fix(spec,objectql)!: retire scale from the currency field type — refused at parse, no longer enforced on writes

[objectstack-fleet]

Fixes #19629

Clause-②: no (narrowing)

Executes maintainer ruling 5791803339 (batch #215 item 1, letter B, 「215 同意」): scale is retired from the currency field type. Its refusal remedy and its ADR-0087 migration entry are worded by maintainer ruling 乙 5805782503 (batch #218 item 2, 「其他同意」), item 1:

PR #19909's refusal remedy and ADR-0087 migration entry are reworded: delete scale on a currency field; the currency's ISO 4217 minor unit decides its display, and its write allowance stays unconstrained (today's contract). ⛔ No pointer to currencyConfig.precision.

The consumer half is objectstack-ai/objectui#10221 and is not in this diff: the field designer, the grid summary footer, and, under ruling 乙 item 2, the dashboard ObjectMetricWidget. B′ (enforcing a currency width on writes) was not taken and is not here.

Landing order (ruling 乙 item 2)

This PR lands after objectstack-ai/objectui#10221 has landed (widened to ObjectMetricWidget) and after this repository's .objectui-sha pin bump has moved past it. That order keeps any zero-decimal window from opening on the grid summary footer or the dashboard metric widget. At this writing the pin is 62597c588072 and objectui#10221 is open with no PR (both re-read in round 3). Holding the landing to that order is the seat's.

What changes

1. @objectstack/spec, FieldSchema (packages/spec/src/data/field.zod.ts): scale on a type: 'currency' field is refused at parse, with one custom issue at scale (the file's per-type superRefine house pattern). No alias and no grace window. The remedy's first sentence is "scale is not valid on a currency field — delete the key." It goes on to say that the currency's ISO 4217 minor unit decides how the amount displays, and that the field's write allowance stays unconstrained. It names no other key to carry the value. The scale describe names the set the key still applies to: number, percent, rating and slider, enforced on writes, and formula, rounded. On currency it now reads "REFUSED on a currency field — delete it there".
2. @objectstack/objectql, record validator (packages/objectql/src/validation/record-validator.ts): the max_scale branch no longer reads def.scale for currency, so the type leaves the enforced set. min, max and the finite-number check still apply to currency. percent, number, rating and slider are unchanged. Nothing else is read in its place: a currency's write allowance stays unconstrained.
3. ADR-0087 semantic entry field-currency-scale-refused (packages/spec/src/migrations/entries/semantic/18.field-currency-scale-refused.ts). Its replacement is "DELETE the key — that is the whole migration", and it says nothing replaces the key. It is deliberately not a mechanical conversion: a conversion that dropped the key would accept it on every load, which is the grace window ruling B refused. registry.ts is regenerated with gen:migration-registry, never hand-edited.
4. Shipped examples that the refusal would reject. scale: 2 was deleted from each, and nothing was added:
   • examples/app-crm/src/objects/account.object.ts, opportunity.object.ts, opportunity-line-item.object.ts (×2)
   • examples/app-showcase/src/data/objects/account.object.ts, client-brief.object.ts, expense-report.object.ts, external/customer.object.ts, external/order.object.ts, field-zoo.object.ts, invoice.object.ts (×3), project.object.ts (×2)
5. Documentation examples. scale: 2 was deleted from 13 currency examples in content/docs/concepts/architecture.mdx (×2), concepts/metadata-driven.mdx (×2), data-modeling/fields.mdx, data-modeling/objects.mdx, data-modeling/schema-design.mdx, protocol/kernel/plugin-spec.mdx, protocol/objectql/schema.mdx (×4) and protocol/objectql/types.mdx. The schema.mdx property table's scale row no longer lists currency. It now says "Refused on currency — delete it there".
6. Generated, never hand-edited: content/docs/references/data/field.mdx, content/docs/references/data/object.mdx and content/docs/references/system/migration.mdx (the scale describe). authorable-surface, api-surface, json-schema.manifest, spec-changes.json and the upgrade guide did not move, because no key or export was added or removed. The nine *.metadata-forms.generated.ts i18n bundles did not move either: round 3 changes a visibleWhen, not a label or help text, and those bundles carry only label and help text.
7. packages/spec/src/data/field-scale.ts: the docblock's currency bullet records the retirement and ruling 乙's remedy. It no longer names CurrencyConfigSchema.precision as the money faces' override; that claim was false at the pin.
8. packages/spec/src/data/object.form.ts (round 3): the object designer's quick-add fields grid, objectForm, is the form METADATA_FORM_REGISTRY serves for the object metadata type. Its scale row's visibleWhen was "data.type in ['number','currency','percent']" and is now "data.type in ['number','percent']", with a one-line comment citing this card and ruling 5791803339 B. The precision, min and max rows are unchanged.

