feat(plugin-kanban,plugin-calendar): declare the seven scalar spec keys both renderers already honour (#8201, slice 1 of 2) - #8309

Part of #8201 — this is slice 1 of 2, so it deliberately does not close the card. See the slice boundary below for what stays open and why.

What this does

Declares seven ComponentPropsMap keys that the two object-* renderers already read and already act on, and deletes the seven matching UNPUBLISHED_EXEMPTIONS entries in the same change.

Both tags of each block (object-kanban/view:kanban, object-calendar/view:calendar) now spread one shared list rather than two hand-copies. The two tags are the same renderer, so a hand-copy was the only way they could ever disagree.

No renderer behaviour changes. Every one of these keys was already read; what changes is that the html tier stops reporting working metadata as unknown-prop (objectui#6678's shape) and authoring tools can discover them.

Clause ② — this widens the authoring surface, and here are the grounds

Declaring an input widens what an author may write, so the grounds are stated rather than assumed. They are the same grounds #8186 (filter) and #8223 (sort) cleared on: the spec already declares the key and the renderer already honours it, so this restores declared = enforced instead of publishing anything new.

That claim was measured, not assumed, because #8172 is the counter-example: limit is taught by four faces and refused by name by this same strict ComponentPropsMap. Every one of the fifteen keys on the card was probed with a control key on the same safeParse call:

CONTROL __objectui_control_probe_8201__: success=false unrecognized_keys=true
  groupBy          refusedByName=false   arms={"string":["accepts"], ...}
  defaultView      refusedByName=false   arms={"string":["refuses-content"], ...}
  ... (all 15 refusedByName=false)

The control draws unrecognized_keys on the same call shape while none of the fifteen does — so "the spec declares it" is a verdict here, not a vacuous read. The probe is preserved as row 4 of each new pin test.

⚠️ The slice boundary, argued rather than silent

#8212 landed a tightened direction: every array/object-armed input must pin its member shape, so declaring such a key owes three things — the inputs entry, the exemption deletion, and a MEMBER_PINS entry whose shape must be measured at the sink rather than assumed.

The boundary drawn here is the declared arm, and it was chosen because it is mechanical rather than tasteful: structuredInputs in the parity gate is keyed off the declared arm, a scalar arm never enters that population, so these seven owe no MEMBER_PINS entry and MEMBER_PIN_EXEMPTION_CEILING stays at 62. A reviewer can check the boundary instead of agreeing with it.

Eight keys stay behind, and the card stays open:

• Array/object-armed — slice 2 (object-kanban.data, .cardFields, .grouping, .conditionalFormatting; object-calendar.data, .staticData). Each needs a member pin measured at its own sink. Note that feat(plugin-calendar): declare the sort input both object-calendar registrations read #8223's lesson applies here and is not yet resolved for any of them: the two filter pins assert identity (toBe), sort's does not, because convertSortToQueryParams builds a new map. Assuming the identity shape is the error that had to be corrected on main once already.
• object-calendar.loading — scalar, but held back for a different reason: it is inert alone. ObjectCalendar.tsx applies externalLoading only under hasExternalData, so an authored loading does nothing unless data is authored too. It ships with data, in the slice that measures data's member shape.
• object-kanban.quickAdd — ESCALATED, not deferred. Measured: this renderer does not honour it. KanbanImpl gates the control on quickAdd && onQuickAdd (two sites), onQuickAdd is an finding(types): 28 zod-mirror keys are declared z.function(), which NO JSON document can satisfy — 14 in-repo corpus files author them and are silently invalid #6124 runtime slot the zod twin refuses by name, and nothing on the ObjectKanban path supplies one. That measurement is already tracked as plugin-kanban: quickAdd is inert on object-kanban / kanban — ObjectKanban never supplies the onQuickAdd runtime slot the control is gated on #8285. What remains is a product ruling — permanent carve-out, or a feature gap whose fix makes the key declarable — so its ledger entry is left exactly as it was rather than rewritten into a carve-out reason this PR has no standing to write.

