The Complete Guide to Claude Code
Resources from the video
Every link I mention in the video, in one place.
Step 1: What Claude Code is (and why it's different)
You've probably used ChatGPT or Gemini. They're chatbots: you ask, they answer, they forget. Every conversation starts from zero.
Claude Code is something completely different. It's not a chatbot. It's a full work environment.
Three key differences:
1. Reach. It lives inside your computer. It reads your files, opens your apps, uses your browser. It plugs into Gmail, Calendar, Drive, Slack . the tools you already use.
2. Action. A chatbot explains. Claude Code does. You tell it what you want. It finds a way and hands back finished work, not instructions.
3. Autonomy. It runs on a schedule, in the background, while you're in a meeting or spending quality time with your family. The work doesn't stop when you stop.
In short: ChatGPT talks to you. Claude Code works for you.
Throughout this challenge we work only from the Code tab. Not Chat, not Cowork. Code.
Step 2: Install Claude Desktop
Claude Code runs inside an app called Claude Desktop. Not in a browser, not in a terminal. A real app on your computer.
Mac:
- Go to claude.com/download
- Download the .dmg file
- Open it and drag Claude into your Applications folder
Windows:
- Install Git first (see red box below). Claude will refuse to run without it.
- Go to claude.com/download
- Download the installer and click Next until the end
Windows only: Git is required
Claude uses Git in the background to keep track of your files so it can roll back its own changes. Without Git, Claude Code won't run on Windows. Git is free and takes a minute to install.
- Go to git-scm.com/downloads/win
- Download and install (Next → Next → Finish)
- Make sure "Add Git to PATH" stays checked (it's the default)
- Verify: open CMD (search "cmd" in Start) and type
git --version. If you see a version number, you're good.
Mac users: skip this. Git is already built in.
Once installed, open Claude Desktop for the first time. It will ask you to sign in or create an account. Follow the on-screen prompts.
Step 3: Pick your plan (start on Pro, upgrade to Max when you're hooked)
The free tier doesn't unlock Claude Code. You need a paid plan. Three options matter:
Doesn't work for Claude Code. You need a paid plan. Minimum is Pro.
The starting point for this challenge. Unlocks Skills, Routines, Connectors.
Where most people end up after a week. More credits, unlocks Auto Mode (see Step 8).
How to upgrade:
- Open Claude Desktop
- Click your profile icon (bottom-left corner)
- Pick Upgrade or Subscribe
- Choose Pro (or Max if you already know you want the headroom)
- Enter your card and confirm
A word on credits
Every plan has a credit budget that resets on a rolling window. The more you use Claude, the faster you'll feel the ceiling. If Pro starts to feel tight, Max gives you roughly 5× more room. My honest take: once you see what Claude Code can actually do, the upgrade pays for itself. Don't try to save on the one tool that will save you hundreds of hours.
Tip: set Sonnet as your default model to stretch your credits. Full model breakdown in Step 7.
Step 4: Claude works with folders
This is the single most important concept to grasp. Claude Code doesn't work "in the air." It connects to a specific folder on your computer.
Before every work session you pick a folder (File → Open Folder). Claude reads the files in it, writes to it, and loads the CLAUDE.md sitting inside.
That's why Donna "lives" in a specific folder (donna-starter-kit). Open that folder in Claude Code and Donna knows who she is and what she can do. Open a different folder and Claude won't recognize her at all.
In short:
Folder = project. Different folder = different project. The CLAUDE.md inside the folder is your instructions to Claude for that project.
Rule I live by: one folder per project. A folder for my business, a folder for each product launch, and one for Donna. Don't mix topics inside the same folder.
Step 5: Tour of the Claude Desktop sidebar
Inside the Code tab, the left sidebar holds everything you'll touch daily. The pieces worth knowing:
New session . opens a fresh conversation. Claude doesn't remember past chats, but it always reloads CLAUDE.md. Every new session starts clean with the memory from that file.
Pinned sessions . drag any session to pin it. Useful for a few important threads you keep returning to. The rest can be grouped by date or project.
Routines . this is where you schedule recurring tasks (morning briefing, evening summary, weekly report). Anthropic just overhauled this feature in the latest release. We configure three of them on Day 5.
Customize . settings, Skills, Connectors, Plugins. We open this next.
Step 6: Settings worth knowing
Get familiar with Settings before you start working. Open them via Customize in the sidebar, or your profile icon in the bottom-left corner.
The must-do:
Set Sonnet as your default model. Smart enough for almost everything and saves credits. Full model breakdown in the next step.
Custom Instructions (recommended)
Your global style guide that Claude follows for you. Things like "no em-dashes," "break complex tasks into steps," "answer in the language I'm writing in." I've published the exact set I use . copy it if you want a solid starting point:
Usage & Billing (good to know)
Shows how many credits you've burned this cycle and when they reset. Check it every few days until you get a feel for your rhythm. If you enable "extra use" (pay-per-credit), be careful . it can rack up fast if you forget about it. You can also toggle it off entirely for peace of mind.
Step 7: The Claude models
Claude ships in a few model tiers. Each one balances intelligence, speed, and credit cost differently:
Fast, light, cheap. Good for simple tasks like translation, a quick summary, or a one-shot question.
The balanced pick. Recommended default for this challenge. Handles 90% of what you'll do, easy on credits.
The strongest. Brand new release. Burns credits fast. Save it for heavy analysis, complex writing, or when Sonnet gets stuck.
To switch: type /model in a session, click the model name at the top of the chat, or change it in Customize → Settings.
Default to Sonnet to save credits.
Sonnet handles 90% of tasks. Move to Opus 4.7 only for deep analysis or complex writing.
Effort setting (optional read)
Next to the model picker there's an effort setting (low / medium / high). Low is fine for quick questions. High is for anything that needs real thinking: "analyze these 100 products," "refactor this skill," "plan a product launch."
From my hacks series: Hack #14: Three Brains, One Claude
Step 8: Permission modes (now with Auto Mode)
Every time Claude wants to do something (edit a file, run a command, send data somewhere), it asks for permission . unless you've told it not to. Five modes to know:
Claude asks before every change. Safest. Good for day one when you want to see each step before approving.
