1 / 75

Public, five-module slide masterclass

Claude MasterclassChat, Chrome, Cowork, Code, and Skill Forge

Go from your first Claude conversation to reusable AI workflows and shareable skills. Work one lesson at a time, with every prompt, screenshot, checkpoint, and download kept in context.

5 modules45 lessonsNo coding background needed

The finish line

One useful workflow you can run, improve, and share

You will know which Claude surface fits the task, how to keep a human in control, and how to package your strongest process as a reusable skill.

Progress is saved in this browser.

Public, five-module slide masterclass

Claude MasterclassChat, Chrome, Cowork, Code, and Skill Forge

Go from your first Claude conversation to reusable AI workflows and shareable skills. Work one lesson at a time, with every prompt, screenshot, checkpoint, and download kept in context.

5 modules45 lessonsNo coding background needed

The finish line

One useful workflow you can run, improve, and share

You will know which Claude surface fits the task, how to keep a human in control, and how to package your strongest process as a reusable skill.

Progress is saved in this browser.

Before you begin

Four things make the Masterclass easier.

A computer and internet accessThe lessons work on macOS or Windows. You do not need a coding background.

A Claude accountFree is enough for Chat and limited Projects. Paid access unlocks Chrome, Cowork, and Claude Code.

Google ChromeModule 2 uses the Claude for Chrome extension and site-by-site permissions.

One real task to improveFollow along hands-on with work you actually repeat. The result will be more useful than a practice-only demo.

What you will be able to do

Move from asking Claude questions to building dependable AI capability.

Hold a productive Claude conversation and write prompts that produce usable answers.

Share files and images, then organize ongoing work with Projects, instructions, and memory.

Use Claude for Chrome safely while approving access one site at a time.

Connect Claude to tools like Gmail and Calendar so it can act, not just advise.

Build and improve repeatable workflows with Skills, connections, and a scorecard.

Use Claude Code for practical non-technical work, from websites to document drafting.

Install the meta-prompting toolkit and reuse strong prompts with slash commands.

Use Skill Forge to build, score, package, and share a reusable Claude skill.

Choose the right Claude surface for a task and protect your data deliberately.

The course stays practical: every lesson ends with an observable checkpoint, and your progress is saved in this browser.

The Masterclass

Move from conversation to reusable AI capability.

Module 1 · BeginnerClaude ChatEveryday chat and promptingSteps 1–9 Module 2 · BeginnerClaude for ChromeBrowser use, with Claude acting on websitesSteps 10–16 Module 3 · IntermediateClaude CoworkAgentic desktop workflowsSteps 17–28 Module 4 · AdvancedClaude CodeClaude Code, plus reusable prompts and skillsSteps 29–40 Module 5 · AdvancedSkill ForgeTurn any workflow into a reusable, shareable skillSteps 41–45

Module 1 · Beginner

Everyday chat and prompting

Claude Chat

This is the version most people meet first. It is a website (claude.ai) and a desktop app where you type questions and Claude types back. By the end of this module you can sign in, hold a useful conversation, share files, and keep your work organized.

Step 1Meet Claude and sign in Step 2Take a tour of the chat screen Step 3Write a prompt that gets a good answer Step 4Share files and images, then ask about them Step 5Use Artifacts for documents you can see and edit Step 6Set up a Project and custom instructions Step 7Install the Adaptive Reasoning Protocol custom instruction Step 8Understand memory and history Step 9Check your privacy settings

Step 1 of 45 · Claude Chat · Beginner

Create an account and understand free versus Pro.

Meet Claude and sign in

Claude is an AI assistant made by Anthropic. You talk to it in plain language, the same way you would message a knowledgeable coworker, and it answers. Free works for everyday questions and includes a limited number of Projects. Pro is the paid plan that gives you more usage during busy periods plus access to advanced products such as Cowork and Claude Code.

Do this

1Open a web browser and go to claude.ai.
2Click "Sign up" and register with an email address or a Google account.
3Confirm your email if asked, then sign in.
4Look at the plan options. Start on Free. You can upgrade to Pro later from Settings if you find yourself running into limits.

Quick reference

Question	Free	Pro
Cost	Nothing	A monthly fee
Usage during busy times	Lower limits	Higher limits
Access to the strongest models	Limited	Full
Projects	Up to five	Included

You are done when

You are signed in and looking at an empty chat box with a cursor ready for your first message.

Step 2 of 45 · Claude Chat · Beginner

Learn the three parts of the interface you will use most.

Take a tour of the chat screen

The screen has a few simple parts, and knowing them removes most of the early confusion. The main area is where the conversation happens. The sidebar on the left holds your past conversations so you can return to them. A "new chat" button starts a fresh, blank conversation with no earlier context carried over.

Do this

1Find the large text box. This is where you type. Press Enter to send.
2Find the sidebar on the left. Each line there is a past conversation. Click one to reopen it.
3Find the "New chat" button (usually top left). Click it to start clean.
4Start a new chat for each separate topic. It keeps Claude focused and your history readable.

The Claude chat home screen showing the sidebar on the left, the message box, and the model picker.

You are done when

You can start a new chat, send a message, and reopen an earlier conversation from the sidebar.

Step 3 of 45 · Claude Chat · Beginner

Three habits that turn vague replies into useful ones.

Write a prompt that gets a good answer

A prompt is just the message you send Claude. The quality of the answer follows the quality of the ask. Three things do most of the work: say clearly what you want, give the background Claude cannot see, and show an example of the result you have in mind. Without context, Claude has to guess. With it, the reply lands much closer to what you needed.

Do this

1State the task plainly. "Write a thank-you email to a client" beats "help with email."
2Add the context Claude lacks. Who is the client, what are you thanking them for, what tone fits.
3Show an example if you have one. Paste an email you liked and say "match this style."
4If the first answer misses, say what was off and ask for a revision. The conversation remembers what came before.

A structured prompt in the chat box with a clear ask, context, and example, alongside Claude's drafted reply.

Step 3 of 45 · Run it

Example prompt to copy and adapt

Do this

4If the first answer misses, say what was off and ask for a revision. The conversation remembers what came before.

Example prompt to copy and adapt

Write a short thank-you email to a client named Dana.
Context: Dana referred two new customers to us this month.
Tone: warm but professional, no more than four sentences.
Here is an email whose style I like, please match it:
[paste your example here]

You are done when

You wrote a prompt with a clear ask plus context, and the reply was close enough to edit rather than restart.

Step 4 of 45 · Claude Chat · Beginner

Upload a document or photo and have Claude work from it.

Share files and images, then ask about them

You are not limited to typing. You can attach files (a PDF, a spreadsheet, a Word document) and images (a screenshot, a photo of a page) and ask Claude questions about what is inside. This is how you get a summary of a long report, a plain explanation of a contract, or the numbers pulled out of a photographed receipt.

Do this

1Click the attachment icon (a paperclip or plus) near the text box.
2Choose a file or image from your computer.
3In the same message, ask your question. For example, "Summarize the main points of this PDF in five bullets."
4Ask follow-ups in the same chat. Claude keeps the file in mind for the rest of the conversation.

The plus attach menu open in the chat box, showing options to add files and photos, take a screenshot, and use Skills and Connectors.

You are done when

You uploaded one file or image and got an answer that clearly used its actual contents.

Step 5 of 45 · Claude Chat · Beginner

When Claude builds something, it opens in a panel beside the chat.

