comes to breaking down tasks and finding time in the calendar, I'm exhausted from my deep thinking and end up doing a poor job at that.

Both scenarios end up being weak on linking strategy to concrete actions.

This command helps you stay on track on both.

The benefit: You get pretty good feedback every week on whether your goals match your calendar, you build momentum by naming what's working (not just what's broken), and you end up with a record of weekly themes and progress you can look back on. After a few weeks, patterns become visible.

Paste this prompt into Claude Code:

I want to set up my weekly review slash command. Please do the following:

1. Create the folder weekly/ in the current directory
2. Create the file ~/.claude/skills/weekly-review/SKILL.md with this exact content:

---
name: weekly-review
description: Weekly review and forward planning. Reflects on last week, then plans next week with intention.
---

Read context/business-profile.md before starting. Use the tone and communication style set in CLAUDE.md throughout.

## Phase 1: Looking back

Ask: "What moved last week on each of your priorities? What are you proud of?"

Wait for the answer. Then:
- Name what's working specifically — this builds momentum, don't skip it
- Give an honest read on what didn't move and why, adapted to their tone preference
- Don't inflate partial progress

## Phase 2: Looking forward — questions first

Based on what they shared and their priorities in business-profile.md, ask 2-3 questions grounded in what they just said. Questions should surface real blockers, name things being avoided, and help them think about what next week should contain.

Do not suggest goals. Wait for their answers.

## Phase 3: Calendar and theme

Ask them to drop a screenshot of next week's calendar.

Read it for:
- Who they're meeting — cross-reference with their landscape in business-profile.md
- What the overall load looks like — is there realistic space for their goals?
- When during the week they could actually work on each priority

Reflect back what you see. Help them decide not just what they want to focus on, but when. If the calendar doesn't support their intentions, name it.

Then propose one sentence as a theme for the week — something that captures what this week is about based on everything they've shared. Ask if it lands or if they'd adjust it.

## Phase 4: Write the weekly note

Once they've confirmed their plan, write weekly/YYYY-MM-DD.md with:
- Theme for the week (confirmed or adjusted)
- What they're proud of from last week (1-2 sentences)
- Next week's goals in their own words
- When in the week they'll work on each goal
- Key todos in - [ ] format

Tell them the file has been created and where to find it.

Confirm both have been created and show the paths.

Once Claude creates the command, test it by typing this in Claude Code:

/weekly-review

It should come up as a suggestion. Run through it once to see how it works.

What you've built

You now have the foundation of a Leadership OS that's yours. Claude knows your preferences, understands your work context, and has a repeatable workflow you can run every week.

3 weeks from now you'll open Claude Code and it will already know what's important to you this week and how it can help. That's what you just built 🥳

The weekly command will get smarter as you use it. Update business-profile.md when priorities shift, add new reference files when you start working on new areas, and the system adapts.

Where to go from here

Today was about getting up and running. With your foundation in place and a working understanding of context and commands, you can start building the rest.

Give Claude more compartmentalised content: Create a more extensive goals file with your own professional goals, have Claude interview you for it, link it from business-profile.md. Create a stakeholder file with relationship health (green, yellow, red) so your weekly review can flag who needs attention.

Connect Claude Code to your other systems. Link Claude to Slack, to your email, to Jira. Build a /business-review command that surfaces what's at risk, what's blocked, and what needs attention based on real data instead of your memory.

Build workflows for specific areas. If you're working on a product launch, create a context file for that project with its own goals, timelines, and stakeholders. If you're managing a team transformation, create a file that tracks progress and blockers. As you work on different areas, keep building context files and commands.

The pattern is the same: create context, build commands, let it compound.

You've got the foundation. Now build the rest!

This is part of a series on building your leadership OS in Claude Code. Get the next one in your inbox by signing up for my newsletter.👇🏻