
You can spend five minutes describing what is wrong with a screen, or you can show the agent the screenshot. Showing is faster, and it is exact - there is no gap between what you saw and what you managed to put into words. "The button is behind the header on narrow screens" plus the picture lands. The paragraph you would have written instead usually does not.
There are several ways to do it - from your desk, from a browser, from your phone - and they behave differently enough to trip you up. Underneath, they are all the same move, and once you see it they become one habit.
The one idea: an agent reads a picture as a file
A command-line coding agent reads a picture the same way it reads your code: as a file on the machine it is running on. So showing it a screenshot is two jobs. Get the image file onto that machine, and put its path in your prompt. That is all any route below is doing.
At your desk: drag it onto the prompt bar
In DevThrottle, the prompt bar is the box under the terminal you type into. Drag an image from File Explorer onto it and its full path is inserted where your cursor was. Nothing is uploaded and no copy is made - the agent is on this machine, so it opens the picture where it already lies. Write your sentence around the path and send.
- One picture per drag. Grab several files at once and only the first one lands.
- Check it is the right picture. Screenshots taken a minute apart have names that look identical. Expand, at the top right of the prompt bar, opens a larger editor in which a path to an image is shown as the image itself. Sending the wrong screenshot is not an error anybody catches - the agent answers confidently about a picture you did not mean to show it.
- Skip File Explorer entirely. The Screenshots panel on the right of the Director lists the newest pictures in one folder; drag a thumbnail onto the prompt bar for the same result. Point it at your own screenshots folder first, under Settings - Directories, where Detect fills in the folder Windows saves screenshots to.
- Paste does not attach here. Pressing Ctrl+V with a picture on the clipboard does not attach it in the Director's prompt bar. Drag the file instead. Most people arrive with the paste reflex, so it is worth knowing before you try it.
In a browser: paste works
Open the same session in the Cockpit and its message box takes a picture three ways: drag one on, paste one in, or press Attach. All three upload the picture from the device you are sitting at to the machine the session runs on, save it there, and drop the saved path into the box, with Image attached to confirm it. Here Ctrl+V is the one to make a habit of. Pasting ordinary text behaves exactly as it always did.
From your phone: take the photo there and then
Standing in front of the screen that is wrong, or a whiteboard someone just filled? Open the session on your phone and press Attach; the phone's own picker offers the camera and your photo library. The photo is uploaded to the machine the session runs on and the path where it landed arrives at your cursor. A photo sent from a train becomes a file on the machine at home, and the agent is told where.
One snag: a phone that photographs in .heic sends a file that is accepted, saved and handed over like any other, but that no DevThrottle list will show you. Whether the agent can open a .heic is its own affair. If it cannot, set the phone to photograph in JPEG.
Pictures coming back
Agents make pictures too - a chart, a rendered page, a screenshot of your app. Do not go hunting for the file. Ask the agent to print the path; nothing obliges an agent to say where it wrote something, and they often do not. Then click the path: the Director offers View File, and in the browser it opens in place.
Handing over the path is where DevThrottle's job ends, and it ends in the same place for every agent it runs. Whether the agent can look at an image is that agent's own ability, so if one ignores your picture, say what you want it to notice as well as showing it.
The step-by-step version is Work with pictures. Which file types each surface recognises, and every button on the Screenshots panel, are on Working with pictures.
Run your agents from one control room
DevThrottle orchestrates command-line coding agents across your machines.
Create account