FLSoft Academy

MODULE 2 · PROMPTING & CONTEXT / 20 MIN Free NOVEMBER 2026 EDITION · CLAUDE CODE 2.1.318

Images, screenshots and visual work

Give Claude screenshots, mockups and diagrams, and build a UI loop where it compares its result to a target image.

Updated 5 November 2026 · Reviewed 5 November 2026

What you’ll be able to do

Use screenshots and mockups as context, and run a “match this design” loop that reliably gets close to the target.

Three ways to give Claude an image

  1. Paste an image from your clipboard into the prompt: Ctrl+V on macOS and Linux (Cmd+V in iTerm2), Alt+V on Windows and WSL. An [Image #1] chip appears, which you can refer to in your prompt.
  2. Drag and drop an image file into the terminal.
  3. Reference a path: “Look at @designs/checkout.png and…”

What images are good for

  • Bug reports: “This modal overflows on mobile. Here’s what it looks like. Expected: it scrolls internally.”
  • Mockups to code: “Build this card component to match the image.”
  • Diagrams: a whiteboard photo of an architecture becomes a starting point: “Turn this into a Mermaid diagram and a list of services.”
  • Error dialogs you can’t copy as text.

Exercise: match a design

Create a tiny static page project:

mkdir visual-practice && cd visual-practice
git init

Find a simple UI you like (a pricing card, a login form) and take a screenshot. Save it as target.png. Then:

Build index.html + styles.css that reproduces @target.png as closely as
possible. Use plain HTML/CSS, no frameworks. When done, list the differences
you expect between your version and the target.

Open index.html in a browser, take a screenshot of your result, and paste it in:

Here's how it renders now. Compare it with @target.png and fix the three
biggest differences.

Repeat two or three times. Each round should get visibly closer.

Let Claude take the screenshots itself. With Claude in Chrome (claude --chrome, or /chrome in a session), Claude can open your page in your browser, screenshot it, read the console and fix what it sees. That closes the loop completely. The Desktop app has a built-in browser pane, and Module 5 shows an MCP-based alternative. For native apps, computer use lets Claude click through a real GUI.

Tips

  • Crop to what matters. A full-screen screenshot of a small bug makes Claude hunt for it.
  • Annotate with arrows or boxes if several things are wrong.
  • Say what’s correct, too. “The layout is right, only spacing and font sizes are off” stops it from changing things that already work.

Verify it worked

  • You gave Claude an image by pasting, by drag and drop, or with a path
  • You ran at least two compare-and-fix rounds against a target image
  • Your final version is recognisably close to the target

Knowledge check

3 QUESTIONS · PASS MARK 80%
1. How can you give Claude Code a screenshot in an interactive session?
2. You're implementing a UI from a mockup. What makes the result much closer to the target?
3. Which is the best use of an image in a bug report prompt?

Tip: select the text that's wrong before clicking, and it's quoted in your report.

Report a problem

Lesson: Images, screenshots and visual work

What's wrong?