Cindy Zhu.
← all free guides
Getting started Claude

How to master Claude in one weekend

Hey, it's Cindy 🌱 This guide walks you through 6 weekend steps that take Claude from a chat tool to a real teammate, no coding and no overwhelm, just two afternoons.

Most people use Claude in a browser tab, ask a few questions, and decide AI isn't ready yet. The truth is they're using the wrong version. This guide walks you through 6 weekend steps that take Claude from chat tool to teammate. No coding, no overwhelm. Just two afternoons.

the outcome

βœ… What you'll have by Sunday night

  • Claude Code installed and running on your machine
  • Three skills that auto-load whenever they're needed
  • One scheduled Routine running on Anthropic's cloud
  • Three of your apps connected directly to Claude
  • A CLAUDE.md file that turns Claude into your second brain every session

No coding required. Just plain English the whole way.


what you need

🧰 Before you start

  • A Mac or Windows machine
  • A Claude account. Free tier works for steps 1, 2, 3, 5, 6. Step 4 (Routines) requires Pro, Max, Team, or Enterprise.
  • 10 minutes of admin patience to install one app and create a few files

the setup

πŸ› οΈ Step 1: install Claude Code

Claude Code is the version of Claude that can read your files, send your emails, post in your Slack, and run scheduled jobs. Same Claude under the hood, way more capability.

Install in 3 steps:

  1. Go to claude.ai/install
  2. Copy the one-line install command for your operating system
  3. Paste it into Terminal (Mac/Linux) or PowerShell (Windows) and press enter

Launch Claude Code: when the install finishes, type claude in your terminal and press enter. A browser tab opens for you to log in. Authorize, return to the terminal, you're in.

Test it: type what can you do in this folder? and press enter. Claude lists everything it can read, write, and run from where you ran the command.

Full docs at docs.claude.com/en/docs/claude-code/quickstart


pick your surface

🎯 Step 2: pick the right Claude for the job

Anthropic ships Claude in four places. Each is built for a different job. Picking the right one is half the work.

1. Chat (claude.ai)
Quick questions, brainstorming, throwaway thinking. Inside Chat there's a feature called Projects: a folder where you can attach files and keep a recurring conversation that always remembers them.

2. Claude Design (claude.ai/design)
Released April 2026. Built for people without a design background. Describe what you want, Claude builds the first draft, you refine inline. Exports to Canva, PPTX, PDF, HTML.

3. Cowork
Task hand-off. You describe a job, Claude works on it independently in the background, and delivers a finished result.

4. Claude Code
The version you just installed. Skills, MCPs, scheduled Routines, the whole automation layer.

Decision shortcut:

"I have a question" β†’ Chat
"I need a design or visual" β†’ Claude Design
"I need a task done in the background" β†’ Cowork
"I want to automate something forever" β†’ Claude Code

stop retyping

🧠 Step 3: save your prompts as skills

Every reel script needs the same setup prompt. Every email needs the same voice rules. Every spreadsheet analysis needs the same KPI list. Retyping each one from scratch wastes the best part of your day.

A skill fixes this. A skill is a small file Claude reads automatically the moment a task matches its description.

Where skills live:

  • ~/.claude/skills/ for skills available in every project
  • .claude/skills/ (inside any folder) for project-specific skills

File format:

Each skill is a folder with a SKILL.md file inside. The file has a short frontmatter header that tells Claude when to fire it, then a body with the actual instructions.

Example: a brand-voice skill at ~/.claude/skills/brand-voice/SKILL.md

πŸ—£οΈ the brand-voice skill
---
name: brand-voice
description: Use whenever writing copy, captions, scripts, or emails. Catches AI clichΓ© and enforces my voice rules.
---

You're writing copy for me. Voice rules:
- Short sentences. Short paragraphs.
- Casual, friend-to-friend tone.
- No em dashes anywhere.

Banned words and phrases (never use):
- delve, dive into, elevate, navigate the landscape, unlock the potential
- boundaries, journey, transform, harness, leverage (as a verb)

Always include a specific example or number when making a claim.

Three skills to write this weekend:

  1. brand-voice (above) catches AI clichΓ© in everything you write
  2. summary takes any long document, transcript, or article and gives you the 5 most important takeaways
  3. planner given a goal, breaks it into a weekly action plan with dates and dependencies

Each takes 10 minutes to write. Saves an hour a day.

Trigger a skill manually: type /skill-name (e.g., /brand-voice). Otherwise Claude auto-fires it when the task matches.

Pro tip: start every skill description with "Use whenever..." or "Use when...". Clarity in the description equals reliability in the trigger.

while you sleep

⏰ Step 4: schedule Claude to run while you sleep (Routines)

The biggest leverage shows up when Claude works without you opening anything.

The feature is called Routines. It runs on Anthropic's cloud, not your laptop. You describe a recurring job, set the schedule, walk away. Laptop can be off. Phone in another room. The job runs.

Requirements: Claude Pro, Max, Team, or Enterprise with Claude Code on the web enabled.

Set up your first Routine:

  1. Open Claude Code (claude in terminal)
  2. Type /schedule followed by a plain-English description. Examples:
    • /schedule daily inbox triage at 7am
    • /schedule weekly project recap every Friday at 4pm
    • /schedule monthly P&L summary on the 1st of each month
  3. Claude walks you through the rest (what data to pull, where to deliver the output, format)
  4. Confirm. Routine is live.

