Set Up Claude for Your Design Work
Give Claude permanent context about your role, your users, and your output preferences. Every session starts already knowing your work.
The situation
You have been using Claude like a search engine. You ask a question, get a generic answer, and close the tab. It helps sometimes. But it does not know you are a designer, who you design for, or how you like to receive feedback.
Every future Claude session will start already knowing your role, your users, and how you want output delivered. You write it once. It loads every time.
What you walk away with
A working agreement that tells Claude who you are, who you design for, and how you like to work
The exact template filled with your real context, not a generic placeholder
A test prompt that confirms your agreement is specific enough to produce useful output
Create your workspace
Set up the folder or Project that will hold your working agreement. Follow the demo for your journey. CLAUDE.md is a plain text file Claude reads automatically at the start of every session.
# First: confirm Claude Code is installed
# (if you see "command not found", install it at claude.ai/download first)
claude --version
# Then: create your project folder and start Claude Code
mkdir -p ~/my-design-project
cd ~/my-design-project
touch CLAUDE.md
claudeWrite your working agreement
Paste this template and fill in your actual details. Generic answers come from generic context. The more specific you are about your users and constraints, the more useful Claude becomes.
# Paste this into Project Instructions (app) or CLAUDE.md (terminal/IDE)
You are assisting a UI/UX designer. Here is the context you need.
ROLE
I am a [junior / mid / senior] UI/UX designer working [at an agency / in-house / freelance].
My day-to-day includes [e.g. mobile app design, dashboard UX, e-commerce checkout flows].
USERS
The people I design for are [describe: e.g. first-time smartphone users on
low-end Android devices in Bangladesh, enterprise procurement managers in Europe].
Key constraints: [e.g. limited data plans, low digital literacy, RTL support].
TOOLS AND WORKFLOW
My primary design tool is [Figma / Adobe XD / Sketch].
I hand off to [developers / a no-code team / build it myself].
OUTPUT RULES
- Write in plain English. Define jargon on first use.
- When I ask for copy, give me 2-3 variants.
- When critiquing my work, be direct. Skip encouragement.
- When I paste a brief, ask clarifying questions before generating anything.
- Flag when a suggestion depends on something you cannot verify.Test it before moving on
Send this prompt now. If the response feels generic, your USERS section needs more specificity. Iterate until Claude's questions feel like they came from a colleague who knows your work.
# Paste this into your Claude session:
I've just been handed a brief. The client wants "a modern, clean app for managing invoices."
What questions should I be asking before I open Figma?Keep it under 300 words and update it when context changes
Most designers find that shorter agreements work as well as longer ones. Beyond 300 words you are likely adding noise rather than signal. Stale context is worse than no context: when you change clients, platforms, or roles, update it. Review your USERS section right now: if it says "general audience" or nothing specific, rewrite it before moving to the next guide.
- Leaving USERS vague. "General audience" gives Claude no user constraints to apply. Name who they are and what they are trying to do.
- Copying someone else's template without adapting it to your actual users. The agreement only works if it describes your context, not a generic designer.
- Never revisiting it after the first session. When you change clients, platforms, or roles, update it. A stale agreement is worse than none.
What's next?
Decode any design briefNew guides, when they ship
One email, roughly weekly. CLAUDE.md templates, workflows I actually use, and the cut-for-length stuff that does not make the public guides. One-click unsubscribe.
Or follow on Substack