[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

[os-steve]

Ruled: 5791803339 · letter B · 2026-09-23T08:49Z

Filed at 2026-09-22T02:11Z by the domain:spec seat 4 (session_01AmH9bKvGoLjiY86Q4Z3og2, seat post #18917), from the
out_of_scope_findings of the #19579 round, marked class (c) by the dev. ⛔ Filed unassigned, ⛔ no
priority:*, ⛔ no type — routing and grading are triage's. ⛔ Not a claim. ⛔ Not a ruling.

⭐ This card exists because a ruling said it must. Ruling batch #194 item 1 letter A′ scoped number and
currency to ride #19579's PR only if a consumer census showed one consistent absent-value per type, and
「where consumers disagree, that type gets its own card with the readings」. The currency census came back
inconsistent. ⇒ this is that card, with those readings.

The defect, in one sentence

FieldSchema.scale on a currency field is offered to authors by a named producer, ignored by every
display face, and still enforced on writes — so setting it buys a narrower write contract and no visible
change. Half-live, in the direction that surprises rather than refuses.

⏱️ Readings — attributed, ⛔ NOT re-taken by this seat

⚠️ Every leg below is the #19579 dev's, taken at objectui origin/main 0cf2d6644b and re-read identically
at the pinned .objectui-sha 87af769e9a. This seat has ⛔ not re-driven them and does ⛔ not assert
them; it files them because the ruling instructs it to and because a card with the readings is worth more
than a card that asks someone to find them again. Whoever takes this re-measures leg 1 first — a producer
is the difference between a dead key and a trap.

Leg 1 — the producer is named, and it is stored metadata, ⛔ not a React prop. The metadata-admin field
designer ships a scale control:
packages/app-shell/src/views/metadata-admin/inspectors/ObjectFieldInspector.tsx:907-909 —
label={tr('designer.field.scale')}, onCommit={(v) => patchDef({ scale: v })}, with the label declared in
i18n.ts:1477 / :3448. ⇒ an author is invited to set it, in a UI this platform ships, and the value is
re-authorable stored metadata.

Leg 2 — no display face reads it on a currency field. CurrencyCellRenderer formats through
formatCurrency(value, currency, locale) and passes no scale; packages/fields/src/widgets/CurrencyField.tsx
reads precision and says so in-line (「No scale passed: this is a currency widget whose precision is a …」);
fraction digits otherwise come from the currency's own ISO 4217 minor-unit count.

Leg 3 — the write path DOES enforce it. packages/objectql/src/validation/record-validator.ts:789-812
enforces max_scale on a currency write from that same declaration. ⭐ This seat did verify the gate's
shape first-hand on origin/main while reviewing #19624, quoted:

if (
  def.scale !== undefined &&
  Number.isInteger(def.scale) &&
  def.scale >= 0
) {

⇒ the enforcement is real and is gated on the declaration being present.

⇒ The observable consequence

An author sets scale: 3 on a currency field in the designer and gets:

• no display change — the cell still renders at the currency's ISO minor-unit width;
• a narrower write contract — values with more than 3 decimals are now refused.

⚠️ ⛔ Nothing tells them either fact. This is the 「declared ≠ enforced」 family, except inverted: the
declaration is enforced, just not where the author was looking.

Why a guessed default is refused here

The ruling names it: a global 0 would turn an undeclared currency from $25.00 into $25, which is
exactly why plain letter A was refused. ⇒ ⛔ this card does not propose a default, and the shape of
the answer is open — align the display faces to the declaration, retire the key from the currency surface, or
split display width from write allowance. ⛔ This card states no preference.

Seam

spec:FieldSchema.scale → producer:ObjectFieldInspector | renderer:CurrencyCellRenderer |
renderer:CurrencyField | objectql:record-validator max_scale

Why it lands in THIS repo

The ruling puts the declaration in packages/spec; the consumer half follows in objectui. 〈多仓协调〉rule 1:
「issue 住在修复落地的仓」.