Claude edits files without asking, but still pauses before running commands. A solid middle ground.
Claude plans before it acts. Shows you the plan, waits for approval. Good for big tasks you want mapped out first.
Claude runs without asking. Fastest, but no safety net. Only use it when you know exactly what you're doing.
Claude runs without asking, but a separate classifier checks every action first and blocks anything risky . production deploys, sensitive data leaving your machine, mass deletions, forced git pushes. You get Bypass-level speed with a safety net. Anthropic added this in late March 2026 and it's what I use every day now.
Recommendation: start with Auto accept edits for a day or two to build trust. Once you're comfortable, switch to Auto Mode and let Claude flow.
On Pro? Read this before picking Auto Mode
Auto Mode isn't available on the Pro ($20) plan . it needs Max, Team, or Enterprise, and Sonnet 4.6 / Opus 4.6+ as the model. If you're on Pro for this challenge, use Auto accept edits as your default. When you upgrade to Max (and most people do), Auto Mode unlocks automatically.
Where to change it: click the icon in the bottom-left corner of the input bar (next to the send button).
Step 9: Useful slash commands
Claude Code has built-in shortcuts that start with /. Think of them as power buttons. The ones I use daily:
/compact . use it when a chat gets long and Claude starts to slow down. Compresses the conversation into a summary and keeps going sharp.
/clear . start a fresh session. Claude forgets everything (but still reads CLAUDE.md).
/model . switch model mid-session. No need to open settings.
/context . see how much context the session has left before it needs compacting.
/help . show every available command.
Try one now:
/help
From my hacks series: Hack #7: Session Getting Slow? Here's How to Fix It
Step 10: Golden tip . import your memory from ChatGPT
Optional . click to expand
Moving over from ChatGPT? Don't start from scratch. Claude can import the memory you've built up over there so it already knows things about you from minute one.
How it works:
- In ChatGPT: ask it to export the memory it has about you. It prints the list.
- Copy everything it gives you.
- In Claude Desktop: Customize → Import memory → paste the text.
Full walkthrough: Memory Import Guide
Note: this is experimental. It works best for work preferences and professional context, less so for personal details. Not required, but it saves time.
From my hacks series: Hack #1: Import Your Memory in 60 Seconds
Step 11: Golden tip . the coach prompt
Already built into Donna . expand to read
A trick I stumbled into: you can turn Claude into a personal coach. You paste a special prompt at the start of a project and Claude keeps working normally, but every so often it drops a tip that helps you learn.
For example, after finishing a task it might say: "Worth documenting this decision so we don't lose time on the same question next week." Or after a long chat: "This session is getting long. Consider /compact to save context."
It doesn't stop working to teach. It works and teaches in parallel, only when there's a real learning moment.
Good news
We already baked this prompt into Donna's CLAUDE.md. Donna ships with "coach mode" built in. Nothing to do on your end.
Step 12: Other ways to reach Claude
Reference . click to expand
Claude Desktop isn't the only way in. Worth knowing the others exist:
Mobile app (iOS / Android) . Chat mode only. Good for quick questions on the go. No Code tab, so Donna can't run from there directly.
Browser extension . Claude can read and work with web pages you're browsing. Useful for summaries and research.
Excel & PowerPoint plugins . Claude can read and analyze spreadsheets and generate slides. Free download from inside Claude Desktop.
Dispatch (new) . lets you control your Mac remotely from the Claude mobile app. Your computer runs the actual work, your phone just sends the instructions. Turn it on in Settings first.
For this challenge we stay in Claude Desktop, Code tab. But after the challenge, explore . there's a lot more surface area than you'd expect.
Bonus from the video: Amazon Landing Page Skill
Bonus skill . click to expand
In the demo section of the video I show a skill I built that turns any Amazon product URL into a full landing page . titles, price, real reviews, images, everything. Handy when you want to run Facebook ads and capture emails instead of sending people straight to Amazon.
What the skill does:
- Scrapes the Amazon listing: title, price, images, real reviews
- Builds a full HTML landing page in the background while you keep working
- Includes a CTA that links back to Amazon
- Can collect emails or add a coupon code if you ask for it
- Not locked to Amazon . tell Claude "same thing but for a Walmart listing" and it figures it out
How to use it:
- Download the skill from the page below
- Drop the file into your project folder. If you're not sure where, just ask Claude: "take this skill and put it in the right place."
- Run
/amazon-landing-pageand paste an Amazon URL - Ask for any tweak you want: coupon code, email capture, a different CTA, a different layout
→ Get the Amazon Landing Page Skill
A note on "scary" steps
GitHub, domains, hosting . a lot of people freeze at these words. You don't need to. Just ask Claude: "how do I connect this landing page to my domain?" or "can you do it for me?" . it'll walk you through it or take over your computer and handle it. Most of the things you think you need to do yourself, you don't.
Tips worth remembering
8 tips . click to expand
- Always work from the Code tab. Not Chat, not Cowork.
- Default to Sonnet to save credits. Move to Opus 4.7 only when you need the extra brainpower.
- If Claude asks clarifying questions, answer them. It's trying to understand what you actually want.
- If something isn't working, rephrase. "Build me X that does Y and save as Z" works almost every time.
- You can't break anything. Play, experiment, try things.
- Windows: if Claude Desktop refuses to run, double-check Git is installed (Step 2).
- One folder per project. Keep projects separate so memory doesn't get tangled.
- Got time? Skim my Claude Hacks Library . 20+ small wins you can apply today.