Use Artifacts for documents you can see and edit

An Artifact is a workspace that appears next to the conversation when Claude makes something substantial: a document, a table, a checklist, or a small interactive app. Instead of the result scrolling away in the chat, it sits in its own panel where you can read it, copy it, and ask for changes that update it in place. Think of the chat as the discussion and the Artifact as the thing you are actually producing.

Do this

1Ask for something concrete, like "Make a one-page packing checklist for a winter trip."
2Watch for a panel to open on the right. That is the Artifact.
3Request edits in the chat. "Add a section for ski gear." The panel updates.
4Use the copy or download button in the panel when you are happy with it.

The Artifacts library with a New artifact button and a list of saved artifacts.

You are done when

You asked Claude to build something, a side panel opened, and one follow-up request changed what it showed.

Step 6 of 45 · Claude Chat · Beginner

A reusable space that remembers your context and preferences.

Set up a Project and custom instructions

A Project is a folder for related conversations. Projects are available on free accounts, with a limit on how many you can create. You can add reference files once and write custom instructions, which are standing notes about who you are and how you like answers. Every chat inside that Project starts already knowing them, so you stop re-explaining yourself. A freelancer might keep a "Client Work" Project with their bios, rates, and preferred tone loaded in.

Do this

1In the sidebar, find Projects and click "New Project." Give it a name.
2Add knowledge by uploading a few reference files (a style guide, past examples, key facts).
3Write custom instructions. State your role, your audience, and how you want Claude to respond.
4Start chats inside the Project. They use the files and instructions automatically.

The Projects gallery with a New project button and existing project tiles.

Step 6 of 45 · Run it

Example custom instructions to copy and adapt

Do this

4Start chats inside the Project. They use the files and instructions automatically.

Example custom instructions to copy and adapt

About me: I run a small bakery and write our weekly newsletter.
Audience: local customers, friendly and casual.
How to respond: short paragraphs, no jargon, warm tone.
Always suggest one seasonal item to feature.

You are done when

You have a Project with at least one reference file and custom instructions, and a chat inside it reflects them without being reminded.

Step 7 of 45 · Claude Chat · Beginner

Paste one ready-made instruction that makes Claude pick the right thinking method.

Install the Adaptive Reasoning Protocol custom instruction

Custom instructions do not have to be about your bio and tone. They can also tell Claude how to think. The Adaptive Reasoning Protocol below is a single instruction you paste once that teaches Claude to size up each request and switch methods automatically: it breaks problems into small, checkable pieces when you ask it to solve or debug, it explains things plainly like a patient teacher when you ask it to learn or understand, and it does both in order when a task needs both. You do not have to remember which mode to ask for; the instruction handles that for you.

Do this

1Open the custom instructions area, either in a Project (from the previous step) or in your account's settings.
2Copy the Adaptive Reasoning Protocol below exactly as written and paste it in.
3Save the instruction. Every new chat that uses it will follow the protocol without being reminded.
4Test it. Ask Claude to debug something and watch it decompose the problem, then ask it to explain a concept and watch it switch to plain teaching.

The Settings Instructions for Claude field with the Adaptive Reasoning Protocol pasted in.

Step 7 of 45 · Run it

Adaptive Reasoning Protocol custom instruction to copy and paste

Do this

4Test it. Ask Claude to debug something and watch it decompose the problem, then ask it to explain a concept and watch it switch to plain teaching.

Adaptive Reasoning Protocol custom instruction to copy and paste

Use the Adaptive Reasoning Protocol:

Assess the request type, then apply the appropriate method:

If solving, analyzing, or debugging → Atom of Thought: Decompose into atomic reasoning units. For each atom:

State the logical component
Validate independence
Verify correctness
Then synthesize atoms into final answer.

If explaining, learning, or teaching → Feynman Loop: Explain as if teaching a curious beginner. For each cycle:

Use a concrete analogy
Flag confusion points
Ask questions that reveal gaps
Then compress into a teachable snapshot.

If both are needed → Chain them: First solve via Atom of Thought, then explain the solution via Feynman Loop.

You are done when

You pasted the Adaptive Reasoning Protocol into your custom instructions, saved it, and saw Claude switch methods on its own across a solving task and an explaining task.

Step 8 of 45 · Claude Chat · Beginner

What Claude remembers between chats, and how to revisit past work.

Understand memory and history

Two different things are easy to mix up. History is the list of your past conversations in the sidebar, which you can reopen anytime. Memory is a newer ability where Claude can carry helpful details forward across chats so you repeat yourself less. History is always there. Memory you can review and control in settings, which the next step covers.

Do this

1Scroll the sidebar to see your history. Click any conversation to continue it.
2Use the search box (if shown) to find an old chat by a word you remember.
3Open Settings and look for a Memory section to see what, if anything, Claude has retained.
4Rename or delete old conversations to keep the sidebar tidy.

The Settings Capabilities Memory panel with controls to search and reference past chats and generate memory.

You are done when

You can tell the difference between your conversation history and Claude's memory, and you found where each one lives.

Step 9 of 45 · Claude Chat · Beginner

Decide what is kept, what trains models, and what you share.

Check your privacy settings

You control a few important switches. You can delete conversations from your history. You can choose whether your chats are used to improve Claude's models through a training setting. And it is on you to decide what is safe to paste in. These choices are yours to make, and knowing where they live matters more than any single default.

Do this

1Open Settings, usually under your name or initials in a corner.
2Find the privacy or data controls. Review the training toggle and set it to your preference.
3Learn how to delete a single conversation and how to clear history.
4Make a personal rule about sensitive data. Avoid pasting passwords, full account numbers, or anything you could not share with a contractor.

The Settings Privacy panel with the model training toggle, export, shared chats, and memory controls.

Quick reference

Setting	What it controls	A safe starting point
Conversation history	Whether past chats are saved	Keep on so you can return to work
Training toggle	Whether your chats help improve models	Your choice, off if you handle private material
What you paste	The information you hand Claude	Skip passwords and full account numbers

You are done when

You found your privacy settings, set the training toggle deliberately, and know how to delete a conversation.

Module 2 · Beginner

Browser use, with Claude acting on websites

Claude for Chrome

Claude for Chrome is a browser extension that lets Claude see the website you have open and take actions on it for you: clicking, typing, and moving between pages while you watch. In plain chat you copy and paste; here Claude works on the live page. Anthropic calls this browser use, and it is a beta feature on paid plans. It sits between everyday chat and the deeper Cowork workflows on purpose, so you get a gentle first taste of Claude taking actions, with you in control the whole way. Because Claude is acting inside your browser, this module gives safety real weight. Work through the steps in order.

Step 10Understand what Claude for Chrome is and when it helps Step 11Install the extension and check you can use it Step 12Sign in and connect your Claude account Step 13Grant access one site at a time Step 14Learn the safety model for browser use Step 15Try one safe, low-stakes browser task Step 16Set your privacy and data preferences, then look ahead to Cowork

Step 10 of 45 · Claude for Chrome · Beginner

Browser use lets Claude act on the page in front of you.

Understand what Claude for Chrome is and when it helps

Plain Chat is a conversation. You describe a web page or paste text in, and Claude answers in the chat. Claude for Chrome goes a step further: it can see the tab you have open and act on it, clicking buttons, filling fields, and moving between pages the way you would. You reach for it when the work is on a website and copy-paste is tedious, like pulling details off a page into a note or stepping through a form. For a plain question or a draft, regular Chat is still the simpler choice.

Do this

1Notice the difference. Chat talks about the web. The extension acts on the page you are looking at.
2Think of one website task you repeat that involves a lot of clicking or copying.
3Keep that task in mind. You will try a safe version of it later in this module.
4Reach for plain Chat when you only need an answer or a draft, and for the extension when the work lives on a live page.