⛔ Independent card, ⛔ NOT a sub-issue of #19579 and ⛔ not blocked by it. ⚠️ Related but ⛔ not a
duplicate: the sibling card filed in the same act covers number, whose obstruction is different (a
declared 0 also turns grouping off).

Dedupe

currency scale ignored display · currency precision vs scale · formatCurrency ISO 4217 minor units ·
CurrencyField precision · max_scale currency write · designer offers scale currency

[objectstack-fleet]

Round order for PR #19909, at 2026-09-24T15:07Z, by this card's claim holder (session_013RDBh5DqXd2xnLwvHLgLFr, branch claude/issue-19629-currency-scale-retired).

Authorization.

• The maintainer's ruling 乙 on [Decision] Ruling 5791803339 points currency authors at currencyConfig.precision — but nothing reads that key, and removing scale leaves two console faces rendering currency with zero decimals until objectui#10221 lands #19910 (5805782503, batch 🔗 Broken links detected in documentation #218 item 2).
• The maintainer's instruction in this seat's live PM chat, verbatim 「你session 的pr 红了为什么不修复，绿了为什么不 合并。」

This round (ruling item 1).

• Reword the refusal remedy and the ADR-0087 entry 18.field-currency-scale-refused: delete scale on a currency field; the currency's ISO 4217 minor unit decides its display; its write allowance stays unconstrained. ⛔ No pointer to currencyConfig.precision.
• Base-merge origin/main, regenerating registry.ts with the generator.
• Clause-②: no (narrowing) stays as ruled (item 4).

Landing stays held (ruling item 2). objectstack-ai/objectui#10221 must widen to ObjectMetricWidget and land first, and this repo's .objectui-sha must move past it. At this writing that card reads pm:queue, domain:ui, priority:p3, not started. The seat will not arm this PR before the pin covers it.

Ruling item 3. The two findings (currencyConfig.precision read by nothing; objectui CurrencyField reading the field-level precision as decimal places) are this seat's to file through the filing gate, ⛔ not riders on this PR.

[objectstack-fleet]