Measured first, on origin/main 1f89ba0d70 (built dist)

• A currency field with scale: 2 parsed: FieldSchema.safeParse succeeded. Control: the same with type: 'number' also succeeded.
• The record validator refused a currency write over scale: scale: 2, write 1.234 → max_scale {"scale":2,"actual":3}. A currency field without scale accepted 1.23456.
• max_scale read def.scale on all four controls. Each was given scale: 2 and the write 1.23456. number, rating and slider answered {scale:2, actual:5}. percent answered {scale:4, actual:5}, because of the fraction-storage derivation.
• Population census, a TypeScript AST sweep over every tracked .ts/.tsx/.mjs/.js file (7,036 files): 15 example sites, all listed above, and 7 test fixtures, listed below. A JSON walk over 556 tracked .json files found 0 currency-typed objects with scale. A windowed scan of 831 .md/.mdx files found the 13 documentation examples. The one surviving AST hit is packages/spec/src/data/inline-related-columns.test.ts:103. It is InlineGridColumnSchema.scale, a different schema that the ruling does not cover.

Readings the round measured, and where each went

1. currencyConfig.precision has no reader. No face in the pinned console (.objectui-sha 62597c588072) reads it, and no non-test code in this repository does. The cell uses the currency's ISO 4217 minor unit, and the edit widget reads the field-level precision. → Ruling 乙 item 1: the remedy and the migration entry no longer point to that key. Two findings follow from ruling 乙 item 3: currencyConfig.precision is declared and read by nothing, and objectui's CurrencyField reads the field-level precision as decimal places. Both are filed by the domain:spec seat, as [finding] currencyConfig.precision is declared and validated against ISO 4217, but no renderer or runtime reads it — an ADR-0049 enforce-or-remove case, filed on ruling 乙 on #19910 #19992 and [finding] CurrencyField renders the field-level precision (a total-digit count in the spec) as the number of decimal places — filed on objectstack ruling 乙 (objectstack-ai/objectstack#19910) objectui#10276, ⛔ not riders here.
2. Ruling B's control reading undercounted the examples. Fifteen Field.currency declarations in examples/ declared scale: 2. → Corrected on the card by 5805294161.
3. A fourth face. The dashboard ObjectMetricWidget's currency arm reads valueFieldDef.scale ?? 0. → Ruling 乙 item 2: objectui#10221 is widened to it and lands first (see the landing order above).
4. Direction of the validator half. It moves writes from refused to accepted. → Ruling 乙 item 4: Clause-②: no (narrowing) stays as ruled. The widening reading from the finding(spec,objectql): ruling B's percent storage derivation is NOT live — max_scale has no percent arm and the percentScaleOf docblock is silent about scale #19320 precedent is noted and ⛔ not re-declared.

Round 3, on top of 40ef043003

Why. The at-tier contract-review record 5818558748 on this PR (head 40ef043003) read VERDICT: FAIL with one required fix: packages/spec/src/data/object.form.ts:190 still offered scale on a currency row of the object designer's fields grid, the form registered in packages/spec/src/system/metadata-form-registry.ts. After this PR an author using that form is shown a scale input on a currency field and is refused at publish, which is the offer-vs-door shape this retirement removes.

• d7740c324e: the fix exactly as the record states it. The scale row's visibleWhen is now "data.type in ['number','percent']", with the one-line comment. The row was not widened to any other type, and the precision, min and max rows were not touched. The same commit adds the offer-side pin and one changeset sentence (both below). origin/main was not merged: the PR reads mergeable_state: clean, and no file main changed since the merge base c8399867b8 is a file this branch changes (comm of the two name lists is empty), so CI's merge ref needs nothing from a merge here.

Sweep of packages/spec/src/** for any other surface offering scale on currency. All 17 *.form.ts files were read for scale rows, and every non-test source file naming both scale and currency was read at the hit.

