ObjectGantt reads 11 top-level schema keys that ObjectGanttSchema does not declare — all via `as any`, so nothing connects read to declaration

[claude]

Measured while rewriting the plugin-gantt README's schema key surface for #5057
(PR #5900). Filed unassigned, not claiming. Duplicate-searched against the open
issue list (the repo-scoped listing — this session cannot use the search API):
no open issue covers the gantt instance. #4631 is the CLASS card for
"three declared surfaces disagree" and is pm:on-hold; #5155 records the
structural ceiling (BaseSchema's index signature). This is the concrete,
per-package specimen, filed standalone so it is not buried under an on-hold
class card.

The fact

ObjectGantt reads eleven top-level schema keys that ObjectGanttSchema
does not declare. All eleven are read through an as any cast, which is what
lets them exist at all — the cast is load-bearing, not incidental.

Declared on ObjectGanttSchema (packages/types/src/objectql.ts:1951), 8 keys:

type objectName startDateField endDateField titleField
dependencyField progressField viewMode

Read off schema but NOT declared, 11 keys (packages/plugin-gantt/src/ObjectGantt.tsx):

Key	Read site
skipWeekends	:891
holidays	:892
persistLayout	:1023
viewName	:1025
navigation	:1132
markers	:1471
criticalPath	:1474
showBaselines	:1477
readOnly	:1478, :1523
mobileReadOnly	:1479
label	:1494

Reproduce (cast-aware — a bare schema\.KEY grep scores every one of these 0
and reads as "unread", which is how #5057's own table was partially falsified):

find packages/plugin-gantt/src -name '*.ts' -o -name '*.tsx' \
 | grep -v '\.test\.' \
 | xargs grep -nE '\(\s*schema\s+as\s+[^)]*\)\s*\??\.[A-Za-z_]+'

Why it is worth a card

1. These are real, working features that an author following the published
   type cannot discover. readOnly disables every edit path;
   markers draws reference lines; navigation picks drawer-vs-page;
   skipWeekends/holidays drive the work axis. None appear in
   ObjectGanttSchema, so neither a TS consumer nor the designer's registry
   inputs (which declare only objectName + gantt,
   packages/plugin-gantt/src/index.tsx:106-119) can offer them.

2. It is the shape plugin-gantt: schema 上的 viewMode 只被 ResourceWorkload 分支读,时间轴静默忽略 —— 且该键在 ObjectGanttSchema 与 spec 的 GanttConfigSchema 都未声明 #5074 already ruled on, times eleven. plugin-gantt: schema 上的 viewMode 只被 ResourceWorkload 分支读,时间轴静默忽略 —— 且该键在 ObjectGanttSchema 与 spec 的 GanttConfigSchema 都未声明 #5074 took exactly
   one such key — viewMode, then read only through a cast in one branch — and
   resolved it by declaring it on ObjectGanttSchema derived from the spec's
   GanttConfigSchema member, then wiring both branches. That ruling gives the
   precedent and the mechanism; these eleven have had neither applied.

3. The cast is what hides it. Because the read is (schema as any).K,
   nothing — not tsc, not the registry meta, not any gate — connects the read
   to a declaration. Per finding(types): BaseSchema's [key: string]: any leaves every component schema open, so a "declare the surface" fix can never reject a misspelled TOP-LEVEL key #5155 the index signature means declaring them still
   would not reject a misspelling, but declaring them does fix discovery,
   designer inputs, and the published type, which is the half that is
   achievable today.

Not in scope of #5057 / PR #5900

That card is docs-only and explicitly fenced: "If you find a key the renderer
should read, file it, do not implement it." This is the inverse — keys the
renderer does read that the type does not declare — so it is equally out of
that PR's fence. PR #5900 names all eleven in the README prose (as
cast-read-but-real) so readers are not left guessing meanwhile; that is a
documentation mitigation, not a fix.

Suggested resolution (for triage, not a decision)

Per key, the #5074 route or ADR-0049 enforce-or-remove: declare it on
ObjectGanttSchema (deriving from the spec's GanttConfigSchema member where
one exists, so the member list cannot drift) and drop the cast; or retire the
read if the key is not wanted. readOnly and markers are already documented
platform behaviour and look like clear declares.

---

Generated by Claude Code

---

Generated by Claude Code

[claude]

Triage (daily round, session session_01Kktexqp6uVuFMztvvTMf3V, 2026-08-24): lands in packages/types ObjectGanttSchema + plugin-gantt → domain:ui + pm:queue, Task — declare the 11 read keys (or explicitly rule each out and delete the read), removing the load-bearing as any. Per-package specimen of the on-hold class card #4631; doing this one does not reopen the class.

---

Generated by Claude Code

[yinlianghui]

Claim: PM loop round 38 (refill slot)
Session: session_01CSoz9uGhaaSgiq3hshtN7L
Branch: claude/issue-5903-objectgantt-declared-keys
Worktree: objectui-issue-5903
Domain: domain:ui
File surface: packages/types (ObjectGanttSchema — its TS declaration and its zod mirror) + packages/plugin-gantt/src/** (removing the as any) + tests + a changeset
Container & model: M, mode:subagent, model: opus
Clause-②: yes — a published schema's accepted key set changes. Declared; goes out at Opus under this seat's recorded Fable-quota exemption.
Serial constraints cleared: packages/plugin-gantt/** — no in-flight claim. ⚠️ packages/types just came free: #5927's PR #6032 (Group A zod-mirror widenings, touching data-display/form/navigation/objectql/views .zod.ts + zod-mirror-parity.test.ts) is enqueued, not yet merged. Branch from a merge-base that already contains it if it has landed; if it has not, ⛔ expect a conflict on zod-mirror-parity.test.ts and rebase rather than resolving by hand. Concurrent siblings: #5872 (packages/test-support + ~10 parity test files — ⚠️ same package neighbourhood, see below), #5835 (packages/app-shell/src/views/RecordDetailView.tsx), #6010 (packages/components/src/renderers/form/form.tsx), #5808 (packages/plugin-detail/src/renderers/__tests__/).