os-dev-report
{
"issue": 19629,
"status": "done",
"branch": "claude/issue-19629-currency-scale-retired",
"pr": "#19909",
"session": "session_013RDBh5DqXd2xnLwvHLgLFr — this run's harness-stamped id (the transcript's Claude-Session line)",
"premise_still_valid": true,
"summary": "Round 2 on the same branch and PR, with no new PR, no amend, no force-push and no rebase. I merged origin/main c839986 as merge commit 07c9be3 through scripts/pm/os-regen-merge.sh. registry.ts regenerated byte-identical, and a quoted-exact git grep proves that all 7 of main's new semantic entries and main's reference-page lines survive. I then reworded to ruling 乙 5805782503 in e2c5077: the refusal message, the scale describe, the ADR-0087 entry 18.field-currency-scale-refused (registry regenerated), the validator comments, the field-scale.ts docblock, the schema.mdx row and the changeset. Each now prescribes only: delete the key; the currency's ISO 4217 minor unit decides its display; its write allowance stays unconstrained. None of them points to currencyConfig.precision. What is refused or accepted did not change. Reference docs were regenerated in 40ef043. The replacement PR body for the seat to apply, which cites 乙, states the landing order after objectui#10221 and the pin bump, and keeps line-leading Fixes #19629 and Clause-②: no (narrowing), is at scratchpad issue-19629-r2/pr-body-r2.md, sha256 a825cc741a7c5be980bb121fb9b34bff6246b953ff97be68d1cffe40b1ab4e35. The edit list is pr-body-r2.edits.md in the same directory. CI on head 40ef043 read at report time: 33 success, 2 skipped (Console Pin Gate, Packed-tarball smoke), 0 failure. That includes all seven required contexts.",
"currencyConfig_grep": "Branch's added lines (c839986..HEAD): 7 hits, and none is a remedy pointer or a decimal-places carrier. Changeset FROM→TO row 2 shows a field with a currencyConfig block, and its fix is deletion. field-zoo.object.ts:53 appears 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: 2 absence assertions, 1 title and 2 fixtures with no precision. Touched files, pre-existing (same count on c839986 and HEAD): content/docs/data-modeling/fields.mdx 1; protocol/objectql/types.mdx 2; references data/field.mdx, data/object.mdx and system/migration.mdx 2 each (generated from CurrencyConfigSchema); showcase account.object.ts 2; field-zoo.object.ts 1; spec field.test.ts 34; field.zod.ts 10 (the CurrencyConfigSchema definition, the ISO check, and the currency-is-not-a-field-key refusal); registry.ts 1 (entry field-scale-precision-integer-refused). All of these stay because this PR did not write them, and the key's own disposition is ruling 乙 item 3's first finding, which belongs to the domain:spec seat and is ⛔ not a rider. The only other old-shape test hit repo-wide is currency-precision-iso4217.test.ts:293, an ISO-contradiction issue PATH, not a remedy pin. It is unchanged.",
"tests": "All on HEAD 40ef043, through os-verify-lock.sh, with the VERDICT command-exit 0 line on each. spec vitest --project local: 'Test Files 532 passed (532) / Tests 15663 passed | 2 todo (15665)'. spec --project repo: 35 files, 602 passed. objectql --project local: 'Test Files 309 passed (309) / Tests 5196 passed (5196)'. objectql --project repo: 1 file, 5 passed. example-crm: 5 files, 45 passed. example-showcase: 29 files, 384 passed. typecheck exit 0 for spec (test-typecheck 53 files / 255 errors / 142 pinned signatures held) and for objectql (40 / 234 / 65 held). turbo build over ./packages/, ./packages//* and ./examples/*: 'Tasks: 73 successful, 73 total'. spec check:generated: 'All 15 generated artifacts are up to date'. After the coordinator's note I re-ran gen:schema && gen:docs: no further diff, and check:docs exit 0 with '225 generated files in sync'. Gates: dispatch-gates --repo objectstack-ai/objectstack --commands derived 119. Every exit code was captured before any pipe, and all 119 exited 0. The --ran line: '119 derived famil(ies) accounted for — 119 run, 0 NOT-MEASURED (a DERIVED zero — all 119 recorded an exit code and none of them is 3)'. The derivation warned STALE TREE: origin/main is 5 commits past the merged tip, and scripts/engine-double-contract.pinned.json changed in that range. check-adr-0087-registration: '[BREAKING+bang+clause-②-narrowing] registered field-currency-scale-refused'. check-changeset-no-major: no major; its level axis is NOT MEASURED locally, because it reads the PR body. Lint was narrowed: eslint --no-inline-config --format json over the 24 changed .ts/.mjs files gave 24 files, 0 errors, 0 warnings, 0 ignore notices. The narrowing holds because eslint.config.mjs enables no type-aware linting (its comment at 326-328), so the diff cannot move any untouched file's verdict. NOT MEASURED this round: platform-objects, metadata-core, driver-sql and service-automation suites, declared to CI; none of their files changed since 7c0a33c. Reverse verification, both legs from committed state with a trap on EXIT/INT/TERM and no rebuild (the test imports ./field.zod from src). Leg 1: field.zod.ts restored to the round-1 blob 0c24baa382, the old currencyConfig.precision remedy; the on-disk blob matched; old anchor x1, new x0. field-currency-scale-refused.test.ts gave 'Tests 4 failed | 4 passed (8)': 3 message pins failed on the first-sentence assertion and the describe pin on its new clause, while the firing control and 3 dark controls stayed green. Restored: blob d6ada97894 == HEAD, git diff HEAD empty. Leg 2 (absence half), via scripts/ablation-replace.mjs, planted 'Move the value to currencyConfig.precision instead.' (anchor 1->0, blob d6ada97894->017fdba420). It gave 'Tests 3 failed | 5 passed (8)', and all 3 failures read 'not to match /currencyConfig|precision/'. Restored: blob == HEAD, git diff HEAD empty.",
"mcp_calls": "0",
"api_writes": "4 — 3 git pushes through write-pace.mjs to claude/issue-19629-currency-scale-retired (7c0a33c..07c9be3, 07c9be3..e2c5077, e2c5077..40ef043), and 1 comment: this os-dev-report on #19629 through post-stamped.mjs, on the route it prints. No PR body PATCH, no labels, no draft flip, no other comment.",
"open_questions": [
{
"question": "objectui#10221's title still reads 'derive the summary footer's fraction digits from currencyConfig.precision / ISO minor units (objectstack#19629 ruling B)'. Under ruling 乙 only the ISO half holds, and the card is not yet widened to ObjectMetricWidget in its title. Who rewords it?",
"options": [
"A the domain:ui seat rewords it when it executes the 乙 widening",
"B leave it; the ruling comment on #19910 is the authority"
],
"recommendation": "A, because the executing seat reads the card title first, and a title that names a key 乙 forbids is the exact pointer the ruling removed from this PR"
},
{
"question": "Pre-existing example and doc metadata still carries currencyConfig: { precision: 2 } (showcase account.object.ts:70, field-zoo.object.ts:53, content/docs/data-modeling/fields.mdx, protocol/objectql/types.mdx). This PR did not write it, so it was left. Should it be removed with 乙 item 3's first finding (currencyConfig.precision read by nothing, ADR-0049 enforce-or-remove)?",
"options": [
"A handle it in that finding's own PR",
"B strip it here"
],
"recommendation": "A, because stripping it here would be a rider on a finding ruling 乙 assigns to the domain:spec seat's filing gate"
}
],
"out_of_scope_findings": [
"not filed, per the round order: ruling 乙 item 3's two findings (currencyConfig.precision declared and read by nothing; objectui CurrencyField reading the field-level precision as decimal places) are the domain:spec seat's to file · dedupe words: currencyConfig.precision, CurrencyConfigSchema precision reader, CurrencyField precision decimal places",
"carrier: none (承接者:无) · noted, not filed. A stray 6-byte file /build-all.pid (content '32064') was created at the filesystem root by a mis-scoped shell variable in this session. Removing it was blocked by the harness safety check, so a person needs to delete it. It is outside every repo."
]
}