file:line (HEAD d7740c324e)	what it is	changed?
data/object.form.ts:191	the objectForm fields-grid scale row (it was :190)	yes, the record's fix
data/field.form.ts:79	the fieldForm scale row, "data.type == 'number'"	no; it does not offer the key on currency (the record's own control)
data/field.zod.ts:1235-1236	FieldSchema.scale and its describe, "REFUSED on a currency field"	no; that is the door, from round 2
data/field.zod.ts:1025	FieldSchema aliases decimals / decimalPlaces → scale	no; on currency the renamed key lands on the same refusal (acceptance note below)
data/field.zod.ts:420	CurrencyConfigSchema alias scale → precision, inside currencyConfig	no; a different schema, pre-existing, and the surface of #19992 (acceptance note below)
data/field.zod.ts:931	InlineGridColumnSchema.scale	no; a different schema, left alone as instructed
ui/view.zod.ts:3338	FormFieldSchema.scale, a form-view row's widget override with no type gate by design (a form row usually omits type)	no; a different schema, and its describe names no field type
ui/view.zod.ts:1600	the timeline view's scale (hour … year)	no; unrelated
ui/bulk-action.zod.ts:157	scale in BULK_PARAM_WIDGET_CONFIG_KEYS, a list of keys bulk params REFUSE	no; a refusal list, not an offer
data/field-scale.ts:101	ABSENT_SCALE_BY_TYPE, whose one row is percent	no; there is no currency row
data/numeric-column-representation.ts:49-52	a dated measurement docblock that names a currency field with scale: 2 refused at the write seam	no; a record of a reading, not an offer (acceptance note below)
studio/**	designer hints	0 currency hits
packages/platform-objects/src/apps/translations/*.metadata-forms.generated.ts (outside packages/spec/src)	the generated form bundles; fields.scale carries only its label and help text (en :147-150)	no; no predicate lives there, and check:i18n reads all nine packages' bundles in sync

A quoted-exact git grep -n -F "['number','currency','percent']" over the whole tree now answers one line, the precision row at object.form.ts:189, which the record says to leave. Before the fix it answered that row and the scale row.

The new pin (packages/spec/src/data/field-currency-scale-refused.test.ts, final block). It evaluates each row's visibleWhen the way form-delete-behavior-options.test.ts already does: a fail-closed reader of the data.type spellings these forms use (== and in [...], joined by ||) that throws on anything else, so a predicate it cannot read fails the test instead of being assumed visible or hidden.

• CONTROLS (lit): the walk over every form in METADATA_FORM_REGISTRY finds exactly two scale rows, field:scale and object:fields.scale. METADATA_FORM_REGISTRY.object is objectForm. The reader lights on a row that does name currency.
• Firing case: the objectForm scale row is not offered when data.type is currency.
• Dark controls: it is still offered for number and percent, and the door agrees on both types (FieldSchema.safeParse with scale: 2 succeeds).
• Class guard: no registered form offers scale on a currency field.

Old-row pins. Nothing pinned the old row: before the fix, the quoted-exact git grep above found the predicate only in object.form.ts itself, and none of the tests that read objectForm or METADATA_FORM_REGISTRY mention scale. There was nothing to reverse.

Round 2, on top of 7c0a33c6ad

• 07c9be36bf: a merge of origin/main c8399867b8, run through scripts/pm/os-regen-merge.sh. registry.ts text-merged, and gen:migration-registry then wrote no change. check:generated read "All 15 generated artifacts are up to date".
  • Main changed none of the three reference pages this branch changes, so the script kept the branch's bytes for them. The three reference pages main did change (api/package-api.mdx, security/permission.mdx, system/translation.mdx) are byte-identical to origin/main (git diff empty).
  • Quoted-exact git grep survival checks:
    • Each of the 7 semantic entries main added since the merge base appears exactly once in both origin/main's and HEAD's registry.ts: admin-scope-business-unit-blank-refused, filter-equality-array-comparand-refused, flow-predicate-slot-blank-string-refused, package-api-contracts-unmounted-entries-retired, rls-predicate-array-comparand-refused, translation-per-app-settings-platform-only and view-filter-rule-absent-value-refused.
    • Main's added reference line "Required and non-blank: an empty or whitespace-only value names no business unit" appears 2 times in permission.mdx on both sides.
    • Main's deleted line "POST /api/v1/packages/upgrade" appears 0 times on both sides.
• e2c5077421: the rewording to ruling 乙, across the refusal message, the describe, the migration entry (with registry.ts regenerated), the validator's comments, the field-scale.ts docblock, the schema.mdx row, the changeset and the spec pins. What is refused and what is accepted did not change.
• 40ef043003: gen:docs regenerated the three reference pages. A later gen:schema && gen:docs produced no further diff, and check:docs read "225 generated files in sync with packages/spec".

Tests, on HEAD d7740c324e

Every command went through scripts/pm/os-verify-lock.sh. Each count is the runner's own summary line.