Quick reference

You want to	Reach for
Ask a question or draft something	Plain Claude Chat
Summarize the exact page you have open	Claude for Chrome
Pull details off a website into a note	Claude for Chrome
Work with files on your computer	Claude Code, later in this masterclass

You are done when

You can say in one sentence when browser use helps and when plain Chat is the better fit.

Step 11 of 45 · Claude for Chrome · Beginner

Add it from the Chrome Web Store, on a paid plan.

Install the extension and check you can use it

Claude for Chrome is a Chrome extension, so you add it the same way you add any extension, from the Chrome Web Store. Two things to know before you start. It runs in Google Chrome only, not in other browsers or on phones. And it is a beta feature for paid plans (Pro, Max, Team, or Enterprise), so it is not available on the free plan. If you are on Free, you would upgrade first. None of this requires any setup beyond a normal browser extension install.

Do this

1Confirm you are using Google Chrome on a computer. The extension does not run in other browsers or on mobile.
2Confirm you are on a paid Claude plan. Browser use is in beta and not part of the free plan.
3Open the Chrome Web Store and search for the Claude extension from Anthropic.
4Add it to Chrome and approve the permissions Chrome asks for. Pinning it to the toolbar makes it easy to find.

The Chrome Web Store listing for the Claude extension by Anthropic with an Add to Chrome button.

You are done when

The Claude extension is installed in Chrome and you can see its icon in the toolbar.

Step 12 of 45 · Claude for Chrome · Beginner

Open the extension and log in with your usual account.

Sign in and connect your Claude account

The extension uses the same Claude account you have used all along. Opening it for the first time walks you through signing in, and after that it stays connected. This is also where you confirm browser use is available to you. If the extension says the feature needs a paid plan, that matches the access limits from the last step rather than anything going wrong.

Do this

1Click the Claude extension icon in the Chrome toolbar to open it.
2Sign in with the same Claude account you use on the website.
3Let it finish connecting. A panel for talking to Claude should appear.
4If it tells you browser use needs a paid plan, upgrade first, then sign in again.

You are done when

The extension is open, you are signed in, and you can see where you type a request to Claude.

Step 13 of 45 · Claude for Chrome · Beginner

Claude works on a website only after you allow that site.

Grant access one site at a time

This is the control that keeps browser use in your hands. Claude cannot act on a website until you approve that site, and approval is per site. Visiting a page does not hand it over. When you ask Claude to work on a site, it asks for access first, and you decide: allow just this one action, allow that site going forward, or decline. You can review and remove the sites you have approved later in the extension's settings, so access never silently piles up.

Do this

1Open a safe, ordinary website, like a news article or a public help page.
2Ask Claude in the extension to do something with the page, such as summarizing it.
3When it asks for access, read the request. Choose to allow this one action, or grant that single site ongoing access, or decline.
4Look in the extension settings for the list of approved sites, so you know where to review or remove access later.

The Settings panel for Claude in Chrome showing per-site permission controls.

Quick reference

When Claude asks for access, you can	What it means
Allow once	Claude acts a single time on this page, then has to ask again
Allow this site	Claude can keep working on this one site until you remove it
Decline	Claude does not touch the page

You are done when

You granted access to one site, saw that it applied to that site only, and know where to review approved sites.

Step 14 of 45 · Claude for Chrome · Beginner

The most important step. Read it before you trust the extension widely.

Learn the safety model for browser use

Browser use is powerful because Claude acts on real pages, and that is exactly why it carries weight. Two risks matter most. First, a web page can hide instructions meant to trick an AI into doing something you did not ask, such as a page or email that quietly says to find and share your private information. This is called prompt injection. Second, Claude is acting inside your logged-in sessions, so on a site where you are signed in it can reach what your browser can reach. The protections are real but not perfect, so you stay the safeguard. Claude asks before acting and pauses for high-risk actions like purchases, you grant access per site, and you review what it proposes. Keep it away from sensitive places, banking, investments, medical and legal accounts, and anything with other people's private information, until you trust it. Anthropic blocks some of those categories outright. You stay in control the whole time.

Do this

1Treat anything on a web page as something Claude might read, including hidden text. Do not assume a page is harmless just because it looks plain.
2Keep browser use away from banking, investing, medical, legal, and work accounts with sensitive data while you are learning.
3Watch for the moments Claude asks before acting, and read what it plans before you approve. Decline anything you did not intend.
4Write clear, specific requests. A vague ask gives Claude more room to do something you did not want.

Quick reference

Risk	What it is	How you stay safe
Prompt injection	Hidden instructions on a page or in an email try to redirect Claude	Approve per site, review actions, avoid sketchy pages
Your logged-in sessions	On a signed-in site Claude can reach what your browser can	Keep it off sensitive accounts until you trust it
A wrong action	Claude does something you did not intend	It asks before acting and pauses for purchases and the like; you review and decline

You are done when

You can explain prompt injection in one sentence, you know which sites to avoid for now, and you understand that you approve and review every action.

Step 15 of 45 · Claude for Chrome · Beginner

Summarize the open page or pull a few details into a note.

Try one safe, low-stakes browser task

Now you try it for real on something harmless. Good first tasks read the page and hand you back something useful without changing anything important: a short summary of a long article, or a few key details pulled off a page so you do not have to copy them by hand. Starting low-stakes lets you watch how the extension asks for access and how it works the page, with nothing at risk.

Do this

1Open a long but ordinary page, like a news article or a public documentation page.
2Ask for a summary, such as "Summarize the main points of this page in five bullets." Grant access to that one site when asked.
3Try pulling details next, like "List the dates and names mentioned on this page," and watch Claude read the page to answer.
4Read the result and confirm it matches what is actually on the page. Decline anything you did not intend.

You are done when

You completed one safe task on a real page, approved access for that site, and got a result drawn from the actual page.

Step 16 of 45 · Claude for Chrome · Beginner

Manage approved sites, and see how browser use complements Cowork.

Set your privacy and data preferences, then look ahead to Cowork

Two things close out the module. First, tidy your settings: the extension keeps the list of sites you have approved, and you can review or remove any of them so access stays only where you want it. Carry over the same habits from earlier modules about what is safe to let Claude see. Second, a look ahead. The next module, Cowork, connects Claude to specific tools like Gmail and Calendar through dedicated connections. Browser use is the complement to that: when a website has no dedicated connection, Claude can still act on it directly through the extension. They are separate tools, and the extension does not power Cowork, but together they widen what Claude can reach for you.

Do this

1Open the extension settings and review the list of sites you have approved. Remove any you no longer need.
2Apply your usual rule about sensitive information. If you would not let a contractor see it, keep the extension away from it.
3Note the split for later: Cowork uses dedicated connections for tools like Gmail, while the extension acts on websites that have no such connection.
4Keep both in mind. As you move into Cowork next, you will have browser use available for sites the connections do not cover.

You are done when

Your approved-site list reflects choices you made, and you can describe how browser use complements the Cowork connections in the next module.

Module 3 · Intermediate

Agentic desktop workflows

Claude Cowork

Cowork is not a separate product. It is the Claude desktop app set up to do real work: stronger reasoning turned on, safe connections to your tools like Gmail and Calendar, specialized instruction sets called Skills, and Projects that run start-to-finish workflows. This module takes you from a chat assistant to something that completes tasks. Work through the steps in order, since each one builds on the last.

