step-timeline

An optional, framework-neutral sidecar that maps a moment in a video to a named step, so an annotation can land on a step. SpexCode owns the tiny format; any emitter writes it.

For an annotation to land on a step ("at the login step the spinner hung"), a video reading (video-evidence) needs a map from video-time to step. step-timeline is that map: a companion content-addressed blob — an ordered list of {tMs, step} on the clip's own clock, filed beside the clip (spex yatsu eval --video … --timeline <json>, validated LOUD, the reading carrying only its hash). The step at a moment T is the last event at or before T; that lookup is the whole of "which step is this".

The seam SpexCode owns is the format, never a test framework. A small userland emitter (start / mark a step / flush) stamps each step against the recorder's clock while it records; a Playwright reporter, a WebDriver listener, or a computer-use hand narrating each action are all just emitters of the one format — so scripted and ad-hoc measurement share a single contract and no framework is privileged. Aligning the emitter's clock to the clip is the emitter's own job.

Its consumer has arrived: the annotator's step ruler reads this map to seek and to name the step an annotation lands on. A step may carry an optional owning-node id, so a finding at that moment routes to the node actually responsible — not necessarily the node under measure. Deliberately out of scope: keyframe/highlight markers beyond a plain step, any on-video caption rendering, any authoring UI, and any lifecycle beyond writing the map.