ux playground

playground 101

3 of 4

Start your first project

Start a project, see what comes next, and learn how to steer the agent.

Start your project

Done when

Your project card is on main and shows on the playground hub.

Claude Code

In the Claude desktop app's Code tab, start a new session on the ux-playground folder.

Cursor

Open the ux-playground folder with File › Open Folder, in a new window. Cursor keeps one branch per window, so give each project its own window rather than a new chat.

Then say:

I want a new project.

It runs the new-project skill:

  1. It checks whether the problem already has a project. Say "Patients arrive not ready for their scan", and if a project for that exists, you'll work there with whoever started it. One project is one problem.
  2. It asks which surfaces the problem touches. One problem often shows up in several places, like the dashboard, the app and a reminder email, and one project can cover all of them. It starts your project from those kits in its own copy of the repo, so you never manage branches.
  3. It runs your project and shows you the toolbar: synthetic patient, journey stop and concept. Then it starts asking about the problem.

Push your work

Your project lives only on your laptop until you push it. Get it onto main on day one, even if it's just a problem and a kit, so others can find it and join in.

Say:

Push my work.

The agent saves your work, brings in the latest from main, pushes it to GitHub and opens a pull request. A pull request that only touches your own project merges on its own once its checks pass. A few minutes later, your card is on the playground hub (opens another site).

  • Dashboard
  • Mobile app
  • Comms

Appointment prep

Some patients arrive not ready for their scan, so the visit runs late or is rebooked.

From then on, end every session by saying "Push my work". Why it matters.

See what comes next

Every project goes through the same six steps, whoever makes it. Small changes too; each step is just shorter. You don't need to read them now. When you reach one, your agent tells you and links its example.

  1. Understand the problemYou approve the brief
  2. Set a directionYou pick the direction
  3. Explore conceptsYou steer as it builds
  4. Check the workYou pick which findings to fix
  5. Share and testYou choose what feedback you need from real people
  6. Hand it offYou decide when it's ready

See every step, with an example of each →

Work with the agent

Things you can ask for any time

SayWhat happens
"Review this screen"Three reviewers look at something that already exists, without starting a project
"Walk the synthetic patients through this"Each patient tries the concept, and you see who gets stuck and where
"Check the copy"Words checked against the Prenuvo lexicon and plain language
"What am I missing?"The happy, sad and edge cases the concept covers, and the ones it doesn't
"Make a component board"Ten or more options for one component, side by side
"Debate this"Agents argue competing directions so you can choose
"Hand this off"The handoff, once the critique and patient test have run

Wait, I don't see Figma here… why?

Because the design happens in the prototype. Real screens, real states and synthetic patients answer what a static frame can't: what Skye sees on his phone, what happens on the error path, how the card looks at every journey stop.

We see Figma as a tool for visual fine-tuning: spacing, type, a detail you want to push by hand. It's best connected to your agent through MCP. Paste a Figma link and the agent reads the frame, or ask it to send a screen from the prototype into Figma to refine. Connect it in Get set up.