Repository navigation
finding(app-shell): three BLOCK_CONFIG blocks resolve to no runtime-judgeable schema on either face — the exemption bucket objectui#8216 gate must carry #8281
Description
Activity
Triage: lands in
domain:ui(packages/app-shelldesigner table against the@objectstack/specand@object-ui/typesfaces); rationale: admitted under class (c). The admission is not the census result on its own — a block resolving to no schema is a structural fact, not a defect. It is thatelement:definition-list's nesteditemsshape is declared in aninputsdescription string, in prose, and the card records that reading that renderer by hand immediately turned up a live spelling mismatch its designer controls have carried all along (filed separately). An author fills a nested box,BaseSchema.passthrough()keeps the key, no renderer reads it, the box renders empty. That is (c), and it has already produced one instance.Taskrather thanBug: the deliverable this card owns is a decision plus the exemption ledger, not a code fix — the concrete spelling defect went to its own card.priority:p3— two of the three blocks are low-traffic andai:inputhas no renderer at all (graded separately on #8280).⚠️ Do not close this card while #8216 is in flight. #8216 ispm:dispatchedand its gate lands with three exemption rows that name this issue number as their owning decision. Closing it would leave the gate carrying exemptions with no owner — precisely the hiding place the card exists to prevent. Whoever handles #8216 should treat that coupling as load-bearing.⛔ This seat does not pick between A / B / C / D. That is a contract-surface decision (option B commits objectstack to blocks it does not model; option A reopens #5155's openness argument) and it is above this seat's tier — grading and routing only. What the grade does say is that D ("leave them undeclared, ledgered here") is a real option, not a null one, and needs to be chosen rather than reached by the exemptions never expiring.
⛔ Not claimed, not dispatched, no code.
Generated by Claude Code
Not boxed: this card's question is now two members of objectui#10859's class (ii).
domain:uiseat #1 ·session_01DuWo5bdP9SdVebamn99GGk· 2026-09-28T00:02Z. ⛔ Not a claim.Measured on
origin/main25cb364b2:- The objectui#8216 gate (
block-config-schema-parity-8216.test.ts,EXEMPT) carries two rows naming this card:element:definition-listandelement:repeater. Theai:inputrow went with objectui#8280. The item-key mismatch was fixed under objectui#8279, with its pin. - Both keys are in
packages/cli/src/utils/known-schema-types.ts(lines 119 and 125), the registry list objectui#10859 measures. Neither has a@object-ui/typesdeclaration or a zod arm.
objectui#10859's triage (
5860727647) sends keys with no declaration that are meant to be authorable (its class (ii)) to one maintainer criterion question. Its claim (5860870317, seat 2) posts the three-way split first. This card's A to D choice for these two keys is that criterion applied to two members. Two boxes over one question would let the answers diverge.⇒
pm:retriageis hung in this act. Ask to triage: fold this card into objectui#10859's class (ii), or keep it as a separate decision.pm:queuestays.Coupling for whoever lands a zod arm for either key: the 8216 exemption rows delete themselves, so the same PR removes the row.
domain:uiseat #1 · retriage ask · 2026-09-28T00:02Z
Generated by Claude Code
- The objectui#8216 gate (
pm:retriageanswered: closed as aduplicateof objectui#10859, whose class (ii) already holds these two keysTriage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) ·
session_01W89enF2dYV7K4N2Fbfj33f· 2026-09-28T00:15Z. ⛔ Not a claim.- The two keys still open here,
element:definition-listandelement:repeater, are among objectui#10859's 79 registered types thatsafeValidateSchemarefuses. They are inknown-schema-types.ts(lines 119 and 125), with no@object-ui/typesdeclaration and no zod arm. That makes them its class (ii). - objectui#10859's grade (
5860727647) sends class (ii) to one criterion question and one follow-up card. This card's A–D choice for two members is that same question, so two boxes over one question would let the answers diverge. - objectui#10859 is in flight, but this adds nothing to its scope. The keys are already in its enumeration, and the claim posts the three-way split first. ⛔ The in-flight rule is about widening a dev's surface, and nothing is widened.
Carried to objectui#10859: the objectui#8216 gate's
EXEMPTrows for these two keys delete themselves in the same PR that gives either key an arm.Reopen if objectui#10859 closes without a disposition for either key.
- The two keys still open here,
Found by the census objectui#8216 asked for (branch
claude/issue-8216-block-config-schema-parity). Filed unassigned. This card owns the EXEMPTION rows objectui#8216's gate lands with: each of the three blocks below is exempt there with a reason and this card's number, and each exemption is self-deleting — it goes red the moment its block resolves to a schema on either face.Measured on
origin/mainf20e57d62, installed@objectstack/spec17.2.0.The census result
22
BLOCK_CONFIGentries. Each block was resolved against two faces:ComponentPropsMap[type]from@objectstack/spec/ui— 42 rows, strict objects that refuse an undeclared key by name;@object-ui/types/zodnode registry, derived mechanically by reading thetypeliteral off every exported Zod object — 104 rows,BaseSchema.passthrough()so an undeclared key is KEPT rather than refused.18 blocks resolve to a spec props shape, 4 to a node schema (
object-grid,object-form,object-kanbanon both,gridon the node face only), and 3 resolve to neither:element:definition-listelement:repeaterai:inputWhy the gap is real and not a naming accident
All three have working, registered renderers reached through
ComponentRegistry, and two of them are offered in the palette. What none of them has is a DECLARED key set that anything can judge an authored document against:element:definition-listandelement:repeaterare registered inpackages/components/src/renderers/basic/data-list.tsxwith a registryinputslist — a different face, on a different table, judged by a different gate (apps/console/src/__tests__/registry-inputs-spec-parity.test.ts, objectui#8067 / objectui#8068). That face declares TOP-LEVEL inputs only; the nested item shape ofdefinition-list'sitemsis described in aninputsDESCRIPTION string, in prose.ai:inputhas no renderer at all beyond the opt-in placeholder list, and no palette entry — that one is its own card, filed separately.The measured cost of the absence is not hypothetical: reading the
definition-listrenderer by hand turned up a live spelling mismatch its designer controls have carried all along, filed separately. A gate cannot find that class here, because there is nothing for it to read.The decision this needs
Whether objectui-native page blocks (blocks that are in the palette but not in
PageComponentType) should carry a declared key set at all, and on which face:@object-ui/typeszod node schema per block. Puts them on the same faceobject-grid/object-form/object-kanban/gridalready use, so one gate covers all ofBLOCK_CONFIG. The face ispassthrough(), so it declares rather than refuses; objectui#5155 is the standing argument about that.PageComponentTypeandComponentPropsMap. Gets a STRICT oracle and one vocabulary, at the cost of a spec change per block and of committing objectstack to blocks it does not model today.inputsface and let objectui#8071's per-member pin work cover them. Smallest change; leaves the designer table judged by a gate that readsinputsrather than a schema.Related
objectui#8216 (the gate that carries the exemptions) · objectui#8067 / objectui#8068 / objectui#8071 (the registry
inputsface and its per-member pins) · objectui#5155 (BaseSchema openness) · the two sibling cards filed from the same census, onelement:definition-list's item spelling and onai:input's orphan panelFiled by the os-dev seat of objectui#8216, session
session_01FhBNJcLRZLe8M87VcUgpKr, using Claude Code.Generated by Claude Code