• Build: turbo run build over ./packages/*, ./packages/*/* and ./examples/* gave 73 successful, 73 total.
• @objectstack/spec: vitest run --project local gave 532 files, 15667 passed, 2 todo (round 2: 15663; the 4 new offer-side tests). --project repo gave 35 files, 602 passed.
• field-currency-scale-refused.test.ts alone: 12 passed (8 from round 2, 4 new).
• @objectstack/spec typecheck exited 0. check:test-typecheck held 53 files, 255 errors and 142 pinned signatures, the same as rounds 1 and 2. check:type-check-coverage exited 0, so the package's tsconfig reaches the edited test file.
• Consumers that read the registered forms: @objectstack/platform-objects translation suites gave 6 files, 137 passed; @objectstack/lint validate-predicate-path-refs.test.ts gave 1 file, 54 passed; @objectstack/metadata-protocol protocol.meta-types-* gave 3 files, 38 passed.
• NOT re-run this round: @objectstack/objectql (on 40ef043003: 309 files, 5196 passed; 1 file, 5 passed; typecheck exit 0 with 40 files, 234 errors and 65 pinned signatures held), @objectstack/example-crm, @objectstack/example-showcase, and round 1's platform-objects, metadata-core, driver-sql and service-automation suites. No file in those packages changed since 40ef043003. They are declared to CI.

Reverse verification

All legs are one-shot proofs run from the committed state, each with a trap on EXIT/INT/TERM. No rebuild was needed, because the test file resolves ./object.form, ./field.zod and ../system/metadata-form-registry through relative src imports.

Round 3, the offer. scripts/ablation-replace.mjs restored the old row, visibleWhen: "data.type in ['number','currency','percent']", on object.form.ts. The landing was proved on disk: new-row anchor 1 → 0, old row 0 → 1, blob 6e87b93283 → 295afbe69f. Then field-currency-scale-refused.test.ts ran: 2 failed | 10 passed. The two failures are the firing case ("expected true to be false") and the class guard (it received object:fields.scale where it expected none). The lit CONTROLS, the number and percent dark controls and all 8 round-2 tests stayed green. Restored: blob 6e87b93283 == HEAD, git diff HEAD empty, and git status --porcelain empty.

Round 2's two legs, on 40ef043003:

• The ruled wording. field.zod.ts was restored to its round-1 blob 0c24baa382, which carries the old remedy naming the other key. Main changed nothing in this file across the merge, so that blob differs from HEAD only by the round-2 rewording. The landing was proved by blob hash, with the old-remedy anchor ×1 and the new one ×0. Then field-currency-scale-refused.test.ts ran: 4 failed | 4 passed. The three message pins went red on the first-sentence assertion, and the describe pin went red on its new clause. The firing control (every declared value refused) and the three dark controls stayed green. Restored: blob d6ada97894 == HEAD, and git diff HEAD is empty.
• The absence half. scripts/ablation-replace.mjs appended "Move the value to currencyConfig.precision instead." to the new message. Anchor 1 → 0, blob d6ada97894 → 017fdba420. Result: 3 failed | 5 passed. All three failures read "not to match /currencyConfig|precision/", and the describe pin and every control stayed green. Restored: blob == HEAD, and git diff HEAD is empty.

Test pins of the old remedy, reversed

A repo-wide grep finds the old remedy pinned in one file only, packages/spec/src/data/field-currency-scale-refused.test.ts. It had a message toContain on the other key and a control named "the remedy target parses". Here is what replaces them:

• One helper asserts the message's first sentence verbatim, the two ruled clauses, and that the message names neither currencyConfig nor precision. It runs on all three firing paths: the designer shape, the shape beside a currencyConfig block, and the Field.currency() + ObjectSchema path.
• The control is now "following the remedy parses". It takes each refused shape, deletes scale, adds nothing, and the result is accepted.
• The describe pin asserts the new clause and the absence of currencyConfig.
• packages/spec/src/data/currency-precision-iso4217.test.ts:293 names that key only as CurrencyConfigSchema's ISO-contradiction issue path. It is not a remedy pin and is unchanged.

currencyConfig hits that remain (git grep on HEAD 40ef043003, unchanged at d7740c324e)

• In the branch's added lines (c8399867b8..HEAD): 7 hits. None names that key as a remedy or as a decimal-places carrier.
  • The changeset's FROM → TO row 2 shows a field that declares a currencyConfig block. Its fix is still the deletion.
  • field-zoo.object.ts:53 is in the diff only because scale: 2 was deleted from that line. Its currencyConfig: { …, precision: 2 } is origin/main's, unchanged.
  • field-currency-scale-refused.test.ts has 5 hits: two absence assertions, one test title, and two fixtures with a currencyConfig block and no precision (the firing case and the remedy-follow control). Round 3 added none.
• In the touched files, pre-existing (the same count on c8399867b8 and HEAD):
  • fields.mdx 1, types.mdx 2 and the three reference pages 2 each. The reference hits are generated from CurrencyConfigSchema.
  • Showcase account.object.ts 2, field-zoo.object.ts 1.
  • field.test.ts 34 and field.zod.ts 10: the CurrencyConfigSchema definition and tests, the ISO check, and the currency-is-not-a-field-key refusal.
  • registry.ts 1, in another entry.
  • These stay because this PR did not write them. The key's own disposition is ruling 乙 item 3's first finding, for the domain:spec seat.

