You open a run, walk its tests next to the site under test, and mark each one. Nothing ever navigates on its own — you move on when you choose to.
Click a run — or paste a run’s URL from the web app into the search box; the same box also filters the list by title as you type. (New run opens the web app’s create-run page — runs are made there and then show up in this list.)

The run opens as a checklist grouped by suite, with a progress bar and status chips to filter by.

Quick marks live right on the checklist: every row carries its own ✓ / ✗ / − buttons. For everything else — steps, a comment, the log — click the test’s title to open it.

Hotkeys, in the test view only:
| Keys | Action |
|---|---|
Cmd/Ctrl+Enter |
Passed |
Cmd/Ctrl+U |
Failed |
Cmd/Ctrl+I |
Skipped |
N |
Next still-untested test |
↓ / →, ↑ / ← |
Next / previous test |

An open run re-reads itself about every 20 seconds while you look at it, so a colleague’s marks land on their own. The server’s state wins; your own in-flight write is never overwritten by a poll.
Finish run on the run screen asks once — “Finish run? Pending tests will be marked skipped.” — and closes the run. It needs the full (not basic) mode.
A status that fails to save is not lost: it queues in the panel, the row
gets a queued badge, and a bar counts the changes pending. They replay
when the connection returns — or press Retry on the bar.