FAQ
10 common questions . click to expand
Checklist
Glossary
6 terms . click to expand
Terms you'll see throughout the challenge. You don't need to memorize them, just know they exist.
| Claude Desktop | The app you downloaded to your computer. Not claude.ai in the browser. Donna only works inside the app. |
| Vault | A folder Obsidian knows how to read. No magic . just a folder of text files displayed visually. |
| Skill | A command Donna knows how to run. Like /morning-briefing or /draft-reply. A text file with instructions she reads and executes. |
| Connector | A connection between Donna and an external tool (Gmail, Calendar, Drive). This is how she reads emails and sees meetings. |
| Routine | A task Donna runs automatically on a schedule. For example, a morning briefing every day at 9:15. Computer needs to be on. |
| Markdown | A simple text file format (.md extension). Every file Donna writes uses this format. Opens in any text editor. |
You're ready for the challenge.
When Day 1 opens, you'll find it in the next tab.
Meet Donna
Install the kit. Learn the method. Have your first conversation with your AI Hire.
Today: 4 Lessons
4 short videos, each with hands-on parts.
🎬 Lesson 1: Setup & Install
Get the kit on your Desktop · Install Obsidian · Load both apps · Verify Donna responds
🎬 Lesson 2: The Method
The 8 components · 3 things that make it special · How it splits across 5 days
🎬 Lesson 3: Tour the Kit
What Obsidian is · Donna walks you through every folder
🎬 Lesson 4: Talk to Donna
Her personality · 5 first conversations · Your first real win
Setup & Install
Get the kit on your computer. 4 quick parts.
Part 1A - Get the kit on your Desktop
⬇ Download donna-starter-kit.zip
- Unzip: Mac → double-click. Windows → right-click → "Extract All"
- Drag the
donna-starter-kitfolder to your Desktop (not Documents, not Downloads - Desktop, so you find it later)
Part 1B - Install Obsidian (free)
Skip if you already have it. We'll explain WHY we use it in Lesson 3.
- Go to obsidian.md → download for Mac/Windows
- Install like any normal app → open it (don't create any vaults yet)
Part 1C - Load the kit in BOTH apps
- Claude Desktop: Switch to Code tab (top center) → click the folder picker (bottom of chat input) → choose "Open folder..." → select your
donna-starter-kitfolder on Desktop - Obsidian: Click "Open folder as vault" → select the
donna-starter-kitfolder → Open
Same folder. Two windows. Claude = where you talk to Donna. Obsidian = where her memory is visible.
Part 1D - Verify
In Claude Desktop, type:
What folder are we in? List the main folders.
✅ Success:
Donna lists the folders correctly: .claude/, brain/, guides/, samples/, templates/, output/. This confirms the kit loaded properly.
If she can't see the folders: Re-select the folder via the picker. Start a new conversation. Try again.
Note: This question doesn't necessarily trigger Donna's full personality (it's a factual question). You'll see her personality come alive in Lesson 4.
The Method
Most people use AI like a chatbot. We're building something different: an AI Hire. She knows who you are, sounds like you, and does things without you.
Part 2A - The 8 Components
Eight components make Donna work. Each lives somewhere specific in your kit.
1. Identity - Who Donna is. Her personality, voice, values.
2. Memory - Who YOU are. Your business, people, decisions.
3. Voice - How YOU write, so Donna sounds like you.
4. Access - Connection to your real tools (Gmail, Calendar, etc.).
5. Skills - One-command shortcuts (she ships with 14, ranging from morning briefings to onboarding).
6. Quality - Self-check before every output.
7. Routines - Automation. She runs on a schedule.
8. Learning - She gets smarter every day.
📌 Quick recap: Identity, Memory, Voice, Access, Skills, Quality, Routines, Learning. We'll build all 8 over the 5 days.
Part 2B - 3 Things That Make It Even More Special
The 8 components are the foundation. These 3 things take the method to the next level - they're what students keep coming back for.
⚡ Personality is mandatory - Every output sounds like Donna. Never generic AI.
🛡️ Quality gate before every output - Self-Check Gatekeeper catches vague language and missing details.
📈 She compounds over time - After 2-3 weeks she sounds significantly more like you.
Part 2C - How It Splits Across 5 Days (The Map)
Just so you know where we're going. Don't try to memorize - we'll walk through each day as we get there.
| Day 1 | Meet Donna - today (setup + first conversation) |
| Day 2 | Donna learns about your business |
| Day 3 | Connect Gmail, Calendar, Drive, Tasks |
| Day 4 | Test all 12 skills |
| Day 5 | Donna goes autonomous (works while you sleep) |
Tour the Kit
Get familiar with the kit's structure. We'll do this through Donna - that's the pattern you'll use for everything.
Part 3A - What Obsidian Is
Obsidian is a free app that shows your files visually. Your brain/ folder is an Obsidian vault. That's where Donna's memory lives.
Quick note on .md files . Every file in this kit ends with .md (Markdown). It's just plain text with light formatting (headers, lists, bold). Think Word without the bloat. Both Obsidian and Claude read it natively. You don't need to learn the syntax. Donna writes it for you.
⚡ Important: You don't need to learn Obsidian.
Claude does all the work with Obsidian. You just tell Claude what you want:
"Update Obsidian with this fact about Mike."
"Check Obsidian for my Q3 goals."
"Show me what's in brain/People."
Claude does it all. You never need to navigate Obsidian unless you want to.
📂 Quick visual look:
Open Obsidian now. Look at the sidebar on the left. You'll see folders:
- brain/Business/ - your goals, promises, decisions
- brain/People/ - your contacts (empty for now)
- brain/Daily Logs/ - daily briefings
- brain/Preferences/ - your style
Everything is empty right now. By Day 5 it'll be full. That's the whole challenge - filling Donna's brain.
💡 Curious? Click around in Obsidian if you want. Press Cmd+G for Graph View - almost empty now, but on Day 5 it becomes a wow moment. Optional sightseeing.
Part 3B - Let Donna Walk You Through the Files
Instead of clicking around yourself, let Donna explain the structure. This is the pattern you'll use for everything: ask Donna, get an answer.
Type this in Claude Desktop:
Walk me through every folder in this kit. Tell me what each one is for and what I'll find inside. Be brief but complete.
Donna will describe everything: CLAUDE.md, .claude/skills/, brain/Business/, brain/People/, templates/, samples/, output/, and guides/.
