New to Claude? 5 Things to Do Next: Part 2

Part 1 taught Claude to know you. Part 2 teaches Claude to actually do work. These five turn it from something you talk to into something that runs.

Part 1 was setup. This is where the real leverage kicks in.

The five things below are the ones people quit on… because every article explains what they are and never tells you where to click. So this is the click path, the plan you need, the first real thing to build with each one, and the mistake that makes each one flop.

Before you touch any of this, flip one switch:

Go to Settings → Capabilities → turn on “Code execution and file creation.” One toggle. It unlocks two of the five things below (Skills and Artifacts). If you go hunting for either one later and the option is greyed out or missing, this is why. Thirty seconds. Do it now.

Thing 1. Build one reusable Skill.

A Skill is a set of instructions Claude loads automatically when the job matches. Write it once, stop re-explaining yourself forever after.

The plan you need: Pro or up.

The route that takes zero technical skill. You build the Skill by talking to Claude.

1. Open a normal chat and say the thing. “I want to create a skill for quarterly client reviews.” Swap in whatever repeats in your week. New client onboarding. Weekly recaps. Sales page copy. Meeting recaps.
2. Upload examples. Templates. Past ones you loved. Brand guidelines. This is the step most people skip, and it’s the one that makes your Skill sound like you instead of like a robot.
3. Answer Claude’s questions. It’ll ask a bunch of clarifiers. The Skill is only as good as your answers, so treat it like a real interview.
4. Claude writes the file and packages it. Save it.
5. Go to Settings → Capabilities → Skills and toggle it on.
6. Test it. Ask for the exact thing it was built for.

Where Skills live in the app: Customize → Skills in the left sidebar. The direct link is claude.ai/settings/capabilities. To browse pre-built ones: Customize → Skills → the + button → Browse skills.

Fair warning: Anthropic keeps moving this menu around. If Skills aren’t under Customize, look under Settings → Capabilities.

Your first real run. Do not test it with “use my skill.” Ask for the work like you normally would. “Draft the quarterly review for the Nolan account.” Then watch Claude’s thinking in the sidebar for a line that says “Using [skill name].” That’s proof it fired. No line, no skill… you just got regular Claude with extra steps.

The mistake that makes it fail. A vague description on the Skill itself.

Claude decides whether to load a Skill by reading its description. Something like “helps with docs” will never trigger. Something like “creates quarterly client reviews using the pricing structure, past QBRs, and the client’s Slack history” will fire every time. Write your description like the trigger it is.

Second thing to know: Skills don’t sync. One built in the app doesn’t exist in Claude Code, and vice versa. They’re also private to your account… you can’t hand one to your team by sharing a link.

Thing 2. Set up one scheduled task.

This is the one that makes Claude work while you’re asleep.

Read this before you go hunting. Scheduled tasks live inside Claude Cowork, not regular chat. If you’ve been looking for them in your normal chat window, that’s why they’re not there. Cowork is currently rolling out to Pro accounts… Max users have it now. If you don’t see it yet, that’s not you doing anything wrong.

The exact click path:

1. Go to claude.ai and open the Home tab.
2. In the message box, select “Cowork” at the bottom left.
3. Click “Scheduled” in the left sidebar.
4. Click “New task” in the upper right.
5. Pick “Create with Claude” (it asks you multiple-choice questions and builds the task for you) or “Set up manually” (name, prompt, approval mode, frequency, optionally a model and folder).

Shortcut: type /schedule in any Cowork session and skip straight to the setup.

Your frequency options: hourly, daily, weekly, on weekdays, or manually. “Manually” means the task only runs when you press the button, which is more useful than it sounds when you want a repeatable task you kick off yourself.

The part that surprises people. Scheduled tasks run remotely on Anthropic’s servers. Your laptop does not need to be awake and the app does not need to be open. Close your laptop, go to sleep, the task still runs at 6am. That’s different from what a lot of older posts still say.

Your first real run: a daily briefing at 7am.

Set up a task that says: “Summarize my Slack messages, emails, and calendar events from the last 24 hours. Group by sender. Flag anything that needs a reply today. Keep it under 200 words.”

Run it daily. That’s the task that earns its keep in week one, guaranteed. The other one worth building right after: a weekly report from your connected tools, delivered the morning before your team meeting (not Friday afternoon when everyone’s checked out).