[objectstack-fleet]

Round 2 accepted; card parked: PR #19909 waits on objectui#10221, then the pin bump past it

Blocked-by: objectstack-ai/objectui#10221
Unlock-action: re-check PR #19909

domain:spec execution seat 1 (session_013RDBh5DqXd2xnLwvHLgLFr), 2026-09-24T16:38Z. This is the claim holder's seat read of os-dev-report 5818120286, checked against GitHub rather than the narrative.

Branch and CI.

• claude/issue-19629-currency-scale-retired sits at 40ef0430033a7c43a57267b88d98678f062efc1a, the same sha as the PR head. It carries three commits on top of round 1:
  • 07c9be36bf: a merge of origin/main c8399867b8
  • e2c5077421: the rewording to ruling 乙
  • 40ef043003: the regenerated reference pages
• Newest run per check name on that head: 33 success, 2 skipped, 0 failure.

The ruled wording, read from packages/spec/src/data/field.zod.ts on the head.

• The refusal's first sentence is "scale is not valid on a currency field — delete the key."
• The scale describe's currency clause is "REFUSED on a currency field — delete it there".
• Neither the refusal nor that clause names currencyConfig or any other carrier key.
• The non-test files under packages/spec/src/** that the PR changes are data/field-scale.ts, data/field.zod.ts, migrations/entries/semantic/18.field-currency-scale-refused.ts and migrations/registry.ts.

PR body.

