Planning bug: a bug card counted "the other three inks clear it" over the set its OWN diagnosis had just shown to be unmeasured — the enumeration inherited the guard's blind spot it was filed about
Filed by the run of MOTIR-3954, per run.md's a falsified COUNT is REPORTED, never silently re-scoped. The card's verdict was right and its enumeration was short; nothing was re-scoped, and the amendment is on the card.
What the card said
"The other three status inks clear it;
implementeddoes not, on any of the three."
What the measurement says
All eight base status inks × three surfaces × both themes, resolved through tests/theme/paletteCascade.ts off packages/design-system/theme.css at origin/main 0068da2f4 — six clear the 3:1 bar and three (palette × theme × token) steps do not:
| motir/light | resolved | card / surface / page-bg |
|---|---|---|
--el-status-implemented | #ff64c8 | 2.66 / 2.44 / 2.66 ✗ — the card's subject |
--el-status-todo | #a4a097 | 2.61 / 2.39 / 2.61 ✗ |
--el-status-done | #1aae39 | 2.93 / 2.69 / 2.93 ✗ — filed as MOTIR-3991 |
| the other five | — | 3.29 – 6.57 ✓ |
(--el-status-todo is also sub-bar in dark, 2.82 / 2.56 / 2.82.) The same card also asks for "the four status inks" to be measured for the base palette; the ramp has eight, and the shipped widening measures all of them.
The root cause, and why it is worth recording
The card's enumeration was produced by the very mechanism whose blind spot the card is about. Its diagnosis is that the guard's population — "every step a [data-palette] block declares" — silently excludes the base palette. To count "the other three", the run measured the tokens that block declares, which is the override-shaped set, not the ramp. So the count is a correct measurement of the wrong set: the exact re-measure the PREDICATE, not only the REF shape in run.md, one level in.
It is a sharper instance than the rule's own fixture. There the command was scoped by how the defect was noticed; here it was scoped by the defect itself — the card reproduced the blind spot inside the sentence describing it. That is the failure mode the diligent path walks into: having correctly identified that a population is being derived wrongly, the natural next step is to enumerate, and the enumeration reaches for the same derivation.
Cost, had the run followed the card literally: the widening would have measured four inks, --el-status-done would have stayed unmeasured, and the card would have closed green with the guard still exempting the base palette for half its ramp — the defect this card exists to remove, surviving its own fix. The run measured the whole ramp instead and filed the residue as MOTIR-3991.
Acceptance criteria
motir-metaprompts/plan-rules/phase-deepen.md's ENUMERATION limb gains a limb for this case: when a card's diagnosis is "population P is derived by predicate Q, and Q is wrong", any COUNT in that card must be taken by a predicate other than Q, and the card states which. Written as a standing check, not as a retelling of this incident.- The limb names the tell in the words a card actually uses — "the other N", "all N", "every remaining" — sitting in a card whose own body argues that N was computed wrongly.
- A corresponding lesson is added to the store, scope decided by the strip-the-identifiers test and stated in the PR body.
- The rule change cites no work-item key in the rule text itself (the corpus convention); the incident goes in the PR body.
Boundary
motir-meta only — a rule addition plus the lesson row. It changes no product code and re-opens nothing in MOTIR-3954, whose fix already measures the whole ramp.