Step 17Download the Skill Library Step 18Install Claude Desktop and sign in Step 19Turn on stronger reasoning (Atomic Thought)Step 20Configure your privacy settings Step 21Enable Connections (MCP)Step 22Turn on Skills and related features Step 23Install your Skills Step 24Create a new Project Step 25Add your workflow's connections to the Project Step 26Write the workflow instructions Step 27Run the workflow and steer it live Step 28Iterate and build new Skills

Step 17 of 45 · Claude Cowork · Intermediate

Get the pre-built instruction sets you will install later.

Download the Skill Library

A Skill is a file of instructions that gives Claude focused expertise for one kind of task, like triaging support tickets or drafting a weekly update. The files end in .md, a plain text format you do not need to understand to use. A Skill Library is a bundle of these ready-made files. Downloading the library now means you have proven starting points instead of writing everything from scratch.

Do this

1Open the Skill Library link provided with this masterclass.
2Download the bundle to a folder you can find again, like Documents or Desktop.
3If it arrives zipped, unzip it. You should see several files ending in .md.
4Leave them there for now. You will install the ones you want in a later step.

You are done when

You have a folder of .md Skill files saved somewhere you can locate.

Step 18 of 45 · Claude Cowork · Intermediate

Move from the website to the desktop app, which unlocks more.

Install Claude Desktop and sign in

The website is fine for chatting, but the desktop app is where the advanced features live. Only the desktop app can connect to your other tools and run the Skills you just downloaded. Installing it is the gateway to everything else in this module.

Do this

1Go to claude.ai and find the "Download" option for the desktop app, or visit the downloads page directly.
2Choose the version for your computer (Mac or Windows) and run the installer.
3Open the app and sign in with the same account from Module 1.
4Confirm you are on Pro. The workflow features below need it.

The Cowork intro screen reading Hand off complex tasks with an Open in desktop app option.

You are done when

Claude Desktop is open on your computer and you are signed in.

Step 19 of 45 · Claude Cowork · Intermediate

Tell Claude to break problems into small, checkable pieces.

Turn on stronger reasoning (Atomic Thought)

For multi-step work, you want Claude to slow down and reason carefully rather than rush to an answer. This masterclass calls that "Atomic Thought." The idea is simple: Claude breaks a problem into independent pieces it can check one at a time, which catches mistakes a single fast pass would miss. You turn it on through a setting and by asking for it in your instructions.

Do this

1Open Settings in the desktop app and look for "Extended thinking" or a similar reasoning option. Turn it on.
2In prompts for complex tasks, add a line like "Think through this step by step and check each part before moving on."
3Save this phrasing in your Project instructions later so you do not retype it.
4Use it for planning and analysis. For a quick fact, you do not need it.

The Instructions for Claude field showing an Atom of Thought reasoning instruction.

You are done when

Extended thinking is enabled, and you have a sentence you can drop into prompts to ask for careful, step-by-step reasoning.

Step 20 of 45 · Claude Cowork · Intermediate

Set history, training, memory, and past-chat search on purpose.

Configure your privacy settings

Before you connect real tools and accounts, decide how your data is handled. The desktop app gives you control over saved history, whether chats train models, what Claude remembers, and whether it can search your past conversations. Setting these deliberately now means you connect Gmail or Drive in the next steps with clear expectations.

Do this

1Open Settings and find the privacy or data controls.
2Go down the recommended table below and set each one.
3Decide your training toggle based on how sensitive your work is.
4Note that memory and past-chat search are conveniences. Turn them on if the time saved is worth it to you.

Quick reference

Setting	What it does	Recommended start
Conversation history	Saves chats so you can reopen them	On
Training toggle	Lets your chats improve models	Off if you handle client or private data
Memory	Carries useful details across chats	On for convenience, off for strict privacy
Search past chats	Lets Claude pull from earlier conversations	On if you reference past work often

You are done when

Each of the four settings reflects a choice you made, not just the default.

Step 21 of 45 · Claude Cowork · Intermediate

The step that turns a chatbot into an agent.

Enable Connections (MCP)

This is the heart of Cowork. A Connection uses something called MCP (Model Context Protocol), which is a safe, standard way for Claude to reach an outside tool such as Gmail, Calendar, Google Drive, Linear, or Stripe. You approve each connection, so Claude only touches what you allow. Once connected, Claude can read and act in those tools instead of only talking about them. That shift, from describing work to doing it, is what makes it an agent.

Do this

1In the desktop app, open Settings and find Connections (sometimes labeled MCP or Connectors).
2Browse the available connections. Pick one you actually use, such as Google Calendar.
3Click to connect and follow the sign-in prompt. You are granting access through the tool's own secure login.
4Confirm it worked by asking Claude something only the connection can answer, like "What is on my calendar tomorrow?"

The Connectors panel listing tools like Airtable, Gmail, and Slack with per-tool permission controls.

You are done when

At least one connection is active and Claude correctly answered a question using live data from it.

Step 22 of 45 · Claude Cowork · Intermediate

Switch on Artifacts, file creation, code execution, and thinking.

Turn on Skills and related features

A few capabilities power real workflows, and some live behind toggles. Artifacts give you editable documents beside the chat. File creation lets Claude produce actual files you can save. Code execution lets it run small tasks to compute or transform data. Extended thinking is the careful reasoning from earlier. Turning these on now means your Skills have everything they need.

Do this

1Open Settings and find the features or capabilities area.
2Enable Skills, Artifacts, file creation, and code execution.
3Confirm extended thinking is still on from the earlier step where you turned on stronger reasoning.
4If a feature asks for confirmation the first time it runs, that is the safety model working as intended.

The Capabilities settings showing toggles for Visuals and for code execution and file creation.

You are done when

Skills, Artifacts, file creation, and code execution are all switched on.

Step 23 of 45 · Claude Cowork · Intermediate

Add Skills globally or, better, to a single Project.

Install your Skills

Now you load the .md Skill files you downloaded earlier. You have two choices. Global means a Skill applies everywhere, across all your chats. Per-Project means it applies only inside one Project. Per-Project is recommended because it keeps each workflow self-contained and predictable, so a support Skill does not bleed into your newsletter writing.

Do this

1Decide which one or two Skills match the workflow you want to build first.
2Find where the app lets you add a Skill (in Settings for global, or inside a Project for per-Project).
3Add the .md file there. Prefer adding it to a Project.
4Keep it focused. Install only the Skills you will use now. You can add more later.

The Skills list with a skill enabled and its trigger shown.

You are done when

At least one Skill is installed, and you know whether it is global or tied to a specific Project.

Step 24 of 45 · Claude Cowork · Intermediate

Build the container that holds one workflow's instructions and context.

Create a new Project

A Project here works like the one in Module 1, but now it anchors a whole workflow. It bundles the instructions, the reference knowledge, and the connections for one job in one place. Think of it as a dedicated workspace: open it and Claude already knows the task, the background, and which tools it may use.

Do this

1In the desktop app, create a new Project and name it for the job, like "Weekly Customer Update."
2Write a short description of what this Project is for.
3Add any reference files the workflow needs, such as a template or past examples.
4Confirm the Skill you installed a moment ago is attached to this Project if you chose per-Project.

You are done when

A named Project exists with a clear purpose, its reference files, and its Skill in place.

Step 25 of 45 · Claude Cowork · Intermediate

Map each thing the workflow needs to do to a connection.

Add your workflow's connections to the Project

A workflow usually touches more than one tool. A weekly update might read your calendar, pull notes from Drive, and send an email through Gmail. This step is where you match each need to a connection so Claude can reach everything in one run. Doing it deliberately keeps the workflow from stalling halfway because a tool was not connected.

