Claude Code Cheat Sheet for Architects and Designers

What is Claude?

Claude is an AI assistant made by Anthropic. You give it instructions in plain English and it reads files, writes documents, runs research, and handles multi-step workflows. With the right skills installed, it becomes a specialist — for site analysis, space programming, specifications, and more.

We use it for every project at Alpaca Labs — from deploying sites to producing client deliverables.

Two ways to use it

There are two interfaces to Claude — pick the one that fits how you work:

	Claude Desktop	Claude Code
Interface	Desktop app with a chat window	Terminal — text commands
Best for	Most team members, non-technical users	Developers, power users, automation
Skills	Browse and install from a plugin library	`claude install` one-liner
Setup	Download the app, log in	Install via npm, log in from terminal

We recommend starting with Claude Desktop — it’s the same Claude, same skills, no terminal required. Everything in this guide works in both interfaces. Where the steps differ, we’ll show both paths.

Before You Start

You’ll need a Claude subscription (Pro, Max, Team, or Enterprise). Skills are free and open-source — you only pay for the Claude subscription.

Claude Desktop users: Download Claude Desktop and log in. You can skip the terminal setup below and jump straight to Try It Out.

Claude Code users: You’ll need two things installed:

A terminal. On Mac, open Terminal.app (search for “Terminal” in Spotlight). On Windows, open PowerShell (search for “PowerShell” in the Start menu).
Node.js. Download the LTS version (the one on the left) from nodejs.org. Accept all defaults during installation. This also installs npm, the package manager that installs Claude Code.

Getting Started

Claude Desktop users — skip to Try It Out. You’re already set up.

Claude Code users — open your terminal and run:

# Install Claude Code globally
npm install -g @anthropic-ai/claude-code

# Navigate to a project folder
# Mac/Linux:
cd ~/Documents/my-project
# Windows PowerShell:
cd $HOME\Documents\my-project

# Launch Claude
claude

# Log in to your account
/login

After typing claude and pressing Enter, you’ll see a prompt where you can type requests in plain English. Claude will read the files in your current folder and respond.

Try It Out

Once Claude is running (in either Desktop or the terminal), just type what you need:

What files are in this folder?
Read the PDF spec in this folder and summarize the finishes schedule
Organize these files by date into subfolders
Create a spreadsheet of all the room names and areas from this document

There’s no special syntax — write like you’re messaging a colleague.

Essential Commands

These slash commands control your session. Start with these five:

Command	What it does
`/help`	Show all available commands
`/clear`	Clear the current conversation and start fresh
`/compact`	Compress earlier parts of the conversation so long sessions keep running smoothly
`/resume`	Resume your most recent conversation — pick up where you left off
`/exit`	Exit Claude Code

Skills and Plugins

Claude has a plugin system that lets you extend it with domain-specific capabilities. Here’s how the pieces fit together:

Skill = a specialist capability you invoke with a slash command. A skill tells Claude what data to reference, what phases to follow, and what to output. For example, /environmental-analysis turns Claude into a site climate researcher.
Plugin = a group of related skills. Some plugins include a skill that runs the others in sequence — for example, /site-due-diligence-nyc runs all four site planning skills plus zoning in one go.

Skills for architects and designers

We’ve built and open-sourced 16 skills across 6 plugins for AEC workflows:

Plugin	Skills
Site Planning	Environmental Analysis, Mobility Analysis, Demographics Analysis, Historic Analysis, `/site-due-diligence-nyc`
Zoning Analysis	Zoning Analysis NYC
Programming	Workplace Programmer, Occupancy Calculator, `/space-program`
Specifications	Spec Writer
Product & Materials Research	Product Research, Spec Fetch, Spec Cleanup, PDF Parser, Image Processor, `/spec-package`
Presentations	Slide Deck Generator, Color Palette Generator

Installing skills

Claude Desktop: Open Customize in the left sidebar → Browse plugins → click + → Add marketplace from GitHub and enter AlpacaLabsLLC/skills-for-architects. The plugins appear under the Personal tab — click Install on the ones you want. No GitHub account required.

Claude Code (terminal): Install with one command:

claude install github:AlpacaLabsLLC/skills-for-architects

Browse the full collection at /skills.

Building your own skills

A skill is a SKILL.md file inside a named folder. When you type the command, Claude reads the file and follows its instructions.

Here’s an example that organizes a client folder for handoff:

# Clean Up Client Folder

Organize a client project folder for handoff.

## Steps
1. Move loose PDFs into a `Submittals/` subfolder
2. Archive old Revit backups (*.0001.rvt) into `_Archive/`
3. Rename files to match the project naming convention
4. Report what was moved and renamed

Some ideas for your practice:

Submittal log — parse a spec section and generate a submittal register with CSI divisions, responsible parties, and due dates
Meeting minutes — turn rough notes into formatted OAC minutes with action items, owners, and deadlines
Punch list — walk a photo folder and generate a punch list from image metadata and descriptions
Code review — check a floor plan against IBC egress requirements and flag non-compliant conditions
Fee proposals — calculate staffing, hours, and fee from a scope description and your standard rate table

If you’ve done it three times, make it a skill.

CLAUDE.md — Teaching Claude About Your Projects

The CLAUDE.md file is how you give Claude standing instructions — rules and preferences it follows automatically in every session. Think of it as a note you leave for Claude so it remembers how you like things done.

Location	Scope
Your home folder: `CLAUDE.md` in `~/.claude/` (Mac/Linux) or `%USERPROFILE%\.claude\` (Windows)	Global — applies to all sessions
Your project folder: `CLAUDE.md` in the project root	Per-project — applies when working in that folder

Use it to define project structure, naming conventions, deploy commands, and team preferences. Claude reads it at the start of every session.

MCP Servers — Connecting External Tools (Advanced)

This section is optional and requires the terminal. You can use Claude without MCP servers — but they unlock powerful integrations.

MCP (Model Context Protocol) servers let Claude connect to external services like Google Drive, spreadsheets, and other tools your team already uses.

# Check connected servers
/mcp

To set up a server, you add a configuration to a settings file. Here’s an example that connects Google Drive:

{
  "mcpServers": {
    "google-drive": {
      "command": "npx",
      "args": ["-y", "@anthropic-ai/google-drive-mcp"]
    }
  }
}

(npx is a tool included with Node.js that downloads and runs packages automatically.)

Once connected, you can ask Claude to find, read, and update files in your Google Drive without leaving the conversation.

Tips

Be specific. “Organize the FF&E schedule by room number” works better than “clean up the spreadsheet.”
Let Claude read first. It works best when it understands your files before making changes.
Use /compact often. Long conversations fill up Claude’s working memory — /compact compresses earlier parts so it stays responsive.
Share your CLAUDE.md with the team. It helps every team member (and Claude) stay aligned.

Full Command Reference

For when you need it — the complete list of slash commands.

Command	What it does
`/help`	Show all available commands
`/clear`	Clear the conversation and start fresh
`/compact`	Compress conversation to free up working memory
`/resume`	Resume your most recent conversation
`/exit`	Exit Claude Code
`/login`	Authenticate with your account
`/skills`	List all available skills
`/mcp`	Show connected MCP servers
`/config`	Open Claude Code settings
`/cost`	Show token usage and cost for the current session
`/fast`	Toggle fast mode (same model, faster output)