Then ask follow-ups:
What's the difference between brain/Business/ and brain/People/?
What's inside .claude/skills/?
What's the difference between templates/ and output/?
💡 The lesson here
This is how you'll learn the kit going forward. You don't memorize the folder structure. You ask Donna. She knows. This is your new pattern: "Donna, show me X" / "Donna, explain Y."
Talk to Donna
Now you meet Donna's personality and have your first real conversations with her. By the end of this lesson, you'll have used her for actual value - not just tested her.
Part 4A - Meet Donna's Personality
Donna is named after Donna Paulsen from the TV show Suits. Sharp. Confident. Theatrical. Fiercely loyal. Never sugarcoats anything. That's the energy your AI is built on. (2-min taste 👆 if you've never seen the show.)
Open Claude Desktop and ask her this:
Walk me through your personality, Donna. Your tone, your style, your humor, the way you handle a conversation. Give me real examples, not a description.
Read her answer slowly.
Notice the tone, the confidence, the humor. This isn't ChatGPT. This is Donna.
💬 Feel free to keep chatting with her
Ask whatever you want. Test her. Push her. Get a feel for who she is before we move on.
⚠️ Donna's personality is FIXED
You don't customize how she talks to you. She's direct, sharp, funny. That's the point.
What IS customizable: how Donna writes AS you (emails on your behalf). That's Day 2.
Part 4B - 3 Quick Wins You Can Get Today 🎯
Donna doesn't know your business yet. That's Day 2. But she can already give you real value today. Pick one (or all three) and try it now:
1. The decision you've been putting off
I have a decision I've been avoiding. Let me dump it on you, then walk me through it.
2. The tough conversation you're dreading
There's a conversation I need to have and I keep pushing it off. Help me prep for it.
3. The week that feels chaotic
Brain dump my next 7 days with me, then turn it into an actual plan.
🎯 This is real value, today
Tomorrow she'll respond with YOUR business context loaded in. The difference will surprise you. But what you get today is already worth the price of admission.
FAQ
Show me the contents of .claude/donna-character.mdHomework: Prep for Day 2
Tomorrow Donna learns everything about your business. The quality of Day 2 depends entirely on this prep. Open a notebook or text file and write down:
- Your full name + nickname Donna should use
- Your role (Founder, CEO, Freelancer, etc.)
- Where you live (city, time zone)
- Your business in 2-3 sentences
- Top 3 priorities this quarter
- 5-10 key people (name, role, one line about each)
- 3-5 actual emails you wrote . copy/paste into a text file. this is what teaches Donna your voice
- How you greet and sign off emails
- 3 words you use a lot, 3 words/phrases you HATE
Want to give Donna even more to work with? (optional)
- 3-5 things you've promised people recently
- 3-5 things people promised you that you're waiting for
- Current state of business (revenue range, biggest challenge)
- Tools you use (Gmail/Outlook, Asana/Notion, etc.)
Skip if you're tight on time. The list above is enough.
⚠️ Don't skip the homework
Tomorrow Donna reads your prep and fills out your brain automatically (about-me, business, goals, voice style, key people, email samples). If your prep is sparse, Donna stays generic. Solid homework = a Donna that knows you. Skipping it = generic AI.
Checklist
Donna, Meet My Business
Today Donna stops being generic and starts being yours.
Today: 4 Lessons
4 short videos. By the end, Donna knows your business and writes in your voice.
🎬 Lesson 1: The Big Idea
Why Memory + Voice are the components that turn a generic AI into YOUR AI.
🎬 Lesson 2: How /learn-me Works
The skill that does all the writing for you. Two modes: dump or interview.
🎬 Lesson 3: Run /learn-me
You dump your homework into chat. Donna fills 6 categories of files in your brain.
🎬 Lesson 4: The Wow Moment
Run /business-one-pager. Donna writes a one-pager + a sample email in your voice.
The Big Idea (Memory + Voice)
Generic AI starts every conversation from zero. You re-explain your business every single time. That's exhausting and forgettable. Today we fix that permanently.
Part 1A. What "Memory" means here
ChatGPT is stateless. You = a stranger every time. Donna with a filled brain/ folder is the opposite: your business, your goals, your people are loaded into her at the start of every conversation. This is component #2 of Jay's AI Hire Method.
Part 1B. What "Voice" means here
Voice is HOW you write. word choices, tone, sentence length, things you'd never say. When Donna drafts emails AS you tomorrow, she should sound like you. Not like an AI. This is component #3.
Part 1C. The "AI Hire vs ChatGPT" moment
Day 2 is the day Donna becomes YOUR Donna. Not someone else's. Not generic. Yours.
When she nails your voice in tomorrow's drafts, you'll feel the difference. That's the magic of today.
How /learn-me Works
Concept first. you'll actually run it in Lesson 3.
Part 2A. The principle: you talk, Donna writes
- You don't open Markdown files
- You don't fill templates
- You don't navigate Obsidian
- You type one command in Claude.
/learn-me. That's it.
Donna drives the conversation. She receives, she writes, she shows you the result.
Part 2B. Two modes: dump or interview
📤 Dump mode (default)
"Send me everything you prepared. Across as many messages as you want. Say DONE when finished."
💬 Interview mode
Say "walk me through it" and Donna asks one question at a time. Useful if you didn't prepare or prefer a back-and-forth.
You can switch mid-flow. Just tell her what you want.
Part 2C. What gets filled behind the scenes
Donna writes 6 categories of files automatically:
brain/Business/about-me.md. your raw wordsbrain/Business/my-business.md. Donna's organized readbrain/Business/Goals.md. your top prioritiesbrain/Preferences/Style.md. your voice patternsbrain/People/[name].md. one file per key personsamples/email-N.md. your email examples saved
Run /learn-me
Time to actually do it. Open Claude Desktop, Code tab, kit loaded.
Part 3A. Type the command
/learn-me
Donna kicks off with a clear ask: "Send me what you prepared, in any number of messages, in any order. Say DONE when you're finished."