Do this

1List what your workflow must do. For example: read this week's meetings, gather project notes, draft and send a summary.
2For each item, name the connection it needs (Calendar, Drive, Gmail).
3Make sure each of those connections is among the ones you enabled earlier and available to this Project.
4Write the mapping into your notes so the next step's instructions can reference it.

Quick reference

Workflow need	Connection it uses
Read this week's meetings	Google Calendar
Gather project notes	Google Drive
Send the summary	Gmail

You are done when

Every action your workflow needs is matched to a connection that is live and reachable from the Project.

Step 26 of 45 · Claude Cowork · Intermediate

A clear framework Claude can follow the same way every time.

Write the workflow instructions

Instructions are where you turn a vague wish into a repeatable routine. Good ones cover six things: the purpose, the trigger that starts it, the steps in order, the output format you want, quality checks, and what to do when something goes wrong. Spelling these out means Claude runs the workflow consistently instead of improvising. The skeleton below is a starting template you can copy.

Do this

1Open your Project's instructions area.
2Fill in each section of the framework below for your own workflow.
3Be specific about the output. State the format, the length, and the tone.
4Add quality checks and an error plan so Claude knows how to verify its own work and when to pause and ask you.

The Instructions for Claude field where workflow instructions are written.

Step 26 of 45 · Run it

Workflow instruction skeleton to copy and fill in

Do this

4Add quality checks and an error plan so Claude knows how to verify its own work and when to pause and ask you.

Workflow instruction skeleton to copy and fill in

Purpose: What this workflow accomplishes and for whom.

Trigger: When I run this (for example, "when I say 'weekly update'").

Steps:
1. Read this week's meetings from Calendar.
2. Pull related notes from the Drive folder named [folder].
3. Draft a summary in the format below.
4. Show me the draft before sending anything.

Output format:
- A subject line.
- Three to five bullet points of progress.
- One short paragraph on next week.

Quality checks:
- Every bullet ties to a real meeting or note.
- No names misspelled.

Error handling:
- If a tool is unreachable or a folder is empty, stop and tell me what is missing.

You are done when

Your Project holds instructions with all six parts, written for your specific workflow.

Step 27 of 45 · Claude Cowork · Intermediate

Trigger it, watch it work, correct it in the moment, save the good results.

Run the workflow and steer it live

Instructions written, now you run it. You start the workflow, watch Claude move through your steps, and step in if something drifts. This is normal and expected on early runs. You are coaching it toward the result you want, and good outputs are worth saving as examples to make future runs even better.

Do this

1In the Project, give the trigger phrase from your instructions.
2Watch each step. When Claude pauses to show a draft, read it.
3Correct in plain language. "The second bullet is about the wrong meeting, fix it."
4When the result is right, save it. Keep a copy as a reference example in the Project.

The Cowork intro screen reading Hand off complex tasks.

You are done when

The workflow ran end to end, you corrected at least one thing in real time, and you saved a good output.

Step 28 of 45 · Claude Cowork · Intermediate

Score each run, improve the weak parts, and turn wins into Skills.

Iterate and build new Skills

The first run is rarely the best one. A simple scorecard tells you what to fix instead of guessing. You rate the run against a few metrics, compare target to actual, and note what to change. When a workflow consistently works well, you can capture it as a new Skill (a fresh .md file) and reuse it elsewhere. This is the loop that compounds.

Do this

1After a run, fill in the scorecard below.
2Pick the lowest-scoring row and adjust your instructions to address it.
3Run again and compare. Repeat until the scores hold steady.
4Once it is reliable, save the refined instructions as a new Skill file for reuse.

The Skills library showing a skill detail view.

Quick reference

Metric	Target	Notes
Accuracy of the summary	All points correct	What to fix
Format match	Matches my template	What to fix
Time saved	Faster than by hand	Worth it or not
Edits needed	Minor only	Where it slipped

You are done when

You scored a run, made one improvement that raised a score, and know how to save a working workflow as a Skill.

Module 4 · Advanced

Claude Code, plus reusable prompts and skills

Claude Code

Claude Code is the version of Claude that runs as a desktop app and can create files, edit them, and run tasks for you directly on your machine. It sounds technical, and it can be, but plenty of everyday uses need no programming at all. This module eases you in: what it is, when to use it, a few real tasks you can try safely, and a free starter pack that powers it up with reusable prompts and skills.

Step 29Understand what Claude Code is and when to use it Step 30Know that basic uses do not require programming Step 31Get to know the Claude Code app Step 32Install the Claude Code app Step 33Open it for the first time and sign in Step 34Try a few real, non-technical tasks Step 35Learn the safety model Step 36Meet the meta-prompting starter pack Step 37Install the starter pack Step 38Run your first prompt Step 39Level up with workflows and loops Step 40Know when to graduate and where to get help

Step 29 of 45 · Claude Code · Advanced

A short comparison of Chat, Claude for Chrome, Cowork, and Code.

Understand what Claude Code is and when to use it

All four are Claude, with different reach. Chat is conversation in a browser or app. Claude for Chrome is a browser extension that acts on the website in front of you. Cowork is the desktop app connected to your online tools, running workflows. Code goes one level deeper: it works directly with the files and folders on your computer, creating and changing them for you. You reach for Code when the job is about files on your machine rather than a conversation, a website, or an online tool.

Do this

1Read the comparison below to place each one.
2Notice the dividing line. If the work lives in your computer's folders, Code fits.
3For now, just recognize the cases. You do not have to use Code today.
4Keep Chat, Claude for Chrome, and Cowork for everything they already handle well.

Quick reference

Version	Where it runs	Best for
Claude Chat	Browser or app	Questions, drafts, learning
Claude for Chrome	A Chrome extension in your browser	Acting on the website in front of you for you
Claude Cowork	Desktop app with connections	Workflows across Gmail, Calendar, Drive
Claude Code	Its own desktop app	Creating and editing files and folders on your own computer

You are done when

You can say in one sentence when Code is the right choice instead of Chat, Claude for Chrome, or Cowork.

Step 30 of 45 · Claude Code · Advanced

Reassurance before you open anything unfamiliar.

Know that basic uses do not require programming

The name scares people off, and it should not. You talk to Claude Code in plain English, the same as every other version. You can ask it to tidy a folder, rename files, or draft documents without writing a single line of code. Programmers get more out of it, but the front door is open to everyone.

Do this

1Drop the assumption that you need to code. You do not, for the tasks here.
2Plan to type requests in normal sentences, like "rename these photos by date."
3Expect Claude Code to explain what it is doing as it goes.
4Keep going. The next steps walk you in slowly.

You are done when

You feel ready to try a tool that takes plain-language requests, even though it looks technical.

Step 31 of 45 · Claude Code · Advanced

One plain paragraph on the app window before you do anything.

Get to know the Claude Code app

The Claude Code app is a normal window, the same as any other program on your computer. You open a folder so Claude knows which files you are working with, you type your request in plain language, and Claude works through it while you watch. When it wants to create or change a file, it pauses and asks first. Knowing where those three things live removes most of the early confusion before you type anything real.

Do this

1Open the Claude Code app and look around the window. There is nothing to break by looking.
2Find where you open a folder. This tells Claude which files and folders it is allowed to work with.
3Find the box where you type your request, the same way you type a message in any other version of Claude.
4Notice where it asks for your approval. That is where you will accept or decline anything it proposes.

The Claude Code app window showing New session, Routines, Recents, and Pull requests.