Correction to the dispatch's expectation: conditionalFormatting was flagged alongside quickAdd as a likely product ruling. It is not — KanbanImpl reaches it through getCardStyles(card, conditionalFormatting, …) at every card render, so it is a plain declare-case. It is deferred only because it is array-armed and owes a member pin.

Ledger

The backlog ceiling moves 15 → 8, and both assertions carrying it move together (toBeLessThanOrEqual and toBe, same it()); moving only toBe would leave the other banking headroom the ratchet exists to refuse. Proven below.

Corrected in passing: the prose above that ceiling read "Sixteen … ten on object-kanban, six on object-calendar" while both assertions already read 15 — #8223 moved the numbers and not the sentence. A stale count there is what makes a reader distrust the assertions instead of the prose.

Evidence

Parity gate assertion count: 198 before, 198 after — no move. Measured by reverting the three tracked files to origin/main on this worktree, running, and restoring by state. No move is expected: the gate's rows are static it()s, and the change moves counts inside arrays those rows read, not the row set.

Every new pin proven able to fail — red BY NAME, per registration, restored BY STATE.

1. Ablate groupBy from the kanban list and defaultView from the calendar list (one line each, on-disk deletion verified by grep -cF going 1 → 0 and git diff --stat showing the deletions):

   × 'object-kanban'   — the html tier accepts an authored `'groupBy'`
   × 'view:kanban'     — the html tier accepts an authored `'groupBy'`
   × 'object-kanban'   — the registration declares `'groupBy'`
   × 'view:kanban'     — the registration declares `'groupBy'`
   × 'object-calendar' — the html tier accepts an authored `'defaultView'`
   × 'view:calendar'   — the html tier accepts an authored `'defaultView'`
   × 'object-calendar' — the registration declares `'defaultView'`
   × 'view:calendar'   — the registration declares `'defaultView'`
   × 'object-calendar' — the declared `defaultView` members are DERIVED from the spec …
   × 'view:calendar'   — the declared `defaultView` members are DERIVED from the spec …
   × both tags publish ONE shared list, so a hand-copy cannot drift   (×2 files)
   Tests  12 failed | 28 passed (40)
   

   Per-registration redness, not per-file: each tag reddens on its own named row, and 28 rows stayed green.

2. Ablate the ledger — re-add one deleted exemption while the declaration stands:

   × carries no stale unpublished-key exemption — a published key must lose its entry
       AssertionError: expected [ 'object-kanban.groupBy' ] to deeply equal []
   × the objectui#8176 backlog has a ceiling — it may shrink, never grow
       … the backlog list is shrink-only: expected 9 to be less than or equal to 8
   

   The second row is the proof that both halves of the ceiling moved: it is the toBeLessThanOrEqual assertion firing.

3. Ablate the honour claim — neutralise the coverImage mapper in bucketCardsIntoColumns:

   × honoured — `coverImageField` becomes the card cover `KanbanImpl` renders
   Tests  1 failed | 25 passed (26)
   

   One row, precisely.

Every restore was verified by state, never by an exit code: git diff HEAD empty, and git hash-object <path> equal to git rev-parse HEAD:<path> for each mutated file.

Suites and gates (all from the repo root — pnpm --filter PKG exec vitest run FILE is refused here by #3378's guard):

Repo-wide pnpm lint and the full suite are left to CI.

Found in passing, filed not fixed

#8308 — ObjectKanban.tsx resolves the cardTitle / titleField pair with || at :403 (card list) and ?? at :1077 (detail drawer). An authored cardTitle: '' therefore titles the cards from titleField while the drawer heading falls through to the record floor: one document, two answers. Not fixed here — this PR declares the two spellings without changing how either resolves — but declaring them makes the divergence more reachable, which is why it is filed.

Reviewer note

⛔ Draft on purpose. The PM flips and lands; auto-merge is not enabled.

🤖 Generated with Claude Code

https://claude.ai/code/session_01YBWFb5YgMU5dw8p2VKj16S

Generated by Claude Code