-
Notifications
You must be signed in to change notification settings - Fork 0
Description
During real-world testing with an AI coding agent (Claude), the agent specifically requested screenshots at annotation timestamps rather than continuous video or a single session screenshot.
From the agent feedback:
"Automatic screenshots at annotation timestamps are exactly right. Screenshot per annotation — frame captured at [00:15] when they drew the circle, [00:20] when they drew the arrow. Those are the moments the user is pointing at something specific."
Currently PointDev captures an element-scoped screenshot on element selection, but not on annotation events (circle, arrow). Each annotation should trigger a captureVisibleTab call and include the frame in the output.
The agent also suggested burning the annotation overlay onto the screenshot so the recipient sees what the user drew, overlaid on what they were looking at. This would make each annotation a self-contained feedback unit: screenshot + drawing + voice transcript + selector.
This came from a live validation session where an AI agent parsed PointDev output and described what would make it actionable in one shot.