council: /retry sends the last brief again, to the seats that owe an answer

[sanlee-ys]

what a user can now do

A turn where four seats replied and one failed, or was cut with x, left the operator retyping the brief and the mentions to finish it. That is arithmetic at the keyboard, on the one line where getting it wrong bills seats that already answered.

/retry puts the last dispatched brief back in the composer, addressed to the seats that did not answer it:

› /retry
codex, agy did not answer turn 3 — enter re-sends that brief to them, and to no other seat

› @codex @agy the brief that half landed
                                                    → codex, agy  (2 seats) · enter dispatch

It arms; it does not dispatch. setDraft re-derives the route from the draft, and the footer prices it through the same State.SeatsIn intersection dispatch gates on (§9.21) — so the operator reads the bill before paying it, and can edit the draft first. It is an ordinary draft in the grammar that already exists, so nothing new is learned.

what counts as an answer

Defined against §9.37's four endings, and narrower than "the column looks empty":

• PhaseDone answered, including the seat whose body is [Turn completed with 0 text chunks streamed]. That is a measured zero, not a missing reply (§4a.1) — re-sending on it would overrule a vendor's honest empty answer and bill a seat that did the work.
• PhaseFailed and PhaseCancelled did not. Cancelled covers ctrl+c and the per-seat give-up together: a seat the operator cut is the case this verb exists for.
• A seat that sat the turn out is not a candidate at all. dispatch never calls startTurn on it, so its Column.TurnN still names the last turn it took. Without that, /retry after an @codex turn would widen the bill to four seats nobody addressed.

the rest of the rules

• Bare-only, on /read and /write's rule: /retry the failing test is a brief, and a verb that took it as an argument would run a re-send and discard the sentence.
• Three refusals, three sentences. A turn in flight (the phases are not settled, and the draft is kept, per postureCommand); no brief on record (turn 0, and the degenerate turn whose brief sanitized to nothing, are one fact); every seat answered.
• No second copy of the vocabulary. The verb sits in roomVerbs, so §9.31's walked refusal teaches it for free. The notice still fits the reference width — both slash-refusal goldens are regenerated and read as expected.
• --brief stays unfiled. The verb re-sends text unchanged and never reads Model.brief, so §9.17's re-briefing question is still open and still to be decided on its own.
• After a race it re-sends the race's brief as an ordinary turn: Column.Prompt holds the brief, never the /arena wrapper, and the composer shows exactly that before enter.
• The help panel does not name it, on /adopt and /arena drop's precedent — the room-controls row is at its budget.

verification

Offline, on an Intel Mac. go vet ./... clean; go test ./internal/council -timeout 20m and the rest of the packages green, both before and after merging origin/main (PR #291's per-seat quota work).

roomcmd_test.go pins the four endings producing the right seat list, the measured zero counting as an answer, the three refusals, the bare-word rule, the appearance in the walked refusal table, and — at the dispatch level with countSpawns — that the re-send spawns one process per seat owing an answer and none for a seat that replied. No test spawns a vendor.

No live run. Nothing here is a claim about vendor behaviour, and every process the verb can cause is an ordinary turn's spawn.

🤖 Generated with Claude Code