Overview
What to do next, in one screen. Ship-blocking findings ranked by confidence, coverage underneath, and the four states a finding can be stuck in.
Those four are not decoration. Each is a specific gap between “the tool found something” and “this is actually handled”, and each links to the list of what is sitting in it.
Findings
Every open finding, with its surface, severity and confidence. Two views: a flat list, and a board grouped by page. Which one you are on is remembered.
The inspector
Clicking a finding opens it beside the list: the evidence, the selector, the confidence and the frame the browser was looking at when it filed.
Review
One finding at a time, keyboard-first, nothing else on screen. This is where verdicts happen, and it exists as its own screen because giving thirty findings a verdict in a list is a different job from reading one.
Runs
Every run: the module, what it found, what it cost, and how long it took. Opening one gives you its report and its replay.
Changes
The review loop. Every surface is captured and diffed against the accepted baseline for the branch; only what actually changed is queued.
This is the screen that turns “the site looks different” into a decision somebody made on purpose, with a note attached.
Reports
The assembled documents, and where you get the shareable one. See Report.
Settings
The product profile, your model keys, and the console password if you are hosting it. Nothing here leaves the machine on a local install.