Key moment

Updated: July 2026

A key moment is a timestamped instant in an agent-readable recording, carrying an extracted frame, a click target, and a plain-language description of what happened.

Key moments are an ordered list inside the markdown twin. Each one anchors a moment to the timeline (timestamp), points to the still image at that instant (frame), records the UI element and coordinates interacted with (click_target), and narrates the intent (description). Together they let an agent step through a flow without decoding video.

For example, in the Export CSV recording one key moment is the click on the Export CSV button at its coordinates, the next is the button entering a spinner state and staying there, and a third is the DevTools console printing the TypeError from exportWorker.ts:142. The agent reads the sequence and reconstructs the failure.

Every key moment gives an agent the frame and the click, so it learns not just that something happened but what was clicked and where.

Related terms: click target, markdown twin, AREC (Agent Recording).

Learn more: an annotated walkthrough of the markdown twin. See the AREC spec.