Screenshots in questions

Some questions live on the screen rather than in the conversation: a table someone is sharing, a diagram, an error message. Capture window (⌃⌥S, Ctrl+Alt+S on Windows) attaches a picture of the active window to your next question.

How it works

1. Bring the window you want into focus — the shared screen, the document, the dashboard.
2. Press ⌃⌥S — Ctrl+Alt+S on Windows. jhint takes a picture of that window only, not the whole desktop.
3. The attachment appears in the conversation window; press Answer to send it with your question.

Nothing is captured automatically, and nothing is captured while you don’t ask. Each screenshot is taken by you, at the moment you press the shortcut.

Which models read them

Not every model can look at a picture. The answer panel says so: text means the model reads only what was said, text + images means it reads screenshots too.

• Built-in — Qwen3 VL 4B reads screenshots; the other two don’t.
• Your provider — most current Claude, GPT and Gemini models do. jhint knows which and turns the button off for the ones that don’t, instead of sending an attachment that would be silently ignored.

Before it leaves your computer

With an external provider, a screenshot is a piece of your screen going to another company. jhint shows you the picture that is about to be sent, and sends nothing until you press Answer. With the built-in model the picture never leaves the machine at all.