Test fixtures re-judged (round 1)

These fixtures pinned a currency field with scale, or max_scale on currency:

• packages/spec/src/data/field.test.ts: "should accept number field with precision and scale" spelled type: 'currency'. It now spells number, a spelling change only.
• packages/spec/src/data/object.test.ts: the account fixture's annual_revenue lost scale: 2.
• packages/spec/src/data/field-scale.test.ts: the resolver's "declared scale wins" example moved from currency to slider.
• packages/objectql/src/validation/record-validator.test.ts: the finding(spec,objectql): ruling B's percent storage derivation is NOT live — max_scale has no percent arm and the percentScaleOf docblock is silent about scale #19320 CONTROLS loop dropped currency. A new [finding] FieldSchema.scale on a currency field is offered to authors by the field designer, ignored by every display face, and still enforced on writes — half-live in the direction that surprises #19629 block holds the three ruled pins, and its header now cites ruling 乙 for the unconstrained write allowance.
• packages/services/service-automation/src/flow-field-expression-scale.integration.test.ts: the Flow field expressions can call no function but NOW()/TODAY() — every other identifier is rewritten to null, so a computed money value can never be rounded to its field's declared scale #11060 oracle needs a live scale gate, so its total field moved from currency to number.
• packages/drivers/driver-sql/src/sql-driver-external-unprovisioned-sort-anchor.test.ts and packages/metadata-core/test/injected-column-provenance.test.ts mirror the showcase external customer. Both lost scale: 2.

Changeset

.changeset/19629-currency-scale-retired.md sets @objectstack/spec: minor and @objectstack/objectql: minor, with a BREAKING banner. Per AGENTS.md, Clause-②: no (narrowing) is breaking, and check-changeset-no-major refuses major. Round 2 reworded it to ruling 乙:

• Its remedy sentence and FROM → TO table now prescribe deletion only.
• Its write sentence says the allowance stays unconstrained.
• Its console bullet says both faces derive from the currency in the console this release bundles, because of the landing order above.

Round 3 added one sentence to its @objectstack/spec paragraph, because the form change is visible in Studio: "Studio's object editor no longer offers scale on a currency field: the fields grid of the objectForm this package registers in METADATA_FORM_REGISTRY now shows it only for number and percent." It names no key to carry the value, so it holds under ruling 乙.

It carries the ADR-0087 marker registered field-currency-scale-refused. check-adr-0087-registration reads "[BREAKING+bang+clause-②-narrowing] registered field-currency-scale-refused". The test-only packages (driver-sql, metadata-core, service-automation) ship nothing that changed, and the examples are private.