The mistake that makes it fail. Writing the prompt like a note to yourself.

Nobody is going to be there to answer follow-up questions when a scheduled task runs at 6am. So “check my email” comes back useless. You have to write it as if you’re leaving detailed instructions for a substitute assistant… every constraint, every format, every “flag this” or “ignore that” rule spelled out.

The other thing: your tools have to already be connected before you schedule the task. Run the prompt by hand in a normal chat first. If it works there, schedule it. If it doesn’t, fix the prompt before you automate it.

Thing 3. Build one Artifact and publish it.

Artifacts are things Claude builds in a live window next to your chat… calculators, interactive tools, one-page websites, dashboards, quizzes. The building part is well-covered elsewhere. The half nobody covers is publishing, which turns something on your screen into a link you can send a client.

Plans, and read this table carefully because it’s backwards:

A Free account can publish an artifact publicly. An Enterprise account cannot. Also worth knowing: artifacts aren’t on mobile. Do this at a desktop.

To build one:

1. Click Artifacts in the sidebar (this cues Claude that you’re building, not chatting).
2. Describe the problem you want solved, not the code you want written. “I want a pricing calculator my clients can use to estimate their project” is a good prompt. “Build me a React component that…” is going to send you sideways.
3. Let Claude ask clarifying questions. Answer them.
4. Say “Can you create this for me?” It builds in a live window on the right.

Claude only makes an artifact when the thing is significant and self-contained (usually over 15 lines of code). Small asks stay in the chat.

To publish it:

1. Go to the artifact.
2. Check you’re on the right version. Artifacts keep every version you’ve made. You publish the version you’re currently viewing.
3. Click Publish and copy the link.

Whoever you send that link to does not need a Claude account. They open it and use it. That’s the whole point.

Two more buttons worth knowing. Customize lets other Claude users remix your artifact into their own version (yours stays untouched). Get embed code shows up after publishing so you can drop it on your own site. Embeds need an “Allowed domains” list, comma-separated.

Your first real run. Build the thing you keep rebuilding in a spreadsheet.

A pricing calculator for your services. An intake form for new clients. A quiz that sorts people into the right package. A checklist tool your team uses on every project. Anything you’ve been re-doing by hand.

Build it. Publish it. Send the link to one person. Watch them use it. That loop… build → link → someone else’s screen… is the point. Most people build ten artifacts and send zero.

The mistake that makes it fail… and this one is permanent.

Unpublishing is one-way. Once you unpublish an artifact, you cannot publish that same artifact again. Anthropic permanently deletes all associated storage data. No undo. If you unpublish something to clean it up and later want the link back, you have to build a brand new artifact and send everyone a brand new URL.

Rule of thumb: unpublish only when you truly mean it. If you’re unsure, leave it up.

Thing 4. Add one plugin.

Plugins bundle Skills, connectors, and specialized agents into a single ready-made setup for a role or task. Anthropic’s own line is the clearest: “You get a ready-to-go setup from your first conversation.”

If Skill / Connector / Plugin all blur together, here’s the difference in one table. Keep it.

Plan you need: any paid plan. Pro, Max, Team, or Enterprise. Not on Free.

To install one in the app:

1. Customize in the left sidebar.
2. Click the Plugins tab.
3. Click Browse plugins, then Install on the one you want.

The Knowledge Work marketplace is added by default so there’s a starter set to browse. You can also add Anthropic’s Financial Services, Legal, and Life Sciences marketplaces. Full catalog is at claude.com/plugins.

To add one from GitHub: Customize → Plugins → Personal plugins → + → “Add marketplace” → “Add from a repository” → paste the URL.

Read this warning from Anthropic’s own docs. “Anthropic doesn’t control what MCP servers, files, or other software are included in plugins and can’t verify that they work as intended.”

Translation: a plugin is somebody else’s code running with your access. Install from Anthropic’s own marketplaces first. If you’re pulling from a random GitHub repo, check who made it before you let it touch your Gmail.

Your first real run. Install one plugin that matches what you did last Tuesday. Then, in a new conversation, hand it a real task from your actual week. Not a test question… a real thing.