You are done when

You opened the app and can point to where you open a folder, where you type a request, and where approvals appear.

Step 32 of 45 · Claude Code · Advanced

Download and install the Claude Code desktop app.

Install the Claude Code app

Claude Code has its own desktop app, separate from the Claude Desktop app you installed in Module 3. That one is for working across your online tools. This one is for working with the files and folders on your computer, so it is a different download for a different job. Installing it is a normal app install, the same as any other program.

Do this

1Go to the official Claude Code download page from Anthropic.
2Choose the version for your computer, Mac or Windows, and download it.
3Run the installer and follow the prompts, the same way you would install any other app.
4Wait for it to finish. The installer tells you when it is done.

The Claude Code download page with a Download for macOS button.

You are done when

The Claude Code app finished installing without errors and is ready to open.

Step 33 of 45 · Claude Code · Advanced

Launch the app and connect your account.

Open it for the first time and sign in

With it installed, you open the Claude Code app and sign in with the same Claude account you have used all along. The first time, it usually opens your browser to log in, then hands you back to the app. After this, opening it works like any other program.

Do this

1Open the Claude Code app.
2When it prompts you to sign in, follow the steps. Log in with your usual account.
3Return to the app once you are signed in.
4You should see a box where you can type a request.

You are done when

The Claude Code app is open, you are signed in, and it is waiting for your first request.

Step 34 of 45 · Claude Code · Advanced

Three things a non-programmer can actually do today.

Try a few real, non-technical tasks

This is where it gets concrete. You can ask Claude Code to build a simple one-page website, tidy and rename a messy folder of files, or draft and revise documents that live in a folder. You describe the goal in plain words and let it do the work, reviewing as you go. Start with a practice folder so nothing important is at stake.

Do this

1Make a new, empty folder for practicing, then open that folder in the Claude Code app so it works only on those files.
2Try a website: "Build a simple one-page website for my dog-walking service with my name, services, and a contact email."
3Try tidying files: "This folder has photos with messy names. Rename them by the date each was taken." Review the proposed names before approving.
4Try documents: "Draft three short blog post outlines in this folder, then revise the second one to be more casual."

Quick reference

Task	What you type	What you get
One-page website	"Build a one-page site for my service"	A web page file you can open in a browser
Tidy a folder	"Rename these files by date"	Cleanly named files after you approve
Draft documents	"Draft and revise outlines in this folder"	Document files you can edit further

You are done when

You completed at least one of the three tasks in a practice folder and saw the result.

Step 35 of 45 · Claude Code · Advanced

It asks before acting, and you review changes before they stick.

Learn the safety model

This is the part that should put you at ease. Claude Code does not silently change your files. It shows you what it intends to do and asks permission first. You see the proposed changes and approve or decline them. Nothing becomes permanent without your say-so, which means you can experiment without fear of breaking things.

Do this

1When Claude Code proposes an action, read the summary it shows you.
2Approve only what you understand and want. Decline or ask questions otherwise.
3Review file changes before accepting them, the same way you would proofread an edit.
4Keep important work backed up anyway, as a sensible habit for any tool.

Claude Code settings showing session classification, model switching, and an Accept edits control.

You are done when

You saw Claude Code ask for permission before acting and understand that you approve changes before they take effect.

Step 36 of 45 · Claude Code · Advanced

A small free download that adds reusable prompts and skills.

Meet the meta-prompting starter pack

Once Claude Code feels comfortable, you can teach it a few new tricks. The meta-prompting starter pack is a small free download (about 166 KB) that adds two slash commands and two skills to the Claude Code you already have. A slash command is a shortcut you type with a slash in front, like /create-prompt. These help Claude Code write better prompts, save and reuse them, and break a big job into steps. These commands and skills are different from the Cowork Skills in Module 3. The Cowork Skills live inside the Cowork desktop app. These live in a hidden folder on your computer called .claude, and Claude Code loads them every time it starts.

Do this

1Read the table below so you know what each of the four pieces does.
2Notice that two are slash commands you type, and two are skills Claude Code uses on its own when a job calls for them.
3You do not need all four at once. /create-prompt is the one most people start with.
4When you are ready to add them, the next step walks through the install.

The technical-leaders.com Claude meta-prompting landing page.

Quick reference

Piece	Type	What it does for you
/create-prompt	Slash command	A coach that helps you write a strong prompt. You answer a few quick questions and Claude writes the well-worded prompt, then saves it so you can reuse it.
/run-prompt	Slash command	Your saved-prompts shelf. Pick a prompt you saved earlier and run it again, with no copy-pasting or rewriting from memory.
create-meta-prompts	Skill	A recipe for prompts that build other prompts. When a job needs research, then a plan, then the work, in that order, it stitches those three steps together. Good for repeatable, multi-step jobs.
create-plans	Skill	Breaks a big, fuzzy goal like "migrate our docs to the new site" into small, testable steps Claude can finish one at a time.

You are done when

You can name the four pieces and explain how these Claude Code commands and skills differ from the Cowork Skills in Module 3.

Step 37 of 45 · Claude Code · Advanced

Download it, then drag the files into your hidden .claude folder.

Install the starter pack

Installing means putting the downloaded files where Claude Code looks for them. It is all drag and drop, with nothing to run. You unzip the download, open a hidden folder called .claude on your computer, and drag the command files and skill folders into place. The one thing to do first: open Claude Code at least once, because that first launch is what creates the .claude folder you are about to drop files into.

Do this

1Make sure you have opened Claude Code at least once already, so the hidden .claude folder exists.
2
Download the starter pack and unzip it.
- Use the download button below to get claude-meta-prompting.zip.
- On Mac, double-click the file to unzip it. On Windows, right-click it and choose Extract All.
- You now have a claude-meta-prompting folder with a commands folder and a skills folder inside.
3
Open the hidden .claude folder using the row for your computer in the table below.
- On Mac, open Finder, press Command + Shift + G, type ~/.claude, and press Return.
- On Windows, open File Explorer and type %USERPROFILE%\.claude in the address bar, then press Enter.
4Drag the two .md files from the unzipped commands folder into the commands folder inside .claude. If there is no commands folder yet, drag the whole commands folder over instead.
5Drag the create-meta-prompts and create-plans folders from the unzipped skills folder into the skills folder inside .claude. If there is no skills folder yet, drag the whole skills folder over instead.
6Open Claude Code and type /create-prompt. If it appears in the menu, the install worked. The .md files now live in commands, the two skill folders live in skills, and Claude Code reads both folders automatically every time it starts.

Remember: About 166 KB. A single zip file. Nothing runs on its own; you just drop the folders into place.

Quick reference

Your computer	How to open the hidden .claude folder
Mac	In Finder, press Command + Shift + G, type ~/.claude, and press Return.
Windows	In the File Explorer address bar, type %USERPROFILE%\.claude and press Enter.

You are done when

The command files and skill folders are inside your .claude folder, and typing /create-prompt in Claude Code shows the new command in the menu.

Step 38 of 45 · Claude Code · Advanced

Try /create-prompt, then reuse it later with /run-prompt.

Run your first prompt

Now you use what you installed. You type /create-prompt followed by a plain description of a task, starting with "how might we." Claude asks a few questions about how the task should work, then writes a well-worded prompt for you and saves it. Later, when you want that same prompt again, /run-prompt pulls it off your saved-prompts shelf so you do not rewrite it from memory.

Do this

1Open Claude Code and type /create-prompt followed by a task, like the example below.
2Answer the few questions it asks about how the task should work.
3Read the prompt it writes for you. It is saved automatically for next time.
4Later, type /run-prompt to pick that saved prompt and run it again.

