Glossary
The words, in plain English
Every term the guide leans on, defined once, honestly. If a definition still reads like jargon, tell me and I will rewrite it.
link any term: /glossary#token
- Claude
- The AI assistant this whole site is about. It comes in several surfaces: the app for conversation and thinking, Claude Code for building in your files, and Cowork for multi-step jobs. Same brain, different rooms.
- The Claude app (desktop)
- Where you chat, keep Projects, and connect your tools. For thinking, drafting, analysing and planning. If a task is a conversation, it lives here.
- Claude Code
- Claude working inside your project files instead of replying in a chat box. It reads, creates and edits real files on your machine, with your approval. If chat is advice, Claude Code is production.
- Cowork
- A desktop feature where you describe a multi-step job and Claude plans and runs it across your files and connected apps, asking before it acts. Built for people who never want to see a terminal. On paid plans.
- VS Code
- A free program that shows you your project as a file tree, with every change visible. Not a programming course: it is simply the room where you watch Claude work. The guide teaches it as a workbench, nothing more.
- The terminal
- The text window where you type commands to your computer directly (Terminal on Mac, PowerShell on Windows). You need about five commands’ worth of it, and the guide gives you each one.
- CLAUDE.md
- A plain text file at the top of your workspace that teaches Claude your business, your voice and your rules. Claude reads it automatically at the start of every session. The single most useful thing you will build; chapter 6 builds it with you.
- Memory
- What Claude saves about you as you work, so you stop re-introducing yourself. You can see and edit it in settings, and import it from ChatGPT, Gemini or Grok with the official import flow. For anything that matters, prefer the CLAUDE.md: a file you own beats a black box.
- Workspace / project folder
- One folder that holds a project’s files, with your CLAUDE.md at the top. Give Claude the folder and it has your world; keep things in it and nothing gets lost in chat history.
- Context window
- Claude’s working desk space: everything it can hold in mind at once in a conversation, measured in tokens. When it fills, older detail falls off the edge, which is why long chats drift and why the guide teaches fresh sessions per task.
- Token
- The unit AI text is measured in. Roughly three-quarters of a word; “jewellery” is a couple of tokens. Plans meter usage in tokens, and the context window is a token budget.
- MCP
- Model Context Protocol: the plug-socket standard that lets Claude connect to real tools and data (your shop, your files, your calendar) instead of guessing. MCPs sound technical; they are plug sockets.
- Connector
- An MCP wired into your Claude app: Shopify, Google Drive, your calendar. Once connected, Claude can check the real thing rather than asking you to paste it in.
- Skill
- A reusable move: instructions you (or someone else) wrote down once, that Claude calls on whenever the job comes up, instead of you re-explaining. A skill is not a prompt; it is a recipe card.
- Agent / subagent
- Claude given a job description: a goal, tools and a process, so it runs a whole task (“research these five competitors and build me a comparison”) rather than answering one question. A subagent is a helper agent spawned for a piece of the work.
- Plugin
- A bundle of skills, commands and agents installed in one go, like an app for your Claude setup. Chapter 9 covers where to find them and what to trust.
- Slash command
- A shortcut typed into Claude Code starting with / (like /help). Each one triggers a saved behaviour; you can make your own.
- PATH
- The list of places your computer looks when you type a command. “Command not found” straight after an install usually just means the terminal has not refreshed this list: close the window, open a new one.
- Homebrew
- The Mac’s app shop for command-line tools, driven from the terminal. The guide uses it to install what Claude Code needs; you will touch it once and rarely again.
- WSL2
- A way of running Linux inside Windows. One of two install routes on a PC; the guide recommends the native route first and keeps this as the fallback.
- Git / GitHub / repo
- Git is a save-history for a folder of files, so any change can be undone. GitHub is the site where those folders (repos, short for repositories) live online. You need only the basics, and Claude handles most of it for you.
- Markdown (.md)
- Plain text with light formatting marks (# for a heading, ** for bold). What your CLAUDE.md and most working notes are written in. If you can write an email, you can write it.
- Brief (versus prompt)
- What you give Claude. A prompt is an incantation you copied from someone; a brief is what a director gives: the outcome, the constraints, what good looks like, what to never do. This site teaches briefing.
- Parametric
- Built with controls instead of hard-coded. A parametric Shopify section has settings for heading, images, spacing and layout, so you build it once and control it forever. The section is not the design; the controls are the design.
- Reverse engineering (a website)
- Studying how something works underneath (structure, rhythm, navigation logic, mobile behaviour) rather than copying how it looks. Steal the UX logic; never steal the design.
- Slop
- What AI produces when nobody with taste is directing it: averaged, glowing, interchangeable. The reason this site exists is that the fix is not avoiding AI; it is keeping the say.