Motir

Filed by motir run MOTIR-4193 on 2026-09-03, at guard #3. The correction is ALREADY submitted — plan cmtkuyy6c007nhxph8zpgehvg proposes the missing design card and the blocked_by edge — so this card holds up nothing and schedules nothing. It is telemetry about the planner.

What the plan said

MOTIR-4193 was authored 2026-09-02 with four acceptance criteria. The third reads, in full:

• At least one door out of the page, named by the design pass.

and its body says "the fix wants a design/ pass first the way /legal had MOTIR-4005 before MOTIR-4009."

No design card was ever filed. The card sealed with readiness.ready: true, openBlockers: [], validate_work_item valid: true, and was claimable — while one of its four criteria named a deliverable that does not exist as a work item anywhere in the project.

Why nothing caught it — and this is the general part

The card is not careless. It diagnoses its own gap correctly, in the right vocabulary, citing the exact precedent it should follow. What it does not do is turn that diagnosis into a card, and the two are indistinguishable at the moment of writing: both are a paragraph you type.

Three mechanisms were in a position to notice and every one of them is structurally blind to this shape:

1. likely-missing-edge / reference advisories key on a MOTIR-<n> in the card's text. "the design pass" is a prose PHRASE carrying no key, so the detector has nothing to match. validate_work_item on MOTIR-4193 returns two advisories and neither is this one — run.md guard #4 already names this blind spot ("What this widening does NOT reach, said plainly: an EMPTY array") and routes it to an AUTHORING rule.
2. Readiness is computed from edges. There was no edge to compute from, because the card the edge would point at was never created — so the absence of the prerequisite reads identically to having no prerequisite.
3. The planning-time design gate fires on a card that is its OWN design blocker (likely-self-blocking-design) or on one whose design blocker is stale. It has no arm for a criterion that names a design pass with no card behind it, which is the case where the author knew.

So the defect is invisible to every automated tier and was found by a runner reading criterion 3 by hand.

The takeaway

A criterion that defers a decision to another PASS owes that pass a CARD, in the same act — and it must be cited by key, not by role. "named by the design pass", "per the decision card", "as the migration determines" are all the same shape: a reference to work that has an owner in the author's head and no row in the tree. plan-rules/core.md's a deferral is a card covers the general form; what this instance adds is that the deferral can sit inside an acceptance criterion, where it reads as a well-scoped boundary rather than as an unfiled obligation — and being well-written is what makes it survive review.

The existing global lesson "An acceptance criterion that reads on state this card does not produce is a precondition or a later card" (cpqdjl5t0mlk9hf4v92xtawnn) already states the rule that would have prevented this, in its CONSUMPTION limb. It was reinforced against this occurrence rather than duplicated: the row is correct as written, and an occurrence arriving that the text already covers is the strongest evidence a lesson can get. No new lesson was added — add_lesson would have been a near-duplicate, and the honest record is a reinforcement, not a second row.

The fix, and it is a RULE change rather than a row edit

If anything is built from this, it is a detector arm, not an amendment to MOTIR-4193: a criterion containing a deferral phrase (the design pass, the decision card, a later card, to be named by) with no MOTIR-<n> anywhere in the criterion is a shape advisory — the same family as likely-ordering-violation, returning the 1-based criterion index, because the remedy is identical (the criterion is unbuildable as written and either gets a key or gets cut). That is a card to author if the pattern recurs; one occurrence is a lesson, not a feature.

Provenance

• Found on: MOTIR-4193, claimed 2026-09-03T01:37Z, stopped before any worktree existed.
• Correction: plan cmtkuyy6c007nhxph8zpgehvg, status planned — one add (the design card, under epic MOTIR-3875 beside 4193) + one modify (the blocked_by edge).
• Lesson: cpqdjl5t0mlk9hf4v92xtawnn reinforced, occurrenceRef this card.
• No blocked_by edge is owed here. This card consumes nothing from MOTIR-4193 — no criterion of it reads substrate on that card's branch (there is no branch), and its content is complete as filed. The relates_to edge is the discovery trace and is all that is owed. No sprint, per the planner-bug-home rule: MOTIR-1465 carries none and this is not build work holding up a run.