Acceptance notes

• docs/qa/platform-checklist/areas/records-forms.json: the field-zoo fixture line still reads "f_currency scale 2". Its knownGaps text, which says scale is not enforced on write, has been stale since A number field's declared scale is never enforced — values with more decimals are accepted and stored verbatim (min/max on the same field are enforced) #7501. This is checklist-owned text and is left for that seat.
• Four documentation YAML examples keep precision: 18 on a currency field (schema.mdx ×3, types.mdx). They are meant as DECIMAL(18,2) total digits, but the pinned edit widget reads field-level precision on currency as fraction digits. This predates the PR. It is the surface of ruling 乙 item 3's second finding.
• data-modeling/fields.mdx ("Currency Configuration") and protocol/objectql/types.mdx show currencyConfig: { precision: 2, … } as a currency's configuration. This predates the PR and is the surface of ruling 乙 item 3's first finding.
• FieldSchema's alias table renames decimals / decimalPlaces to scale. On a currency field that rename now leads to this refusal, and its remedy tells the author to delete the key.
• hotcrm declares scale: 2 on a currency quote field (hotcrm#1206, cited by the flow oracle). That becomes a refused parse downstream, and the migration entry prescribes the deletion.
• objectui#10221's title still says "derive the summary footer's fraction digits from currencyConfig.precision / ISO minor units", which is ruling B's wording. Under ruling 乙 only the ISO half holds. Rewording that card belongs to the domain:ui seat.
• Round 3 sweep: CurrencyConfigSchema (field.zod.ts:420) accepts currencyConfig: { scale: … } and renames it to currencyConfig.precision, the key [finding] currencyConfig.precision is declared and validated against ISO 4217, but no renderer or runtime reads it — an ADR-0049 enforce-or-remove case, filed on ruling 乙 on #19910 #19992 records as read by nothing. This predates the PR and is not FieldSchema.scale; it goes wherever [finding] currencyConfig.precision is declared and validated against ISO 4217, but no renderer or runtime reads it — an ADR-0049 enforce-or-remove case, filed on ruling 乙 on #19910 #19992's enforce-or-remove decision takes that key.
• Round 3 sweep: the field editor form's precision row (field.form.ts:78) is shown on currency with the help text "Decimal places (e.g., 2 for $10.50)", while FieldSchema.precision's describe (field.zod.ts:1211) says "Total digits" and objectForm's own precision row says "Total digits". That is a precision row, which this PR does not touch; it is the spec-side neighbour of [finding] CurrencyField renders the field-level precision (a total-digit count in the spec) as the number of decimal places — filed on objectstack ruling 乙 (objectstack-ai/objectstack#19910) objectui#10276.
• Round 3 sweep: numeric-column-representation.ts:49-52 records a write-seam measurement in which a currency field declaring scale: 2 was refused. After this PR the seam no longer reads scale on currency, so the docblock describes a past reading. It is an internal comment, not an offer, and it is left alone.

Gates, on HEAD d7740c324e

• node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands derived 120 families from this worktree. Every one was run with its exit code captured before any pipe, and all 120 exited 0. The --ran reconciliation read: "120 derived famil(ies) accounted for — 120 run, 0 NOT-MEASURED (a DERIVED zero — all 120 recorded an exit code and none of them is 3)". The derivation warned of a stale tree: origin/main e8f163fc3a is 23 commits past the merge base, and 18 derivation inputs changed in that range (among them .github/workflows/lint.yml and package.json). CI runs the PR's merge ref with the current copies.
• pnpm --filter @objectstack/spec check:generated: "All 15 generated artifacts are up to date".
• check:i18n: "OK (9 package(s) — all bundles in sync, no undeclared authoring keys)". check:nul-bytes: OK, no raw ASCII control bytes.
• check-changeset-no-major: no major bump. The level axis reads this PR body, so it is NOT MEASURED locally.
• CI on d7740c324e, read at 2026-09-24T18:25Z: 35 check-runs, 33 success and 2 skipped (Console Pin Gate, Packed-tarball smoke (opt-in)), none failed. All seven required contexts and Check Changeset read success.
• Lint was narrowed to the diff and was not run repo-wide. eslint --no-inline-config --format json over the 25 changed .ts/.mjs files (c8399867b8..HEAD, round 3 adds object.form.ts) reported 25 files, 0 errors, 0 warnings and no ignore notices. The narrowing holds because eslint.config.mjs never enables type-aware linting (its own comment at lines 326-328: no parserOptions.project, no typed rules), so this diff cannot move any untouched file's verdict. The full pnpm lint belongs to CI.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 2 package(s): @objectstack/objectql, @objectstack/spec, touching 3 documentable anchor(s). ⚠️ 1 changed file(s) yielded no anchor (packages/spec/src/data/field-scale.ts), so the pages documenting them are NOT COVERED by this run — this is not a clean bill of health for those files.

12 hand-written doc(s) NAME something this change touched and may need an implementation-accuracy re-verification:

• content/docs/concepts/metadata-driven.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/data-modeling/external-datasources.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/data-modeling/field-types.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/data-modeling/validation-rules.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/deployment/troubleshooting.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/deployment/validating-metadata.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/getting-started/quick-reference.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/kernel/contracts/data-engine.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/protocol/backward-compatibility.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/protocol/objectql/types.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/protocol/objectui/concept.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/ui/forms.mdx (via FieldSchema (symbol, a top-level const))

⛔ 2 release-owned page(s) also name something this change touched. These are read-only:

• content/docs/releases/v17/17-0.mdx (via FieldSchema (symbol, a top-level const))
• content/docs/releases/v17/17-1.mdx (via FieldSchema (symbol, a top-level const))

content/docs/releases/ is RELEASE-OWNED (AGENTS.md "Documentation Guardrails"): release
notes are written centrally at release time, and a code PR that edits them is the exact PR
that guardrail exists to stop. They are still audited — read-only. If one of them is actually
wrong, file an issue or open a dedicated docs-only PR; do not edit it here.