Part 3B. Dump your homework
Open the notes you prepared in Day 1 (or just type freely). Send it across as many messages as you want, in any order. Examples:
- Message 1: "My name is Sarah, I run a wholesale real estate company in Dallas..."
- Message 2: "Top 3 priorities: hire a closer, hit $500K in Q2, automate dispositions..."
- Message 3: "Key people: Mike (acquisitions partner), Lauren (VA), David (lawyer)..."
- Message 4: [paste 3 actual emails you wrote]
Donna acknowledges each message in 1-2 lines. She's just collecting. She doesn't write any files yet.
💡 Flexibility note
Bringing extra info we didn't ask for? Great, Donna takes it. Don't have something we asked for? Fine, she'll skip it. No rigid format. Each owner and business is different.
Part 3C. Say DONE and watch her work
DONE
Donna processes everything, fills 6 categories of files in your brain, and shows you a clean summary in chat. If something critical is missing, she asks gently before writing.
Then open Obsidian and look:
brain/Business/my-business.md. now filledbrain/Preferences/Style.md. now filled with voice analysisbrain/People/. one file per person you mentionedsamples/. your emails saved
You just talked to Donna. She wrote 6+ files. You never opened a single Markdown file. That's the operator approach.
The Wow Moment
She knows you. Now make her prove it.
Part 4A. Run /business-one-pager
/business-one-pager
Donna writes a one-page business summary AND drafts a sample email in your voice (a "voice test"). Both appear shortly after you send the command.
Part 4B. Open the one-pager
Switch to Obsidian and open output/business-one-pager.md.
It includes: what you do, who you serve, your offer, key numbers (if you shared any), top priorities, key people, AND a "Strategic Read" section where Donna shares her honest interpretation of where your business is and what to focus on. Not generic AI hedging. Her actual opinion.
Part 4C. The voice test
Look at the sample email Donna drafted in chat. Compare it to the actual emails you sent her in Lesson 3.
- Greeting matches?
- Sign-off matches?
- Sentence length matches?
- Tone matches?
- None of your "HATE" words?
If it sounds like you. Day 2 worked.
Donna is now YOUR Donna. Tomorrow we connect her to your real Gmail and Calendar so she stops working with hypothetical data and starts working with your actual life.
If something feels off, just tell Donna what to fix and run /learn-me again to update.
FAQ
/learn-me again anytime to add more. she'll treat it as an update, not a wipe./learn-me again./learn-me again with more detail and re-run /business-one-pager./learn-me again.Homework: Prep for Day 3
Tomorrow Donna gets connected to your real tools. Gmail, Calendar, Drive, and your task manager. Quick prep:
- Make sure you're logged into the Google account you want Donna reading from (Gmail + Calendar + Drive)
- If you use a task manager (Asana, Monday, ClickUp, Notion), have your login ready
- If you use Outlook instead of Gmail. note it. we'll discuss alternatives in Day 3
⚠️ Privacy note
Donna only READS by default. She never sends, deletes, or moves anything without your approval. Day 3 sets up read access. you stay in control of every action.
Checklist
Badge & Access
Today Donna gets her badge. Real emails, real calendar, real documents.
Today: 4 Lessons
4 short videos. By the end, Donna can read your real inbox, see your real calendar, search your real Drive, and prep meetings using all three at once.
🎬 Lesson 1: The Big Idea
Why Access matters + the privacy promise (read first).
🎬 Lesson 2: Connect Google
Gmail + Calendar + Drive in one auth flow. Verify each.
🎬 Lesson 3: Connect Your Task Manager (optional)
Asana, Monday, ClickUp, Notion. or skip if you use an industry-specific tool.
🎬 Lesson 4: The Wow Moment
Real email triage + draft reply in your voice + meeting prep from 3 sources at once.
The Big Idea (Access + Privacy)
Until today, Donna only knew what you told her on Day 2. She had your business profile, your voice, your people. but she couldn't actually SEE anything. Today she gets a badge.
Part 1A. What "Access" means here
This is component #4 of Jay's AI Hire Method. Once you connect Gmail, Calendar, and Drive, Donna can read your real inbox, see your real meetings, search your real Drive. Every skill from now on runs on YOUR data, not hypotheticals.
Part 1B. The Privacy Promise (read carefully)
🛡️ Day 3 is the trust moment. Here's what's guaranteed:
- ✅ Donna READS emails, calendar, files. she never sends, deletes, or modifies on her own
- ✅ She DRAFTS email replies. you approve before send
- ✅ She SUGGESTS calendar events. you approve before create
- ✅ Connectors grant read access by default. anything that creates or sends needs explicit approval
- ✅ You can revoke any connector instantly: Settings → Connectors → Disconnect
Donna never deletes anything. Ever.
The connection is between Claude Desktop and your Google account directly. nothing routes through third-party servers we don't control.
Part 1C. What gets connected today
Donna reads inbox, categorizes, drafts replies
Donna sees meetings, preps you, flags conflicts
Donna searches docs, surfaces relevant info
Asana, Monday, ClickUp, or Notion
Connect Google (Gmail + Calendar + Drive)
Part 2A. Open the Connectors menu
In Claude Desktop (the app, not the browser):
- Click the gear icon (⚙️) in the bottom-left corner of the sidebar
- Click Connectors (sometimes labeled Integrations)
- You'll see a list of available connectors
Reminder: Connectors only work inside the Claude Desktop app you downloaded. Not at claude.ai in the browser.
Part 2B. Connect Gmail
- Find Gmail in the list → click Connect
- Sign in with your Google account
- Grant the permissions requested (read access)
Multiple Google accounts? Connect the work one first. You can add a second later.
Verify:
What email address is connected? Show me my 5 most recent unread emails.
If Donna lists your real emails, Gmail is in.
Part 2C. Connect Calendar + Drive
Often these auto-connect with the same Google auth from Gmail. If not:
- Settings → Connectors → Google Calendar → Connect
- Settings → Connectors → Google Drive → Connect
Verify Calendar:
What's on my calendar today? What about tomorrow?