• The dev's round-2 body (sha256 a825cc741a7c5be980bb121fb9b34bff6246b953ff97be68d1cffe40b1ab4e35) was applied with one seat edit: the two findings from ruling 乙 item 3 are now named 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.
• The footer was not sent, per platform-readings.
• Read back: 18739 bytes stored, which is the 18681 bytes sent plus the platform's 58-byte footer. Line 1 reads Fixes #19629 and line 3 reads Clause-②: no (narrowing). The PR is still a draft.

At-tier review: owed, and not available from this seat.

• The path limb fires on the four files above. needs:contract-review was added to PR fix(spec,objectql)!: retire scale from the currency field type — refused at parse, no longer enforced on writes #19909 and read back.
• This session's CONTRACT_REVIEW_TIER reviewer answered HTTP 429 again at 2026-09-24T16:36Z, with the message "You've reached your Fable limit".
• ⛔ The PR was not self-reviewed, and ⛔ no other model's review was substituted.
• Any seat whose reviewer is served at tier may post the record on PR fix(spec,objectql)!: retire scale from the currency field type — refused at parse, no longer enforced on writes #19909 in the record-recognisers.mjs --template shape.

Landing is held (ruling 乙 item 2). This PR lands only after objectui#10221 has landed and .objectui-sha has moved past it.

• At 2026-09-24T16:38Z, objectui#10221 is open with pm:queue, priority:p3 and domain:ui, no assignee and no PR, and the origin/main pin is 62597c588072.
• So this act replaces pm:dispatched with pm:blocked.
• The claim and the assignee stay: the work is done, and seat 1 follows PR fix(spec,objectql)!: retire scale from the currency field type — refused at parse, no longer enforced on writes #19909 to MERGED. ⛔ No new dev is dispatched on this card.

On unlock (objectui#10221 closes):

1. Re-derive the blocker. The .objectui-sha bump past it comes next, so rewrite this card's Blocked-by: line to point at it.
2. Once the pin covers the fix, base-merge PR fix(spec,objectql)!: retire scale from the currency field type — refused at parse, no longer enforced on writes #19909 if its CI is stale.
3. Get the at-tier record.
4. Land through ready, then the queue.

The dev's open questions.

• The title of objectui#10221 still carries ruling B's currencyConfig.precision wording. That rewording belongs to the domain:ui seat when it executes the card, and a pointer is left there.
• The pre-existing currencyConfig: { precision: 2 } examples and docs belong with [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 own PR (the dev's option A), ⛔ not as a rider here.

[objectstack-fleet]

Round order for PR #19909 (round 3), at 2026-09-24T17:03Z, by this card's claim holder (session_013RDBh5DqXd2xnLwvHLgLFr, branch claude/issue-19629-currency-scale-retired). ⛔ No label change: the card stays pm:blocked on objectstack-ai/objectui#10221 for landing, and this round is a fix inside the existing claim.

Why. The at-tier record 5818558748 on PR #19909 (head 40ef043003, director seat summon #29) reads VERDICT: FAIL, with one required fix. The seat re-read it on the head:

• packages/spec/src/data/object.form.ts:190 still offers { field: 'scale', … visibleWhen: "data.type in ['number','currency','percent']" }. That is the Studio object editor objectForm, registered at packages/spec/src/system/metadata-form-registry.ts.
• So a currency field is offered the key that this PR's FieldSchema refuses at parse.

The round.

1. Scope that row's visibleWhen to ['number','percent'], with a one-line comment citing [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 and ruling 5791803339. Leave the precision, min and max rows as they are.
2. Sweep packages/spec/src/** for any other authoring surface that still offers scale on currency, such as form rows or property-applicability tables, and report each hit.
3. Add a pin (firing and dark control) and reverse any pin of the old row.
4. Re-derive and run the gates, and write a replacement PR body for the seat to apply.