裁决 — triage's grade and its fence

Triage seat, 2026-08-24 (daily round), domain:ui + pm:queue, Task. Verbatim:

lands in packages/types ObjectGanttSchema + plugin-gantt — declare the 11 read keys (or explicitly rule each out and delete the read), removing the load-bearing as any. Per-package specimen of the on-hold class card #4631; doing this one does not reopen the class.

Two admissible outcomes per key, and you choose per key on evidence:

1. Declare it — the renderer reads it and the read is legitimate ⇒ it belongs in ObjectGanttSchema and its zod mirror.
2. Delete the read — the key is dead or the read is wrong ⇒ ADR-0049 enforce-or-remove, and the read goes, not the declaration (there is no declaration).

⛔ A third outcome is not available: leaving the as any in place. It is the load-bearing thing here — it is what makes the read invisible to every gate, and it is why "declared ≠ enforced" has no detector on this surface.

⛔ Do not reopen class card #4631. Triage says so explicitly. If your work suggests the class deserves a different treatment, ⛔ report it; ⛔ do not widen this card past ObjectGantt.

PM mechanism assumptions — measure, do not assume

1. The card's count of 11 is a claim from 2026-08-23 and packages/types has moved a lot today — feat(types)!: retire the ActionCondition branch shape from ActionSchema.condition #5981, feat(types): close the dashboard widget type vocabulary and gate the plugin-dashboard catalog #6003, feat(types,components)!: ui:icon names its glyph with icon, not the identity key name #6012 and (imminently) fix(types): widen seven zod mirrors the renderer already implements (#5927 group A) #6032 all landed or are landing in it. ⛔ Do not work from the card's list. Re-derive the read set on your merge-base and report the delta key by key. ⚠️ In particular, feat(types): close the dashboard widget type vocabulary and gate the plugin-dashboard catalog #6003 closed the dashboard widget vocabulary and feat(types,components)!: ui:icon names its glyph with icon, not the identity key name #6012 renamed IconSchema's glyph key — if either moved a key you were about to declare, say so.
2. ⚠️ as any makes the read set unmeasurable by grep alone. A key read as (schema as any).foo and a key read as (schema as any)['foo'] and a key destructured off an any-typed local are the same defect and do not share a pattern. State the method you used to enumerate, and give a counter-probe: a key you know is legitimately declared must be found by the same method, or the method is blind.
3. Declaring a key is a published-surface change and the mirror must move with it. This repo's zod-mirror-parity ratchet measures TS-vs-mirror drift in both directions over 163 registered pairs — ⚠️ declaring a key on the TS side alone will turn it red, and adding a KnownDrift entry to silence that is ⛔ forbidden: the ledger is shrink-only and 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927 just shrank it from 17 to 13. Declare on both sides at constant requiredness.
4. "The renderer reads it" is a claim about running code. For each key you declare, show the read site. For each read you delete, show that nothing else reaches it and give a counter-probe (a sibling key the same method does find live).

Verification

Reverse-verify at the boundary the as any was hiding, direction predicted before running: with the as any removed and the keys declared, restore the declaration to its pre-card state and show type-check goes red naming the keys — that red is the entire value of this card, because it is the signal that did not exist before.

Counter-probe required: authoring a key that is not declared must now fail (type error, or a zod parse rejection if the mirror is strict on that node) — otherwise "the keys are declared" is satisfiable by a declaration nothing enforces, which is the exact defect one layer up. ⚠️ Note BaseSchema is .passthrough(), so a zod parse may not reject an undeclared key — measure which enforcement you actually get and say so plainly rather than assuming the mirror rejects.

Prove every mutation on disk by grepping the injected text and separately the removed text, and ⚠️ print the landing site rather than assuming it — measured twice this round (#4730's mis-anchored injection that still moved the number; #5927's anchor hitting 3 sites because sibling schemas carry byte-identical lines). ⚠️ Assert anchor uniqueness before writing, not after. Restore under trap … EXIT INT TERM; git diff HEAD --stat empty afterwards.

Gates by name with exit codes: pnpm --filter @object-ui/types type-check, pnpm --filter @object-ui/plugin-gantt type-check (⚠️ hyphenated type-check; echo the script name so a zero-match silent pass cannot read as green — and a first RED reading Cannot find module @object-ui/... is the unbuilt-closure signature, not a real failure), root-form vitest over packages/types/src and packages/plugin-gantt/src (⚠️ this repo's vitest guard refuses the --filter forms — use pnpm exec vitest run <paths> from the repo root), and the downstream prefix-form sweep pnpm --filter '...@object-ui/types' with the contract direction stated. ⚠️ Known gotcha, measured on #5927 today: that prefix form fails at apps/site, and the downstream type-check fails in packages/components, until @object-ui/react-runtime is built separately — it does not depend on @object-ui/types so is never in the prefix closure, while packages/components needs its dist. A root build includes it. ⛔ Do not read that as a finding. eslint on changed files with the merge-base delta stated; ⚠️ plain eslint . per package — ⛔ do not add --no-inline-config, it manufactures phantom errors.

Changeset required. Draft PR, Fixes #5903. ⛔ Do not mark it ready, ⛔ do not enqueue, ⛔ do not enable auto-merge — the PM lands it. Post the terminal report here (plain-text first-line marker os-dev-report; ⛔ avoid short angle-bracket placeholders in prose — this repo's sanitizer eats them) and return it.

---

Generated by Claude Code