Step 38 of 45 · Run it

Example to type in Claude Code

Do this

4Later, type /run-prompt to pick that saved prompt and run it again.

Example to type in Claude Code

/create-prompt how might we write a weekly update from my meeting notes

You are done when

You ran /create-prompt on a real task, got a saved prompt back, and know that /run-prompt brings it back later.

Step 39 of 45 · Claude Code · Advanced

Hand Claude Code a bigger, open-ended job and let it run on its own.

Level up with workflows and loops

Once single prompts feel natural, the next step is a workflow: one plain-language request that sets Claude Code off on a multi-step job, often spinning up several helper agents that work in parallel. A few words unlock the real power. "Use a workflow" tells it to plan the steps and split the work. A loop keeps it going on its own: /loop reruns a task on a schedule, and /goal tells it not to stop until a goal is met. "In a worktree" lets agents try ideas in separate, throwaway copies of your files so nothing collides. And it can interview you with the AskUserQuestion tool when it needs your judgment. You still approve real changes, so a workflow is ambitious but not reckless. The examples below show the range, from babysitting code to ranking resumes.

Do this

1Pick one example below that sounds like a job you actually have.
2Open the right folder in Claude Code, then paste the example and adjust the details to your situation.
3Watch it plan the steps and, where useful, fan out into parallel agents. Answer its questions and approve changes as they come.
4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 1

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 1

/goal Build report.md comparing the five options in brief.md.

Finish line:
- Every section in brief.md is answered.
- The comparison table has no unexplained blank cells.
- Every factual claim includes a source link and source date.
- Assumptions and unresolved questions are listed separately.

Prove it:
- Print all section headings.
- Print the completed comparison table.
- Print every unresolved question and blocked source.

Show me:
- Recommend one option in three sentences.
- Name the strongest reason not to choose it.

If blocked:
- Mark the cell "Not verified" and explain why. Do not guess.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 2

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 2

/loop babysit all my GitHub PRs. Auto-fix build issues, and when comments come in, use /create-prompt in a worktree agent to fix them.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 3

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 3

This test fails maybe 1 in 50 runs. Set up a workflow to reproduce it, form theories, and adversarially test them in worktrees. /goal don't stop until one theory works.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 4

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 4

Use a workflow to dig through #incidents in Slack for the past six months and find recurring root causes where nobody has filed a ticket.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 5

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 5

Take my business plan and run a workflow where different agents tear it apart from an investor's, a customer's, and a competitor's perspective.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 6

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 6

Here's a folder of 80 resumes. Use a workflow to rank them for the backend role and double-check the top ten. Interview me using the AskUserQuestion tool for a rubric.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 7

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 7

I need a name for this CLI tool. Use a workflow to brainstorm a bunch of options and run a tournament to pick the top 3.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 8

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 8

Use a workflow to rename our User model to Account everywhere.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 9

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 9

Go through my blog post draft and, using a workflow, verify every technical claim against the codebase. I don't want to ship anything wrong.

Step 39 of 45 · Run it

Workflow and loop examples to adapt and paste into Claude Code 10

Do this

4For a job that should keep running, start the request with /loop to repeat it on a schedule, or /goal to have it keep working until the goal is met.

Workflow and loop examples to adapt and paste into Claude Code 10

Using a workflow, go through my last 50 sessions and mine them for corrections I keep making, then turn the recurring ones into CLAUDE.md rules.

You are done when

You ran one workflow from a single plain-language request, saw it break the job into steps, and know that /loop and /goal keep a job running on their own.

Step 40 of 45 · Claude Code · Advanced

Signs you are ready for Code, and where to turn when stuck.

Know when to graduate and where to get help

You do not need Claude Code on day one. It earns its place when your work moves onto your own computer in volume: lots of files to organize, documents to generate and revise in batches, or a simple site to stand up. When you hit something confusing, the fastest help is often Claude itself, plus Anthropic's official docs and community.

Do this

1Watch for the signal: repetitive file work on your machine that Chat and Cowork cannot touch.
2Start small with the practice tasks before trusting it on real folders.
3When stuck, ask Claude Code to explain what it is doing or to suggest the next step.
4Keep Anthropic's official documentation and help channels bookmarked for deeper questions.

The official Claude documentation home at docs.claude.com.

You are done when

You can name one situation where you would choose Claude Code, and you know two places to get help.

Module 5 · Advanced

Turn any workflow into a reusable, shareable skill

Skill Forge

Skill Forge is a meta-skill: it teaches Claude how to build other skills. A skill is a modular package, a SKILL.md file plus optional scripts, references, and assets, that turns Claude from a generalist into a specialist for one specific task. This final module has you download Skill Forge, understand how skills work, and run its seven-step forge to create one from scratch, improve an existing one, or package domain expertise into a portable .skill file you can hand to anyone. It builds directly on the Skills you installed in Module 3 and the Claude Code setup from Module 4.

Step 41See what a skill gives Claude Step 42Understand progressive disclosure, the core idea Step 43Download Skill Forge Step 44Run the seven-step forge Step 45Score against the rubric, then package and share

Step 41 of 45 · Skill Forge · Advanced

The four things a skill adds on top of Claude's general ability.

See what a skill gives Claude

You met ready-made Skills in Module 3 and installed Claude Code commands and skills in Module 4. A skill is a modular package: a SKILL.md file of instructions plus optional scripts, references, and assets. It turns Claude from a generalist into a specialist for one specific task. Before you build your own, it helps to see exactly what a skill can add.

Do this

1Read the four things a skill gives Claude in the table below.
2For each row, think of one task in your own work that fits it.
3Notice the last row: a skill can carry its own scripts and reference docs, not just written instructions.
4Pick the one task you would most like to turn into a skill. You will forge it later in this module.

Quick reference

What a skill gives Claude	What that means
Specialized workflows	Multi-step procedures for a specific domain
Tool integrations	Instructions for file formats, APIs, and browser automation
Domain expertise	Your business logic, schemas, and proprietary know-how
Bundled resources	Scripts and reference docs that load only when needed

You are done when

You can name the four things a skill adds and have one task of your own in mind to turn into a skill.

Step 42 of 45 · Skill Forge · Advanced

How a skill budgets Claude's attention in layers so it stays fast.

Understand progressive disclosure, the core idea

Skills share Claude's context window with everything else, so Skill Forge budgets attention in layers. The skill's name and description are always visible, and that is what makes Claude decide to use it. The main instructions load only when the skill is triggered. Heavier resources, such as scripts, reference docs, and templates, load on demand, and scripts can even run without being read into context at all. The guiding rule behind all of it: Claude is already smart, so only include what it doesn't already know.

Do this

1Read the three layers in the table below, from always-on to loaded on demand.
2Note that the description is the trigger: it is the one part Claude always sees, so it decides whether the skill fires.
3Keep the guiding rule in mind: leave out anything Claude already knows, and include only what it does not.
4This is why a well-built skill stays lightweight even when it carries a lot of material.

Quick reference

Layer	When it loads	What lives here
Name and description	Always visible	The trigger Claude reads to decide whether to use the skill
Main instructions	When the skill is triggered	The core step-by-step procedure
Resources: scripts, references, templates	On demand	Heavier material; scripts can run without being read into context at all

You are done when

You can explain progressive disclosure in one sentence and know why the description is the most important part of a skill.

Step 43 of 45 · Skill Forge · Advanced

Get the .skill file and install it like any other skill.

