A tester that drives the real drone‑ops cockpit like an operator and checks every claim on screen against the drone simulator's actual state — catching wrong, stale or unreachable UI by meaning, not by matching pixels or selectors, with near‑zero false positives.
Does each screen show what it should — and stay secure? Signed‑out drone commands, off‑screen actions, overlapping map labels, a drone shown active while offline.
Is what the screen shows actually true, over a workflow? The map showing a position 110 m stale, frozen video labelled live, movement‑trail accuracy, 16‑drone performance.
The hardest category: two operators seeing the same event, clean recovery after a connection drop, and correctness sustained over a long incident.
Argus drives real Chrome (desktop, tablet, phone; several operators at once), injects changing conditions through the cockpit's own control API, and reads what the user sees — UI text by meaning, the video's decoded frame rate, the live Cesium map's real coordinates, layout geometry.
It then compares against independent ground truth from the simulator and backend. When the two disagree in a way that harms the user, that's a bug — proven, grouped by root cause, and recorded in a video whose HUD narrates the UI's claim next to the truth.
Role, accessible name, context — small markup changes don't break the probes.
Reaches the live Cesium viewer and reads exact coordinates.
Ground truth from the control API and a socket feed, never the UI.
One root cause, one issue; correct behaviour passes cleanly.