Bulgar Studio

A gate could not run

test NOT_RUN: needs a local DATABASE_URL the agent container is not given. Not a product failure. Gates were re-run by the supervisor after agent-run.sh died early (see LESSON-editing-running-scripts).

I could not run the app.

I could not run the app. No Docker, no local Supabase, no .env and no .env.development.local in this container. Every claim below about *rendering* is reasoned from the code and the CSS, not observed in a browser. The visual approval this batch stops for is exactly what I could not do myself.

The test suite cannot go green here

The test suite cannot go green here — see §7. This is the environmental failure CLAUDE.md already records, made total by the absence of any local database.

.cmd-scroll-x:has(...)

.cmd-scroll-x:has(...) — :has() is used once, to switch off horizontal scrolling when the table has become a stack. Supported in every current browser and in the Tauri/Electron shells; if it ever is not, the fallback is a scroller with nothing to scroll. Harmless, but worth knowing it is there.

Home now makes two reads

Home now makes two reads (/api/admin/launcher-summary and /api/admin/action-queue). SG's Home already makes the second, so this is not new load for the product, but it is new load for Command's Home. If that matters, the honest fix is to serve the three launcher counts from the action-queue payload and drop the first read — a loader change, not a component change, and not something I should decide alone.

Open question for Ivan / Kris:

Open question for Ivan / Kris: the waiting lane groups new_intake (an unanswered enquiry), unassigned (nobody has picked it up) and waiting_review (nobody has looked). All three are "waiting on somebody at BULGAR", which is why they are together — but if the studio thinks of new enquiries as their own thing, that is a one-line change in WORK_LANE_CATEGORIES and the test that says every category has exactly one lane will keep it honest.

Still genuinely wrong, and NOT fixed here