A room board reached from a Bot board counts its active cards twice in the rail #2069

Implementation: FlyBot (ours) — PR follows within 24h. Do not pick up without a pickup-ok comment on this issue.

Written-by: FlyBot Lead
Claimed-by: FlyBot Lead
Claim-Id: e4910b8f53384edcbdfb47a9540eefb9
Branch: dev/2-room-rail-double-count
Base: 5b862c6
At: 2026-11-01T17:43:55Z

Implementation: FlyBot (ours) — PR follows within 24h. Do not pick up without a pickup-ok comment.
Written-by: FlyBot Lead

The defect

A room board opened from a Bot board counts the room's active cards twice in its rail. The rail is supposed to read {done}/{total} for the cards the board actually draws; on a room board that is the room's active list plus one archive page, but the hook's own state is never reset when the scope changes from a Bot to a room, and the rail adds both lists.

Code on main 5b862c6e (core/src/client/kanban.js):

• :167 — const [own, setOwn] = useState([]); the effect is keyed on key (:169), which changes when the scope changes.
• :180-181 — the non-room read is actions?.tasks?.() with no argument; the host's default read is every task the Bot may act on (core/src/host/tasks.js:1397-1399, all.filter(mayAct || ancestorOwned)), so room cards land in own as well as in groupTasks.
• :186-189 — setOwn(list) only runs in the scope.kind !== 'group' branch; going Bot → room leaves own holding the Bot board's read.
• :401 — [...own, ...(scope?.kind === 'group' ? groupTasks : []), ...(archive ?? []).slice(0, ARCHIVE_SHOWN)] therefore counts the room's active cards twice.

Reproduction

Product's probe (same harness as core/test/kanban.test.js): a room war with one active Writer card and one done card in the archive page. First read on the real path is the Bot board (the rail's first entry, :377), which fills own with [task-open-s]; then the scope switches to the room:

bot board (default scope)      rail Writer = "0/1"   (own holds: task-open-s)
room board after the rail click rail Writer = "1/3"  drawn: not-started=1 / in-progress=0 / done=1  → true value "1/2"
control own = []               rail Writer = "1/2"

A card from another room left in own reaches a Bot rail the board does not draw at all (rail Chief = "0/1" with no Chief card drawn), which is the same bug from the other side. Read arguments confirm the split: bot board read args = [], archive read args = [{botId:'a', archive:true, limit:20}].

Expected behaviour

The rail counts what the board draws, and only that: on a room board groupTasks plus the archive page; on a Bot board own (already filtered to that Bot) plus the archive page. Minimal fix, one line plus a comment:

[...(scope?.kind === 'group' ? groupTasks : own), ...(archive ?? []).slice(0, ARCHIVE_SHOWN)]

#2045's one-page cap stays as it is.

Red first — and the fixtures that must change with it

Base: live main 5b862c6e5ecdd0a472826162ebb2abab63b91544.

• New case: render the Bot board first (so own is filled by a real read), then switch the scope to the room; railMeta('s') must be 1/2 where it is 1/3 today.
• The existing room cases are false green and must be fixed in the same PR. boardWith({ kind: 'group', … }, [], archive, groupActive) presets own = [], which the real path never produces: on a real board the initial scope is rail.bots[0] (:377), so own is always filled before any room is opened. A test that starts a room board with an empty own cannot see this bug. Seed own from the Bot board read, as the new case does.
• Everything already asserted stays asserted, verbatim: 1/2, 2/3, 0/1, 1/4, 3/5, 19/21, 20+ done, 18+ done, 2+ done, 0/20 (core/test/kanban.test.js:344-431).

Scope

• core/src/client/kanban.js — the one line and its comment (P-364's comment above it needs the room branch stated correctly).
• core/test/kanban.test.js — the new case and the fixture correction.
• The built client and patches/ds-bot/ (0008 diff, files/src/client/…, files/test/…) as generate writes them; build-client --check must be 0.
• docs/subsystems/core-client.md — the P-364 paragraph gains one sentence: a room board counts groupTasks plus the archive page, never own.
• changelog.d/<this issue>-room-rail-double-count.fixed.md.
• T1 / XS.

Gates

TZ=UTC ARCH_DOCS_BASE=5b862c6e npm run verify, ds-bot-patches check + canonical, build-client --check, checkpointReasons = [], maintainer_can_modify: true. FlyBot Test and FlyBot Security self-tests run on the new tip before the reviewer is named; a Verdict request on a tip whose advisory is still running is how a half-checked tip reaches the merge queue.

Dedupe on live main: #2048 and #2063 (both closed, the P-364 halves) and #2045/#2058 (page-full lower bound) are the neighbourhood; no open issue covers this double count.