Then judge it on one measure: did it visibly change the output, or did it just sit there? A plugin that doesn’t visibly change how Claude answers is not helping. Uninstall it. Try another.

The mistake that makes it fail. Installing eight in one afternoon.

Every plugin adds instructions and tools to the pile. When they overlap or contradict, Claude gets less sure what you want, not more. Install one. Use it for a week. Then add the second.

Thing 5. Install Claude in Chrome.

This is the one almost nobody has installed. It’s also the one that changes your day the most.

It’s called Claude in Chrome. In the Chrome Web Store the listing is titled just “Claude,” published by Anthropic. Double-check the publisher name before you install anything… there are copycats.

What you need: it’s currently in beta. No waitlist. Works on all paid plans (Pro, Max, Team, Enterprise). Chrome only. Not Edge, not Brave, not Arc, not mobile.

The install, start to finish:

1. Open Chrome. Actual Chrome.
2. Search “Claude” in the Chrome Web Store, or go directly to the listing published by Anthropic.
3. Click Add to Chrome.
4. Sign in with your Claude account.
5. Pin it. Click the puzzle piece icon in your Chrome toolbar and pin Claude. Skip this and you will forget it exists within 48 hours.
6. Grant permissions when it asks.

What it can actually do, and this is far more than “chat about this page”:

• Read the page you’re on, including things below the fold.
• Click buttons, type into fields, fill out forms.
• Navigate between pages on its own. Take screenshots.
• Open, close, and switch tabs. Manage several at once.
• Read the browser console (gift for anyone debugging their own site).
• Download files. Record workflows so you can repeat them.
• Run scheduled tasks in the browser.

Now the safety part… don’t skim it. This is an AI with hands on your browser. Anthropic built real limits in.

• Blocked outright: adult content sites, known pirated content sites. Claude will not go there.
• Permission required: Claude asks before it touches financial sites.
• Confirmation required: for high-risk actions like downloading a file or entering sensitive information, it stops and asks you first.
• Flat-out forbidden: stock trading, investment transactions, bypassing captchas, entering sensitive data on your behalf, scraping facial images.
• Every action gets checked for risk and for hidden instructions before it runs.

Where the accountability line sits. You are accountable for anything Claude publishes, sends, or purchases on your behalf. Guardrails reduce risk. They do not move responsibility. Anthropic’s own advice: avoid financial transactions, password management, or anything involving sensitive personal data.

One more thing worth knowing: browser interactions are more compute-intensive than regular chats. The extension eats more of your monthly limit. Save it for tasks worth the spend.

Your first real run. The form you have filled out forty times.

Pick one of these three:

• Fill out an application you’ve typed the same answers into before.
• Enter information into your CRM after a call… the job everyone avoids until Friday.
• Book something. A table, an appointment, a flight search you’ve been putting off.

Start with the CRM one. Repetitive, data already in front of you, and you’ll know within 90 seconds whether this changes your week. Watch the first run all the way through. You’re learning what it does well before you trust it with anything.

The mistake that makes it fail. Being vague about which tab.

It’s a browser agent. It needs to know where it’s working. “Update the client record” is a coin flip. “On the tab with the Salesforce contact page for Nolan open, update the notes field with the recap I’m pasting below” works every time. Name the page, the field, and exactly what goes in it.

The other rule: don’t hand it a task on a page you’re not watching. If it stops to ask you a question, that’s the safety system doing its job… it’s not a bug.

The order I’d do them in

Five things is a lot for one sitting. Don’t do one sitting.

1. Flip the code execution toggle (Settings → Capabilities). Thirty seconds. Unlocks two of the five.
2. Install Claude in Chrome and give it one CRM entry or one form fill. Fastest payoff on the list.
3. Build one Skill by talking to Claude about the thing you explain over and over.
4. Publish one Artifact and send the link to one person.
5. Install one plugin and give it a real task in a fresh chat.
6. Set one scheduled task, the daily briefing, if Cowork has rolled out to your plan.

One a day for a week. By Sunday you’re past where 95% of people ever get with this tool, because most people stop at typing questions into a box.

If you only do one thing

Install Claude in Chrome and use it on one real task today. No plan confusion, no beta rollout to wait on, no file format to get right.

Add to Chrome. Pin it. Hand it the boring thing you were going to do anyway.