Repo · motir-meta. Found by motir run MOTIR-3715, which halted on the duplicate this limb would have prevented.
MOTIR-3230 widened validate_work_item's subsumption advisory to see OPEN pull requests, not only merged ones. It is done. Its own body then defers the other half, verbatim, under a "Not in scope" heading:
"The filing-time call — having
log-bug.mdrunvalidate_work_item(or a path lookup) before it writes a bug card — is the prose limb that consumes this one. It is deliberately left out: it is amotir-metachange and this card is pinned tomotir-core(ONE SUBTASK = ONE REPO = ONE PR). Once criterion 1 lands there is something for the protocol to call; today there is not."
The repo split was correct. The deferral named no card, which plan-rules/core.md's a deferral is a card rule forbids — and the sentence that defers it is the one that also states its own release condition. That condition is now met: criterion 1 landed, MOTIR-3230 is done, and nothing was ever filed. search_work_items for log-bug returns 36 items and none of them is this limb.
MOTIR-3713 was filed 2026-08-27T15:35:23Z for MOTIR-1135's re-consent regression on main. MOTIR-3715 was filed 17:16:29Z — 1 h 41 m later, by a second session, for the identical defect: same root cause (usersService.createUser bypasses better-auth's user.create.after), same settleOnHome timeout, same two candidate remedies, no cross-reference either way. Its run claimed it at 20:12Z and halted; MOTIR-3713's fix had merged at 20:01:02Z.
The filing pass had everything it needed and asked nothing. MOTIR-3713's card had existed for 101 minutes and its title carries this card's own subject nearly word for word ("The re-consent gate bounces every E2E signIn to /re-consent — the acceptance lane is RED on main"). This is not the hard case MOTIR-3230 was about — an open PR invisible for 84 minutes — it is the easy one: an existing CARD, findable by a one-line text search, that nothing in the protocol asks for.
Worth stating because it is why "look at the siblings" cannot substitute. Both sessions applied log-bug.md's parent rule correctly and landed in different places: MOTIR-3713 under Story MOTIR-657 (8.4 Legal — the gate's own story), MOTIR-3715 under Epic MOTIR-653 directly (the epic the finding session was working in). A correct routing rule is exactly what puts one card's duplicate out of the other's sibling set, so the search has to be over the whole project, not scoped to a parent.
motir-meta) log-bug.md's pre-file step requires a project-wide prior-art read before a bug card is written — search_work_items on the defect's own words plus validate_work_item's subsumption advisory, which now answers for OPEN pull requests as well as merged ones — and says what to do on a hit: add the evidence to the existing card as a comment, and do not file a second.motir-meta) The step states that the read is project-wide, never scoped to the parent the new bug would hang under, with the reason: two sessions applying the parenting rule correctly to one defect route it to different parents.motir-meta) It names the surfaces a tracker read cannot reach and is cheapest first — local worktrees and remote branches (git worktree list, git branch -r), then pull requests in every state — because a fix branch can exist before any card or PR does.motir-meta) MOTIR-3230's "Not in scope" paragraph is amended to cite this card by key, so the deferral is discharged where it was made and a reader of that card is not left re-deriving whether the limb exists.log-bug.md sentence changes meaning (plan-rules/plan-rules-additions-not-rewordings discipline), and the addition is one place, not mirrored into run.md.Widening the advisory further — to a sibling card at implemented on an unpushed branch — is a motir-core change and a different mechanism; it is already described on MOTIR-3230 and is not re-opened here.
motir-meta/prompts/log-bug.md — the pre-file stepmotir-meta/prompts/plan-rules/core.md — a deferral is a card