Verify Drive:
Search my Drive for [pick any document name you know exists]
Drive search slow? First search can take a moment to index. Try a unique phrase from inside the doc if a name doesn't work.
Connect Your Task Manager (optional)
This step is optional. Donna works fine without one. But 4 of her daily skills get sharper with a connected task manager: /tasks, /accountability, /block-time, and the task section of /morning-briefing.
Path A. You already use Asana, Monday, ClickUp, or Notion
- Settings → Connectors → search for your tool → Connect
- Sign in to that tool
- Grant access to your workspace
Verify:
Show me my incomplete tasks
Path B. You don't use a task manager yet
If you want to add one (recommended), pick a free option:
- Asana (best for solo). asana.com/free
- Notion (most flexible). notion.so
- ClickUp (most features). clickup.com
Create a free account, add 5-10 of your current tasks, then come back and connect via Path A.
Path C. You use an industry-specific tool that isn't supported
Yoga/fitness (Mindbody, ClassPass), real estate (Follow Up Boss, Boomtown), coaching CRMs, etc. Most aren't supported as Claude connectors.
✅ Skip the connection. you're fine.
Donna falls back to using brain/Business/Promises.md for tracking commitments. 10 of her 12 daily skills still work fully.
You can always add a supported tool later (e.g., Notion for personal action items separate from your industry tool).
The Wow Moment (real value)
She has the badge. Now use it.
Part 4A. Email triage with real data
In Claude:
Read my unread emails from the last 24 hours. Sort them into URGENT, IMPORTANT, and FYI. Flag any with dollar amounts or deadlines.
Donna pulls your actual inbox. Categorizes. Highlights what you'd want to act on. First time you see her processing real life.
Part 4B. Draft a reply in your voice
Pick one real email that needs a response. Tell Donna:
Draft a reply to [name or short description of the email]. Match my communication style.
Donna reads the email, pulls your voice patterns from brain/Preferences/Style.md, drafts a response using your greeting and sign-off. You approve before sending.
Part 4C. The "Chief of Staff" moment
Prep me for my next meeting. Pull everything relevant from email, calendar, and Drive.
Donna identifies who you're meeting, scans email history with that person, checks shared calendar context, searches Drive for related documents, and synthesizes a brief.
This is the moment.
She goes from "AI assistant" to "Chief of Staff who actually knows what's going on". Three data sources, one brief, one prompt.
📅 Service business / mostly recurring slots?
If your calendar is mostly classes, sessions, or recurring time blocks (yoga, fitness, content creation, etc.), use this instead:
Show me my schedule for the next 7 days. Flag anything that needs prep. guest teachers, new client intros, one-off appointments, anything outside my usual rhythm.
FAQ
guides/mcp-setup-guide.md in the kit for the advanced setup. The simplest path for the challenge is to forward emails you want Donna to see to a Gmail address./tasks, /accountability, /block-time, task section of /morning-briefing) still work, just less polished. Donna falls back to brain/Business/Promises.md for tracking commitments.Homework: Prep for Day 4
Light prep tomorrow. Day 4 is Skills. you'll learn to call Donna with one-command shortcuts (/morning-briefing, /forgotten-emails, etc.) using the data you connected today.
Just confirm tomorrow morning that:
- Connectors are still active (Gmail, Calendar, Drive open in Claude's Connectors panel)
- You're logged into the same Google account from today
No new prep. Just don't disconnect overnight.
Checklist
Training Day
Today you train Donna's reflexes. One command instead of a paragraph.
Today: 4 Lessons
4 short videos. By the end, you'll have run all 12 daily skills, customized one for your business, and identified your Daily 4.
🎬 Lesson 1: The Big Idea
What a Skill is + where they live + the 12 daily skills overview.
🎬 Lesson 2: Run Your First Skill
/morning-briefing. the headline. one command, full morning routine.
🎬 Lesson 3: Test the Skill Library
5 quickfire skill demos to feel the range.
🎬 Lesson 4: Customize + Daily 4
Open a skill, add ONE rule for your business, identify the 4 you'll use forever.
The Big Idea (Skills + Quality)
Part 1A. What a Skill is
Until now you've been talking to Donna in full sentences. A Skill is the shortcut. instead of a paragraph, you type one command.
❌ Without a skill
"Hey Donna, read my emails from the last 24 hours, sort them by urgency, check my calendar for today, look at my tasks, scan for risks, and give me a summary with my top priority."
✅ With a skill
/morning-briefing
Done. Same job.
This is component #5 of Jay's AI Hire Method. Skills.
Part 1B. Where Skills Live
Skills are just text files (.md) inside .claude/skills/. The filename (without .md) is the command:
morning-briefing.md→/morning-briefingdraft-reply.md→/draft-reply [name]
You don't need to look inside yet. just know they're plain text. We'll open one in Lesson 4.
Part 1C. The 12 daily skills
| Command | What it does |
|---|---|
/morning-briefing | Daily morning routine. emails, calendar, tasks, risks, #1 priority |
/evening-summary | End-of-day wrap. done, not done, tomorrow's plan |
/draft-reply [name] | Draft a reply in your voice |
/prep-meeting [name] | Full meeting brief with agenda |
/tasks | Task overview from your task manager |
/accountability | Track promises (yours and others') |
/forgotten-emails | Missed threads in both directions |
/block-time [task] | Calendar time for things you keep postponing |
/risks | Identify business risks before they explode |
/weekly-report | Weekly summary. wins, misses, learnings |
/learn | Improvement loop (auto in evening) |
/ask-jay [q] | Help with anything, always available |
Plus the 2 onboarding skills you ran on Day 2 (/learn-me + /business-one-pager). 14 skills total.
Heads up: A few skills (/weekly-report, /evening-summary, /learn) will feel sparse today. they need internal logs to build on. They'll get richer after a week of use. Today we just confirm they run.
Run Your First Skill (`/morning-briefing`)
This is the headline skill. It pulls from everything we've built so far: brain/, Gmail, Calendar, Drive, optional task manager. one prompt, full morning routine.
Part 2A. Type the command
/morning-briefing
No need to add anything else. just send.