What this run could not see

• 1 changed file(s) yielded no anchor (packages/spec/src/data/field-scale.ts) — pages documenting those are invisible to this run
• 5 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 60 of 215 client-bound route-ledger rows — the other 155 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 155: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 55 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 100 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 137 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json e8f163fc3a62cc6c65f91d2197f7b516e3a2c90b → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 0141148478a51aa3de9b9a8ca9f58d3c977e998d — the merge of head d7740c324e92789285785ed3632148d078e6b8cf into base e8f163fc3a62cc6c65f91d2197f7b516e3a2c90b, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 0141148478a51aa3de9b9a8ca9f58d3c977e998d && git checkout 0141148478a51aa3de9b9a8ca9f58d3c977e998d
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin e8f163fc3a62cc6c65f91d2197f7b516e3a2c90b d7740c324e92789285785ed3632148d078e6b8cf && git checkout -B drift-repro e8f163fc3a62cc6c65f91d2197f7b516e3a2c90b && git merge --no-ff d7740c324e92789285785ed3632148d078e6b8cf

node scripts/docs-audit/affected-docs.mjs --json e8f163fc3a62cc6c65f91d2197f7b516e3a2c90b

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs e8f163fc3a62cc6c65f91d2197f7b516e3a2c90b → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[objectstack-fleet]

Landing-order note (ruling 乙 item 2): the first half is met. 2026-09-24T20:10Z

objectstack-ai/objectui#10221 landed as objectui 0651e7ab4 (PR objectstack-ai/objectui#10348, merged 2026-09-24T20:03Z), widened to ObjectMetricWidget as ruled. With a currency resolved, the grid summary footer and the metric tile now format with the list cell's formatCurrency, on the currency's ISO 4217 minor unit. Neither reads scale or precision on a currency, and the designer no longer offers Scale on currency.

The second half is still open: this repository's .objectui-sha is 62597c588 (bumped by #19832), which does NOT contain 0651e7ab4 (checked with git merge-base --is-ancestor). This PR still waits on a console pin bump past 0651e7ab4.

Carrier note for the objectui spec-pin bump that eventually ships this PR's refusal. It was recorded at objectui#10221's acceptance and is ⛔ not this PR's to fix. A stored currency field def that still carries scale, including one switched from number to currency (the designer keeps every key), no longer has a control in the objectui designer to clear the key. Under the pinned @objectstack/spec 17.4.0 that is harmless. Once objectui installs a spec with this PR's parse refusal, saving such a draft is refused with nothing on screen to fix it. objectui#4644 is the precedent: a type-conditional strip on the designer's write path. Whoever bumps objectui's spec pin past this PR owns that, and the domain:ui seat will pick it up there.

domain:ui seat #4, session_01BP8CMtACxTdLjqR6rhd33C.

---

Generated by Claude Code

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: d7740c324e92789285785ed3632148d078e6b8cf

① Derived judgments

Round-3 delta is exactly the required fix, and nothing else moved. git show d7740c324e touches three files: packages/spec/src/data/object.form.ts (+2/−1), packages/spec/src/data/field-currency-scale-refused.test.ts (+109), .changeset/19629-currency-scale-retired.md (+1/−1). git diff 40ef043003..HEAD --stat confirms no other file changed between the FAILed head and this one, so every finding the prior review marked right (field.zod.ts:2281-2292 refusal scoped to currency only; ruling 乙 wording with no currencyConfig/precision; record-validator.ts:811 t !== 'currency' && gate with percent/number/rating/slider untouched; ADR-0087 entry byte-identical in registry.ts:8333-8375; 15 example sites and 13 doc examples losing exactly scale: 2; generated reference docs matching the describe) stands unchanged on the full diff c8399867b8...HEAD (37 files, +573/−63).

Blocking item fixed — right. object.form.ts:190-191: the scale row's visibleWhen is now "data.type in ['number','percent']", with a one-line #19629 (ruling 5791803339 B) comment. helpText: 'Decimal places' and type: 'number' are unchanged; the min/max rows (:187-188, still include currency — correct, the validator still applies them to currency) and the precision row (:189, #7918, out of scope) are untouched. The row sits in the field: 'fields' repeater (:59, nested fields: at :71), i.e. the Studio object editor's quick-add grid registered at metadata-form-registry.ts:68 as object.

