Record card. Type chore · executor coding_agent · Repo: none — the fix is a plan change, so targetRepo is deliberately unpinned. The correction is submitted as plan cmtdl09xr003jhvphju1lwjoy and awaits approval; the re-parenting half of it was applied directly (see The correction below).
Found by: motir re-plan MOTIR-1043 "this should be implemented in motir.co" (Yue, 2026-08-28).
Two defects on one card, and the second is downstream of the first.
MOTIR-1043 was authored 2026-06-17 and re-scoped 2026-06-22 as "a PUBLIC (no-auth) page… a marketing / credibility surface" — correctly, because at that date Motir's public front door was motir-core's app/page.tsx.
On 2026-06-30, the 8.3 re-plan (MOTIR-656) moved the public marketing surface out of motir-core into a new motir-marketing repository at motir.co. That re-plan's own banner records the move in full. It never reached MOTIR-1043, and MOTIR-1043 had no edge of any kind to 8.3 — it sat two stories away, under MOTIR-1084 (8.8, launch-readiness polish), in the same epic.
For 59 days the card therefore read: status: todo · ready: true · openBlockers: [] · valid: true, with (motir-core) in its own title, pointing at a repository that had stopped owning the thing the card is for.
MOTIR-1043 is a UI card. Its five blockers were MOTIR-1040, MOTIR-1045, MOTIR-1061, MOTIR-1065, MOTIR-1070 — three axis foundations and two onboarding designs. None of them is a design asset for this surface, and no design/<area>/ asset draws it in either repository. Gate 5 requires BOTH the asset and the type: design subtask, or a STOP. It was never run, in the original authoring or the June re-scope.
This is not a coincidence beside defect 1: a card whose home repository is wrong is a card nobody re-opened, and the design gate is one of the checks a re-open would have run.
Every mechanism that acts on the plan was correct. Readiness answers "is anything blocking this?", never "is this still wanted, and in the place it says?". validate_work_item reasons over the graph, and there was no edge to be wrong about. The reference advisory family scans a card's own description for MOTIR-<n> keys — and a card written in June cannot name a decision made in July, so the reference runs the wrong way for the detector to see it. The likely-repo-straddle advisory reads criteria for paths in a repository other than the card's pin, and this card's criteria named only motir-core paths, which is exactly what a correctly-pinned motir-core card looks like.
So the card got quieter as it rotted, which is the general shape: a card falsified from OUTSIDE, by someone else's decision, has no field that changes.
prompts/plan-rules/op-replan.md, the redefines-a-mechanism / allocation sweep corollary, already carries this case in three of its limbs, verbatim:
(motir-core).So this is a DILIGENCE miss, not a totality gap: no rule change is owed. Measured before concluding — git grep over prompts/plan-rules/*.md on origin/main returns limbs (a)–(e) of that corollary intact, and core.md gate 8 indexes it.
search_lessons (takeaway register, kinds: [subtask], phases: [skeleton]) returned "A re-plan sweeps its dependents and sibling test cards" (cmr11s2bw000tkmqz0l24gs3w, global) at distance 0.4277 — inside the ≤ 0.50 real-match band. Reinforced against MOTIR-1043; recurrenceCount is now 2.MOTIR-1465 card is by construction "you did this and it was wrong", so the default is a lesson; the incident-strip test leaves nothing to reach for that is not already prescribed one limb away.op-replan.md limb (c) is where the wording already exists to borrow.bug cards under MOTIR-1465 against the referrer-sweep shape do not reach three — MOTIR-3723 is duplicate-planning-without-an-edge and MOTIR-3746 is an orphaned deferral. Different causes; no re-shape card owed.Applied directly (2026-08-28): MOTIR-1043 was re-parented from MOTIR-1084 (8.3 → 8.8 polish) to MOTIR-656, the marketing-site story, where the rest of the motir.co work lives, and the two new proposals were re-pointed there with it. move_to_parent has no proposal form, so this half could not ride the plan. The move is what makes the whole edge set legal at one level: MOTIR-1043 blocked_by MOTIR-1152 is now a same-parent sibling edge rather than the cross-parent one the plan originally had to justify, and validate_work_item on MOTIR-656 (projected over the plan) returns valid: true. All three cards were re-numbered 8.8.x → 8.3.x to match their new home.
Proposed, awaiting approval (plan cmtdl09xr003jhvphju1lwjoy):
motir-core → motir-marketing, re-routed /tokens → /design, re-based on the published @motir/design-system, re-sized 3 pt / 45 min → 5 pt / 55 min; blocked_by gains the two new cards and MOTIR-1152, and drops MOTIR-1061 as no longer a prerequisite.type: design card for the surface (the gate-5 repair), and 8.3.15, a type: copy card for its words — mirroring 8.3's own MOTIR-1143 / MOTIR-1144 / MOTIR-1152 shape.⚠️ The plan's SUMMARY is stale on one point and cannot be edited — there is no tool to patch a plan's summary once it is written. It still says the three cards sit under MOTIR-1084 and asks the approver to move them after approving. They are already moved; the instruction is discharged, not pending. The proposals themselves are correct.
reinforce_lesson → counted: true, recurrenceCount: 2.move_to_parent response (parentId: cmqfb4m6l00jr2d0iax6ktrc3).cmtdl09xr003jhvphju1lwjoy was approved and that MOTIR-1043 now reads targetRepo: motir-marketing with a type: design card among its blockers. If the plan was declined, say so here and leave the card open — the scope half of the defect is unrepaired either way, even though the placement half has landed.todo → in_progress → in_review → done) once the check above passes.motir-meta prompts/plan-rules/op-replan.md — the corollary's limbs (b) and (c), which name this exact shape · prompts/plan-rules/core.md gate 5 (the planning-time design gate) and gate 8 (referrers swept)cmr11s2bw000tkmqz0l24gs3w — "A re-plan sweeps its dependents and sibling test cards", global, reinforced here