Part 2B. Watch Donna work
She'll pull from multiple sources at once:
- Yesterday's Daily Log (carries forward unfinished items)
- Unread emails from the last 24 hours (categorizes urgency)
- Today's calendar (flags conflicts and prep needs)
- Your task manager (groups by overdue/today/this-week)
- Promises in
brain/Business/Promises.md - Risks (deadline + task gap, stuck items)
Then she synthesizes the #1 priority for the day.
Part 2C. Read the output. Notice the Quality
Look for these in her output:
- ✓ Donna opening line (personality, not generic AI)
- ✓ Sections with emoji headers (🔴 URGENT, 📅 CALENDAR, ✅ TASKS, ⚠️ RISKS, 🎯 #1 PRIORITY)
- ✓ Specific numbers and names (not "a few emails" but "3 emails. one from Mike Johnson...")
- ✓ Donna closing line
🛡️ Behind the scenes: the Self-Check Gatekeeper
This is component #6. Quality. Every skill runs a Self-Check before showing output. catches vague language, missing numbers, missing personality. If anything fails, Donna fixes it before showing. You never see the failure.
Output also saves to brain/Daily Logs/[YYYY-MM-DD].md. so tomorrow's /evening-summary and /learn have something to reference.
Test the Skill Library (5 quickfire)
Run 5 different skills back to back. Each one is a different kind of help. don't try to memorize. just feel the range.
Part 3A. /draft-reply [name]
Pick one real email that needs a response.
/draft-reply [sender's name or short email description]
A draft using your greeting, sign-off, and tone (from brain/Preferences/Style.md). Flags any dollar amounts. Does not send.
Part 3B. /forgotten-emails
/forgotten-emails
Two lists. emails YOU haven't answered (3+ days), and emails OTHERS owe you. Empty list = win.
Part 3C. /prep-meeting [name]
Pick someone you have a real upcoming meeting with.
/prep-meeting [name or meeting title]
A brief: who they are, last interaction, open items, what they probably want, your suggested agenda, watch-outs.
Part 3D. /risks
/risks
Categorized 🔴 HIGH / 🟡 MEDIUM / 🟢 LOW with action items. "All clear" is also a valid answer.
Part 3E. /accountability
/accountability
Status report on every promise. yours and theirs. Goal progress with traffic-light status, patterns, where you stand.
📌 The other 7 daily skills
Test as homework or call them when you need them: /tasks, /block-time, /evening-summary, /weekly-report, /learn, /ask-jay [question].
Customize a Skill + The Daily 4
Part 4A. Open a skill file
Skills are just text. Let's prove it.
Show me the contents of .claude/skills/morning-briefing.md
Or open it directly in Obsidian. You'll see Markdown: a Meta table, a Steps section (what Donna does in order), an Output Format template, Opening + Closing line banks, a Self-Check checklist. No code. No magic. Just instructions she follows.
Part 4B. Add ONE business-specific rule
Pick the skill you'll use most. Add ONE rule. Tell Donna what to add and where, and she'll edit the file. Or open in Obsidian and edit yourself.
Examples by business type:
- E-com: "Check our top SKU's recent reviews. Flag any 1-3 star review."
- Service / Solo: "Check tomorrow's class roster. Flag low attendance or new client intros."
- Real estate: "Pull pipeline status. flag deals stuck in same stage 14+ days."
- B2B / Agency: "For each active client project, summarize the latest status."
Save. Run the skill again. Your custom rule appears in the new output.
💡 You can also create your own skills. Just add a new .md file in .claude/skills/ and Donna picks it up. Tell her "create a new skill for X" and she'll draft it with you.
Part 4C. Your Daily 4
14 skills total. You won't use them all every day. Your Daily 4 is the rhythm of your week:
| Skill | When |
|---|---|
/morning-briefing | Daily, sets your day |
/draft-reply | Whenever an email needs a response |
/evening-summary | Daily wrap (auto via Routine on Day 5) |
/risks | Weekly check-in |
Tomorrow on Day 5 we set up Routines so Donna runs the Daily 4 automatically. You wake up to a briefing waiting. You get a wrap-up at the end of the day. You get a weekly review on Sunday. without you asking.
FAQ
.claude/skills/ and confirm the file is there./weekly-report and /evening-summary look thin. Why?/draft-reply, or skip a skill if it has nothing to act on. The skill itself is fine. just nothing to chew on right now..md file from .claude/skills/. The command stops working. Or move it to a .disabled subfolder if you want to keep it for later..md file in .claude/skills/. The filename (without .md) becomes the command. Best way: tell Donna "create a new skill for X" and she'll draft it with you, copying the structure of an existing skill.Homework: Prep for Day 5
Light prep. Day 5 sets up Routines so the Daily 4 runs automatically. Just decide what times you want:
- Morning routine.
/morning-briefingat what time? (e.g. 7:00 AM, 8:30 AM) - Evening routine.
/evening-summaryat what time? (e.g. 9:00 PM) - Weekly review.
/weekly-reporton Sunday morning at what time? - Optional:
/prep-meeting1 hour before each calendar meeting (auto-trigger)
That's the prep. you'll set up the actual schedule on Day 5.
Checklist
/morning-briefing ran successfully with real data
.md file (saw it's just text)
She Works While You Sleep
The final day. Donna becomes autonomous.
Today: 4 Lessons
4 short videos. By the end, Donna runs on autopilot. morning briefings, meeting prep, evening summaries, weekly reports. all without you opening Claude.
🎬 Lesson 1: The Big Idea
Routines + Learning. components #7 and #8.
🎬 Lesson 2: Set Up the 4 Routines
Morning, meeting prep, evening, weekly. one-time setup.
🎬 Lesson 3: The Learning Loop
/weekly-report + /accountability + how /learn runs in the background.
🎬 Lesson 4: Graduation
Graph View, the reflection, what you built. you graduate.
The Big Idea (Routines + Learning)
Until today, Donna has been waiting for you. You open Claude. You type a command. She runs it. She goes back to sleep.
