platform-objects(i18n): es-ES and ja-JP carry the SAME source-parity defect as zh-CN at nine of the 26 enumerated metadata-form leaves — the "zh-CN is alone" reading holds only at the two leaves it was measured on

[claude]

Filed unassigned and ungraded by the domain:engine execution seat running #14931. ⛔ domain:*, priority and type are triage's.

The reading this corrects

#14587's census and #14931's triage both record that this defect class is zh-CN's alone:

es-ES and ja-JP are correct at both leaves. zh-CN is alone.

That is true — at the two leaves it was measured on (dashboard.sections.layout.description and dashboard.fields.header.helpText). It does not generalise. Measured just now across the 21 remaining leaves of the census's own 26-leaf table, at origin/main 9f3989771: es-ES and ja-JP carry the same defect at nine of them, including both leaves the card singles out as its case for priority:p2.

Method: the three bundles under packages/platform-objects/src/apps/translations/ flattened to dotted leaf paths (773 leaves each, key sets identical) and each locale's value read against its en counterpart, which is rewritten from source on every extract and never merged. Same test the census applied: does the string assert something the source does not, or drop a distinct concept the source names.

The nine

leaf	source (en)	es-ES	ja-JP
action.fields.body.helpText	Either an L1 expression or an L2 sandboxed JS body	Código JavaScript que ejecutar	実行する JavaScript コード
page.fields.type.helpText	Page kind. "List / Interface" binds a source view into a curated surface — how it looks (grid / kanban / calendar / …) is a visualization set under Interface, not a page type.	Tipo de página: list / record / home / app / utility. …	ページの種類 — list / record / home / app / utility。…
action.sections.advanced.description	AI exposure and API request shape.	Operaciones masivas, exposición a IA y forma de solicitud API.	一括操作、AI 公開、API リクエスト形状。
object.fields.fields.trackHistory.helpText	Summarize this field on the record activity timeline	Conservar historial de cambios	変更履歴を保持
object.fields.isSystem.helpText	System object (protected from deletion; defaults sharing to public)	Objeto de sistema (protegido contra eliminación)	システムオブジェクト（削除から保護）
view.fields.filter.helpText	Filter conditions — same visual builder as the list toolbar, with field-type-aware operators and value inputs	Condiciones de filtro	フィルター条件
report.sections.basics.description	Identity and report type.	Identidad y fuente de datos.	ID とデータソース。
report.fields.columns.helpText	Dimension names across (matrix only)	Columnas que mostrar en el informe	レポートに表示する列
email_template.fields.variables.helpText	[{ "name": "user.name", "type": "string", "required": true, "description": "..." }]	Lista de nombres de variable referenciados en subject/body	subject/body で参照する変数名リスト

Two borderline cases are named rather than counted, so the nine is a floor rather than an estimate: app.fields.defaultAgent.helpText (both drop the 'ask' / 'build' defaults and both name a button the source does not) and email_template.sections.identity.description (both translate an older source sentence about "identifier and content type").

The other ten of the 21 are faithful in es-ES and ja-JP — e.g. permission.sections.identity.description carries both source sentences in both locales, and skill.sections.triggers.description is correct in both. So this is not "the whole table is wrong everywhere"; it is a measured nine.

⭐ Why this matters more than a count

The provenance story on #14931 explains zh-CN's 613 legacy-trusted leaves as a one-time import of a pre-consolidation hand vocabulary (e0077ea36), never reconciled — and concludes: "a one-time reconciliation, not a recurring leak — which is what makes a bounded repair the right shape and a new gate the wrong one."

Nine of these leaves are wrong in all three locales, which a single bad import cannot explain. es-ES and ja-JP were translated correctly from the seeded English by 874cb084e; they read wrong today because the English source moved afterwards and merge fills gaps only. That is the check-i18n-stale-fill shape (#11671) — a leaf serving a superseded source revision — arriving at leaves the digest machinery does not cover, ⛔ not the import story.

⇒ The two mechanisms are different, and the second one is recurring: every future source-string revision strands every locale's leaf. ⛔ This issue does not propose a gate — #14931 is explicit that a second stale-detection instrument is the debt, not the fix, and that whether a gate should exist is a maintainer call. It records that the "one import, one reconciliation" framing is too narrow to cover what was measured.

Scope, and what is already in flight

• Not repaired by PR fix(platform-objects): 21 zh-CN metadata-form leaves say what their source says #16226 (the zh-CN half of 613 zh-CN metadata-forms leaves are LEGACY-TRUSTED — imported once from a pre-consolidation hand vocabulary, never compared against source, and 26 of them are wrong. No instrument in the repo can judge them. #14931): that PR is bounded to zh-CN by its dispatch, and touching these files would collide with open work.
• PR fix(platform-objects): the es-ES and ja-JP dashboard gap help text says what its source now says #16190 is open on exactly these two files right now, but it is bounded to one leaf, dashboard.fields.gap.helpText — it does not reach any of the nine.
• platform-objects(i18n): the es-ES and ja-JP dashboard gap helpText leaves went stale on the same source edit as the zh-CN one — same class, two more locales #16165 is the es-ES / ja-JP card for that single gap leaf. This is the same class, nine more leaves, and is not covered by it.

Refs: #14931 (the zh-CN half and the census) · #14587 comment 5524387962 (the 26-leaf table) · PR #16226 · #16165 / PR #16190 (the gap leaf pair) · #11671 (the superseded-source-revision shape).

---

Generated by Claude Code