Skip to main content

​
Find the right run

Open Reviews and check the repository and immutable source revision. Search and filters narrow the list by repository, status, UTC creation dates, completed run duration, requester, and PR author. Missing identities are Not recorded; a running review has no completed duration. Filters are reflected in the URL. The requester and PR author can be different people. A confirmed rerun is a new review, so inspect its revision and time rather than assuming the first result is current.

​
Follow the evidence

The review workspace puts the recording alongside the score, findings, and timestamped discussion. Select a finding to inspect its evidence and seek to the relevant point. Supporting tabs organize the detail: Available sections depend on the run. Missing artifacts remain unavailable; a demo recording does not replace failed capture.

​
Replay controls and comments

Choose the recorded browser and viewport before interpreting a result. Playback supports play/pause, seeking, speed changes, and timeline events. In the full viewer, Space toggles playback and Left/Right arrows seek when focus is outside text entry and other interactive controls. Select grouped timeline events to seek to their individual timestamps. A viewport preset with an actual recording selects that execution. Other presets and resize handles show Resized preview: resized historical DOM, not another browser run or device emulation. Resizing cannot establish responsive coverage. Authorized workspace users can add timestamped comments tied to a recording and target. Comments and findings can seek the replay. Comments are discussion; they do not become verified findings or change the score by themselves.

​
Findings and scores

An aberration connects observed behavior to expected behavior and evidence. Inspect its target, captured image or recording, and verification status. Inconclusive or not-reproduced candidates are not confirmed regressions. Repository configuration supports a three- or five-point score, defaulting to five. Scores reflect eligible verified findings. A score does not establish coverage of unexecuted journeys, full accessibility compliance, or field performance.

​
Coverage is separate from the score

Missing or failed requested execution and required image-analysis failures leave the result incomplete; the final score may be provisional or unavailable. Read the stated reason. A completed browser action alone does not prove image analysis ran, and an optional diagnostic marked disabled, partial, or unsupported is not clean. Performance, accessibility, SEO, link-health, video observations, and pixel comparisons are report-only. Video observations can identify moments worth inspection but remain outside verified findings and score penalties. A pixel difference establishes changed pixels, not that the change is a defect. Native Playwright results preserve their test/project identity separately from the generated journey. Their videos and traces are separate artifacts; native steps do not claim precise offsets into native video. See evidence and access for privacy and the 30-day expiry policy, or troubleshooting for an incomplete run.