Today she stops waiting.
Part 1A. What "Routines" mean
Routines let Donna run on a schedule, automatically, without you opening Claude. Morning briefing at 7am. Meeting prep at 6pm. Evening summary at 9pm. Weekly report Sunday morning.
This is component #7 of Jay's AI Hire Method. Routines.
Pro plan limit: 5 Routines per day. Donna uses 4. you have 1 left for whatever else.
Computer off when scheduled? Queues. Only the most recent fires when you open Claude. no spam.
Part 1B. What "Learning" means
Donna doesn't stay the same. Every evening, she runs /learn. She reviews what worked, what didn't, and updates her own memory.
- After a week. drafts sound more like you
- After a month. she anticipates patterns
- After 3 months. nothing like the Donna you met on Day 1
Component #8 of Jay's AI Hire Method. Learning.
Part 1C. The "she works while you sleep" moment
Day 5 is the day Donna stops being a tool you use and starts being a teammate who works for you.
Morning briefing waiting before your coffee. Meeting prep arrives the night before. Weekly report lands in your Daily Log every Sunday. You don't summon her anymore. She just shows up.
Set Up the 4 Routines
Part 2A. Where Routines live
In Claude Desktop, in the Code tab, look at the left sidebar. Between Search and Dispatch, you'll see Routines.
Don't see it? Help → Check for Updates first.
Click Routines to open the panel. Then click + New Routine for each one below.
Part 2B. Routine 1. Morning Briefing (7:00 AM)
- Name: Morning Briefing
- Schedule: Every day at 7:00 AM
Prompt:
Run /morning-briefing
After generating the briefing, save it to brain/Daily Logs/ with today's date as the filename (YYYY-MM-DD.md format).
Part 2C. Routine 2. Meeting Prep (6:00 PM)
- Name: Meeting Prep
- Schedule: Every day at 6:00 PM
Prompt:
Run /prep-meeting for every meeting on tomorrow's calendar. Save all briefs to brain/Daily Logs/ under a "## Meeting Prep" heading.
Part 2D. Routine 3. Evening Summary (9:00 PM)
- Name: Evening Summary
- Schedule: Every day at 9:00 PM
Prompt:
Run /evening-summary
After generating the summary, append it to today's Daily Log in brain/Daily Logs/ under a "## Evening Summary" heading.
Part 2E. Routine 4. Weekly Report (Sunday 9:00 AM)
- Name: Weekly Report
- Schedule: Every Sunday at 9:00 AM
Prompt:
Run /weekly-report and save to brain/Daily Logs/[today's date]-weekly.md
💡 Pick whatever times fit your real schedule. The defaults above are a starting point.
✅ You now have 4 routines on autopilot.
That's the "she works while you sleep" promise. starting tomorrow morning.
The Learning Loop
Part 3A. Run /weekly-report manually (preview)
/weekly-report
Expected today: partial output. you only have 4-5 days of Donna data, not a full 7. By next Sunday, full power.
Output covers: wins, misses, numbers, promise tracking, patterns, next week's plan, and what Donna learned this week.
Part 3B. Run /accountability
/accountability
Honest assessment of what you're tracking, what's stuck, where you're slipping. Run it weekly. or whenever you feel things slipping.
Part 3C. How /learn runs in the background
You don't run /learn directly. The Evening Summary routine runs it automatically as part of its end-of-day work.
It reviews the day, notices patterns, and updates brain/Learning/learning-log.md.
After 30 days, that file becomes the most valuable thing in your kit. it's Donna's accumulated wisdom about you.
Graduation
Part 4A. Open Graph View in Obsidian
Open Obsidian. Click the graph icon in the left sidebar (or press Cmd+G on Mac, Ctrl+G on Windows).
This is Graph View. Every wiki link Donna created becomes a visible connection. people, decisions, goals, daily logs.
Today vs Day 30: Sparse today (10-20 nodes). Dense by Day 30. Take a screenshot now and another at Day 30. The difference is dramatic.
Part 4B. The Reflection
Open output/business-one-pager.md (Donna wrote it on Day 2). Read it again.
Then ask Donna:
Compare what you knew about me on Day 2 vs what you know now. What's changed? What's sharper?
Save her response somewhere safe. This is your before/after. In 30 days, ask the same question. The growth will surprise you.
Part 4C. What you built
| What | Details |
|---|---|
| Identity + Memory | Donna knows your business, style, priorities, and her own personality |
| 4 Connectors | Gmail, Calendar, Drive, Task Manager (or skipped) |
| 14 Skills | 12 daily skills + /learn-me + /business-one-pager |
| 4 Routines | Morning (7am), Meeting Prep (6pm), Evening (9pm), Weekly (Sun 9am) |
| Memory System | Daily Logs, People, Goals, Promises, Decisions, Learning Log |
| Self-Check Gatekeeper | Every skill verifies its own output before delivering |
| Learning Loop | /learn updates her memory every evening, automatically |
This is a working AI Chief of Staff.
She costs $20/month. She works on schedule. She remembers everything. She gets smarter every day.
🎓 You hired your first AI employee. Congratulations.
FAQ
/weekly-report looks thin today.Day 6 and Beyond
No homework. you graduated. Tomorrow Donna runs on her own. Here's the rhythm:
Tomorrow (Day 6)
- Morning briefing arrives at 7am (or when you open Claude after)
- Read it in Obsidian:
brain/Daily Logs/[date].md - Reply to Donna with feedback if anything's off. trains
/learn - Use
/draft-replythroughout the day - Evening summary arrives at 9pm
First Sunday
- Weekly report arrives at 9am
- Block 30 min to read it
- Adjust Goals.md if priorities shifted
First month
- Watch
/learnslowly improve voice matching - Add new People as Donna encounters them
- Update Goals and Promises when priorities shift
- Compare Graph View at Day 30 vs Day 5 screenshot
Checklist
/weekly-report + /accountability (preview)
You finished the challenge.
You built a real AI employee. Not a chatbot. Not a toy. A complete system that works for you.
Share with the community: I finished!