Download Skill Forge

Skill Forge is itself a skill, the meta-skill that builds other skills. Download the .skill file below, then install it the same way you added Skills in Module 3, or add it to Claude Code from Module 4. Once it is installed, you trigger it simply by asking Claude to build, improve, or package a skill, and it takes over from there.

Do this

1Use the download button below to save skill-forge.skill somewhere you can find it.
2Install it the same way you installed Skills in Module 3 (globally or, better, inside a Project), or add it to Claude Code from Module 4.
3Confirm it is enabled the same way you checked your other Skills.
4Trigger it by asking Claude to build, improve, or package a skill, like the request in the next step.

Remember: A single .skill file. Install it like the Skills from Module 3, or add it to Claude Code, then ask Claude to build a skill to trigger it.

You are done when

Skill Forge is downloaded and installed as a skill, ready to trigger when you ask Claude to build one.

Step 44 of 45 · Skill Forge · Advanced

The guided process Skill Forge walks you through, from Discover to Iterate.

Run the seven-step forge

Once Skill Forge is installed, you kick it off with a plain "how might we" request describing a workflow you repeat. From there it runs a seven-step forge: it interviews you about real usage, decides where each piece of knowledge belongs, scaffolds the files, writes and tests them, validates and scores the result, packages it, and helps you improve it after real use. You stay in the loop the whole way, answering its questions and reviewing its work.

Do this

1Trigger Skill Forge with a "how might we" request like the example below, naming a workflow you repeat every week.
2Answer its Discover questions: the tasks it should handle, example requests, what a great result looks like, and the edge cases.
3Let it work through the middle steps, architecting, scaffolding, and building, and review each part as it goes.
4Follow the seven steps in the table so you know what is coming at each stage.

Quick reference

Step	What happens
1. Discover	Claude asks targeted questions about real usage: the tasks, example requests, what a great result looks like, and the edge cases.
2. Architect	Each example is mapped to its home, procedures into the instructions, repeated code into scripts, decision docs into references, templates into assets, and a structure pattern is chosen.
3. Initialize	A scaffold script generates the skill's skeleton and folders.
4. Build	The resources are written and tested first, then the SKILL.md, with special care on the description since that is the trigger Claude reads before loading anything.
5. Validate and score	A validator checks the structure, then the skill is scored against a rubric. Target is 4+/5 on each before shipping.
6. Package	A packaging script zips the skill into a distributable .skill file.
7. Iterate	After real use, you spot where Claude struggles, update the instructions or resources, re-validate, and re-package.

Step 44 of 45 · Run it

Example request to trigger Skill Forge, adapt it to a workflow you repeat

Do this

4Follow the seven steps in the table so you know what is coming at each stage.

Example request to trigger Skill Forge, adapt it to a workflow you repeat

How might we turn our weekly investor-update process into a reusable Claude skill: capture the steps, the data sources, and what a great update looks like, then validate it against the rubric and package it as a .skill file I can hand to the whole team?

You are done when

You triggered Skill Forge with a "how might we" request, answered its Discover questions, and can name the seven steps of the forge.

Step 45 of 45 · Skill Forge · Advanced

The quality bar every skill clears before you hand it to anyone.

Score against the rubric, then package and share

Steps 5 and 6 of the forge are where a skill earns the right to ship. A validator first checks the structure, then Skill Forge scores the skill against a five-part rubric, aiming for 4 or higher on each before you package it. Packaging zips everything into a single portable .skill file, the same format Skill Forge itself came in, so a workflow that lived in one person's head becomes something the whole team, or anyone you send the file to, can use on demand. That portability is the whole point: most teams rediscover the same prompts and processes over and over, and this captures that expertise once, in a form Claude reliably applies.

Do this

1When Skill Forge scores your skill, read each rubric row below and push any score under 4 higher before shipping.
2Let it package the skill into a .skill file once the scores hold.
3Share that .skill file with a teammate and have them install it the way you installed Skill Forge.
4After real use, run the Iterate step: fix where Claude struggles, re-validate, and re-package.

Quick reference

Rubric dimension	What it checks
Trigger clarity	The description reliably makes Claude decide to use the skill at the right moment
Context efficiency	Only what Claude does not already know is included, kept in the right layer
Actionability	The instructions are concrete enough for Claude to follow consistently
Completeness	The real tasks and edge cases from Discover are all covered
Modularity	Scripts, references, and templates are split out so they load only when needed

You are done when

Your skill scores 4+ on every rubric dimension, is packaged as a .skill file, and you have shared it with at least one person or know how to.

Summary

Keep these twelve ideas straight.

Prompt

The message you send Claude. A clear ask, useful context, and an example produce the best result.

Artifact

A panel beside the chat where Claude builds a document, table, or small app you can edit in place.

Browser use

Claude acting in a browser tab through Claude for Chrome, with access approved site by site.

Project

A reusable container for instructions, reference files, and connections for one kind of work.

Skill

A plain-text instruction set that gives Claude focused expertise for a specific task.

Connection (MCP)

An approved link that lets Claude reach an outside tool such as Gmail or Drive.

Workspace folder

The folder opened in Claude Code, which defines the files Claude may work with.

Safety model

Claude shows what it plans to do and waits for approval before making important changes.

Slash command

A saved instruction run with a shortcut such as /create-prompt.

Starter pack

Reusable commands and skills loaded from the .claude folder when Claude Code starts.

Skill Forge

A meta-skill that builds, validates, scores, and packages other skills.

Progressive disclosure

Loading only the instructions and resources Claude needs at each stage of a task.

Keep building

Turn one lesson into a habit this week.

Install Claude for Chrome and try one safe browser task on a single approved site.

Build one real Cowork workflow for work you repeat weekly, then refine it until edits are minor.

Create a practice folder and complete the three Claude Code tasks: a one-page site, folder cleanup, and document draft.

Install the meta-prompting starter pack and run /create-prompt on a real task.

Use Skill Forge on one repeated workflow and share the packaged .skill file with a teammate.

Review privacy settings across Claude Chat, Chrome, Cowork, and Code so each reflects a deliberate choice.

Bookmark the official documentation and save your strongest workflow as a reusable Skill.

When something gets stuck

Start with the simplest fix.

Claude's answers are vague or off-target.

Add context and an example, explain what was wrong, and ask for a revision.

A Cowork connection will not return live data.

Confirm the connection is signed in and enabled for that Project, then reconnect if needed.

Claude for Chrome will not act on a website.

Open the extension and approve that site. Browser access is granted per site.

A workflow stops partway through.

Check every required connection and referenced folder. The workflow should report what is missing.

Claude Code will not open or sign in.

Quit and reopen it, check the internet connection, then reinstall from the official download page if needed.

Claude Code changed something you did not want.

Decline the next action, ask it to undo or revise, and review each proposed change before approval.

Keep nearby

Documentation, commands, and reusable skills.

Claude documentationOfficial product guides and reference material.

MCP setup guideOfficial Model Context Protocol documentation.

Meta-prompting starter packInstall the Claude Code commands and skills used in Module 4.

Skill ForgeDownload the meta-skill from Module 5 and forge a reusable skill.

Masterclass complete

You now have a practical path across the Claude ecosystem.

Use the Gear Audit to choose the next capability to strengthen, or continue into the deeper Code and automation guides.

Assess

Longhand AI Roadmap

Rate all 26 capabilities and jump directly to the lesson that closes each gap.

Open training

Build

The Claude Code Playbook

Turn Claude Code into a safe, repeatable operating system for real work.

Open training

Scale

10x Productivity

Move from individual agents to recurring loops and systems.

Open training