Teaching Coding Agents to Tell the Visual Story of a UI Change
Agents can already drive a UI change in a real browser and screenshot every state, but the frames died in /tmp and the PR just said "verified". A rule, an agent skill and a small hosting service turn that browser pass into a visual story on every UI pull request: before, after and a GIF. Here's how it fits together, with setup in three commands.
Try Capture (opens in a new tab)