Useful follow-up commands:

  • /schedule list see all your active Routines
  • /schedule update change an existing one
  • /schedule run trigger immediately for testing

You can also manage Routines visually at claude.ai/code/routines. The web interface lets you trigger them from API calls or GitHub events on top of the schedule.

Example Routine I'm wiring up: every weekday at 7am, Claude pulls my Meta ad spend, ROAS, and top 3 creatives from yesterday. Writes a one-paragraph summary of what changed. Drops it in my Slack. I read it with breakfast.

Other Routines worth setting up:

  • Daily inbox triage (every morning, summarize unread Gmail, flag anything urgent)
  • Weekly content recap (every Friday, what shipped + what's overdue + next week's priorities)
  • Monthly P&L (every 1st, pull Stripe + ad spend, write a one-page financial summary)

the connectors

πŸ”Œ Step 5: plug Claude into the apps you already use (MCPs)

Skills are how Claude knows what to do. MCPs are how Claude reaches into the apps that hold your work.

MCP stands for Model Context Protocol. In plain English, an MCP is a cable that connects Claude to a tool you already use. With MCPs installed, you stop copying and pasting between tabs. Claude reads from Notion, posts in Slack, pulls from Stripe, and writes in Drive, directly.

To install your first MCP:

  1. In your terminal, type claude mcp add notion (or any other tool name)
  2. Claude walks you through the auth flow. Log in, authorize, done.
  3. Test it: type pull my latest content board from Notion and tell me what's missing
Pick MCPs by problem, not hype. The mistake every new user makes is installing 15 MCPs they'll never use. Don't. Ask one question: what do I do every week that makes me open three or more tabs? Connect those tools first.

Common starter MCPs:

  • notion content boards, project trackers, knowledge base
  • gmail inbox triage, draft replies, summarize threads
  • slack read channels, post messages, summarize discussions
  • google-drive read and write files in Drive
  • stripe pull revenue, list customers, track payments
  • github issues, PRs, commits

Browse the full directory: mcpservers.org lists every available MCP server. Updates weekly.

Example prompts after install:

  • "Pull last month's Stripe revenue and post it in #wins" β†’ 4 seconds
  • "Draft a reply to every unread email tagged 'client' in Gmail" β†’ 12 seconds
  • "Summarize this week's content board in Notion and flag what's overdue" β†’ 20 seconds

the foundation

πŸ“ Step 6: write your CLAUDE.md

This is the file that makes everything else click. Skills, MCPs, Routines all work better when CLAUDE.md is in place.

CLAUDE.md is the foundation file Claude reads automatically every single session. It's where you tell Claude who you are, what you do, what you're working on, and how you like to be talked to.

Three valid locations:

  • ~/.claude/CLAUDE.md global, loads in every project
  • ./CLAUDE.md project root, loads in that folder only
  • ./.claude/CLAUDE.md alternative project location

Project-level rules take precedence over global. You can have both.

To create yours:

  1. Open any plain-text editor (VS Code, Cursor, Antigravity)
  2. Create a new file called CLAUDE.md
  3. Save it in ~/.claude/ (your home directory)
  4. Use the template below

Paste-ready template:

πŸ“ the CLAUDE.md template
# Who I am

I'm [your name]. I'm a [your role]. I run [your business or project].

# Who I serve

[Describe your audience or clients in 1-2 lines.]

# What I'm working on this week

- [Goal 1]
- [Goal 2]
- [Goal 3]

# Voice rules

- Short sentences. Short paragraphs.
- Casual but confident. Educational, not preachy.
- No em dashes. Use commas, periods, or parentheses.

# Banned words

delve, dive into, elevate, navigate the landscape, unlock the potential, harness, leverage (as a verb), boundaries, journey, transform

# My skills

- /script
- /summary
- /planner
- brand-voice

# My MCPs

- notion
- slack
- gmail

# Default behaviors

- Always include a specific example or number with any claim
- When uncertain, ask one clarifying question instead of guessing
- Don't write more than 200 words unless I ask for more

When to use CLAUDE.md vs a skill:

Use CLAUDE.md for persistent instructions every session (who you are, voice, defaults). Use skills for task-specific workflows that load on demand (writing a script, auditing an ad, generating a carousel).

Size guidance: target under 200 lines per CLAUDE.md. Longer works, but tighter loads faster.

Bonus: per-project CLAUDE.md.

If you work across multiple projects, give each one its own folder and its own CLAUDE.md. Open the folder, Claude becomes that project's brain. Same setup, different context.


the cheat sheet

πŸ“‹ Quick reference

StepCommandWhat it does
Install Claude Codeclaude.ai/installOne-line install
Run Claude CodeclaudeOpens an interactive session
Create a skill~/.claude/skills/<name>/SKILL.mdAuto-loaded prompt
Trigger a skill/<skill-name>Fires the skill manually
Add an MCPclaude mcp add <tool>Connects an app
Schedule a Routine/schedule [description]Runs in Anthropic's cloud
Write CLAUDE.md~/.claude/CLAUDE.mdFoundation file, every session
Want more like this? Follow @cindiezhu for daily AI you can actually use.