Sweep of every other shipped form surface — clean. grep scale packages/spec/src/**/*.form.ts returns only field.form.ts:79 ("data.type == 'number'", pre-existing, no currency) and the fixed object.form.ts:191. All 16 registry entries (metadata-form-registry.ts:68-92, incl. system/email-template.form.ts) are under that glob. No other packages/** or apps/** non-test source pairs scale with currency in an offering position; the remaining hits are comments (numeric-column-representation.ts:50 describes the no-scale write allowance, which is now the only currency case; field-scale.ts:68-85 already states the retirement).

i18n bundles — need not move, and did not. The extractor keys metadataForms.<type>.fields.* on label/helpText only; packages/platform-objects/src/apps/translations/en.metadata-forms.generated.ts:147-149 carries "fields.scale": { helpText: "Decimal places" } unchanged, and no *.generated.ts appears anywhere in the full diff stat. The predicate string is not extracted (the "fields.visibleWhen" key at :219 is the form field named visibleWhen, not this row's predicate).

New pin test — sound, no new defect. The added block in field-currency-scale-refused.test.ts:146-244 walks METADATA_FORM_REGISTRY (imported from ../system/metadata-form-registry, which is the real registered map, and the test also asserts METADATA_FORM_REGISTRY.object === objectForm). rowsNamed recurses sections[].fields and each row's nested fields, so it reaches the repeater row (fields.scale). isOfferedForType is fail-closed: it reads only data.type == '…' and data.type in [...] joined by ||, throws on any other spelling or non-literal member, and treats a row with no predicate as offered (the conservative direction for this assertion). predicateSource accepts both the bare string and defineForm's { dialect, source } normalization (view.zod.ts FormViewSchema.parse). The CONTROLS case pins the exact roster ['field:scale', 'object:fields.scale'] (matches the grep) and proves the reader can say "offered" on a positive control, so the empty result in the final case is a measured zero. The number/percent case also asserts the door agrees (FieldSchema.safeParse({type, scale: 2}).success). The test imports ../system/metadata-form-registry from within data/ — a data→system→data import cycle, but only at module-evaluation of the test, and CI on this head (39 success / 7 skipped / 0 failed) shows it evaluates cleanly; not run locally because the detached worktree has no node_modules, and re-running gate families is out of scope.

Changeset delta — true. The one added sentence ("Studio's object editor no longer offers scale on a currency field: the fields grid of the objectForm this package registers in METADATA_FORM_REGISTRY now shows it only for number and percent") matches object.form.ts:191 exactly and mentions neither currencyConfig nor precision, so ruling 乙's wording rule is kept. Exactly one <!-- adr-0087: … --> marker remains.

② Semver level

Unchanged and still correct: @objectstack/spec: minor, @objectstack/objectql: minor, **BREAKING** banner, Clause-②: no (narrowing), FROM → TO table, one-line fix, one ADR-0087 marker. The form change is a shipped-metadata narrowing of the same accept set the parse refusal already narrows, so it adds no new level; the launch-window convention (minor, never major) and AGENTS.md Post-Task Checklist §3 are satisfied. The ! in the title is consistent with the banner. The "two console faces" bullet remains conditioned on the landing order (ruling 乙), which is the landing hold and not this review's concern.

③ Boundary flags

• status: done, premise_still_valid: true — still confirmed; the offer-vs-door gap the prior round found is now closed on the spec's own surface and pinned.
• Open question 1 (objectui#10221 title wording) — option A stands; not this diff's surface.
• Open question 2 (pre-existing currencyConfig: { precision: 2 } in showcase/docs) — option A stands; those are [finding] currencyConfig.precision is declared and validated against ISO 4217, but no renderer or runtime reads it — an ADR-0049 enforce-or-remove case, filed on ruling 乙 on #19910 #19992's surface, none is remedy advice, unchanged this round.
• field.form.ts:79 already scopes scale to number alone and object.form.ts:191 now to number/percent; the two forms differ on percent but agree on currency, and percent is inside the door's accept set, so this is not a defect of this PR.
• Out-of-scope findings from the prior round ([finding] currencyConfig.precision is declared and validated against ISO 4217, but no renderer or runtime reads it — an ADR-0049 enforce-or-remove case, filed on ruling 乙 on #19910 #19992, objectui#10276) remain filed separately; nothing new to file.

Implemented-by: claude/issue-19629-currency-scale-retired
Reviewed-by: session_01EcrTi7s5oDYPHS4Pi7h31d

VERDICT: PASS

_{Isolated at-tier reviewer adopted by the director seat, summon #29, on the maintainer's instruction 「执行契约复审」 · fed only the card(s), the governing rulings, the PR body, the dev's flags and the code}

---

Generated by Claude Code