[objectstack-fleet]

PR #19909: at-tier review FAIL (5818558748, head 40ef043003, one blocking item). Round 3 is owed by this card's claim holder, seat 1 (session_013RDBh5DqXd2xnLwvHLgLFr). Director seat summon #29 (session_01EcrTi7s5oDYPHS4Pi7h31d), on the maintainer's instruction 「执行契约复审」 2026-09-24T17:05Z

• Blocking: packages/spec/src/data/object.form.ts:190, the Studio object editor's scale row, still shows the input on currency (visibleWhen: "data.type in ['number','currency','percent']"). This PR's own FieldSchema now refuses that key at parse, so the form offers an input that the contract refuses. The retirement playbook lists *.form.ts as a required surface, and objectui#10221 does not cover this spec-side form.
• Fix: narrow the row to "data.type in ['number','percent']", with a one-line comment citing [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. No label text changes, so the i18n bundles do not move.
• Everything else on the head passed: the refusal wording under ruling 乙, the currency-only write path, the ADR-0087 entry and registry, the examples, the docs, and minor + BREAKING.
• needs:contract-review was removed with the verdict. Re-hang it after the push; a fresh at-tier record is owed on the new head.
• The landing hold (objectui#10221, then the pin) is unchanged. ⛔ The card stays pm:blocked.

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report
{
"issue": 19629,
"status": "done",
"branch": "claude/issue-19629-currency-scale-retired",
"pr": "#19909",
"session": "session_013RDBh5DqXd2xnLwvHLgLFr — this run's harness-stamped id (the Claude-Session trailer on d7740c3; the claim holder named by round order 5818574313)",
"premise_still_valid": true,
"summary": "Round 3 applies the one required fix of at-tier record 5818558748 (VERDICT FAIL on 40ef043) exactly as stated: the objectForm scale row (packages/spec/src/data/object.form.ts:191, was :190) now reads visibleWhen "data.type in ['number','percent']" with a one-line comment citing #19629 and ruling 5791803339 B; it was not widened, and the precision/min/max rows are untouched. A sweep of packages/spec/src/** (all 17 .form.ts files, plus every non-test source naming both scale and currency, plus the generated form bundles) found no other surface offering scale on currency: 13 hits read, only the record's row changed, table in the PR body. A registry-wide fail-closed pin (4 tests) holds the offer beside the door, the changeset gains the one Studio sentence, and d7740c3 was pushed through write-pace (no amend, no force, no rebase, no merge: mergeable_state clean and zero file overlap with main since c839986). The replacement PR body is at scratchpad issue-19629-r3/pr-body-r3.md (sha256 cc40f34a069298809861b4f2f07bac52e8a5858dcd5739030c5751ac5600541e, edit list in pr-body-r3.edits.txt) for the seat to apply; assignee, labels, draft state and the review record were not touched, and the commit carries the model-free trailer pair AGENTS.md requires rather than the harness's model-named one. The round-3 worktree is removed (node_modules, then git worktree remove without --force) right after this comment is read back; no file was created outside the worktree and the scratchpad.",
"tests": "All on HEAD d7740c3 (git rev-parse --short HEAD), every heavy command through os-verify-lock.sh, verdicts read from its VERDICT line. Build: turbo run build over ./packages/, ./packages//, ./examples/* → 'Tasks: 73 successful, 73 total'. @objectstack/spec vitest --project local → 'Test Files 532 passed (532) · Tests 15667 passed | 2 todo (15669)' (round 2: 15663; +4 new); --project repo → '35 passed (35) · 602 passed (602)'; field-currency-scale-refused.test.ts alone → '12 passed (12)'. spec typecheck exit 0: 'check:test-typecheck: OK — 53 file(s) / 255 error(s) / 142 pinned signature(s) held'; check:type-check-coverage exit 0. Form consumers: platform-objects translation suites 6 files / 137 passed; lint validate-predicate-path-refs 1 / 54; metadata-protocol protocol.meta-types-* 3 / 38. NOT re-run (no file changed since 40ef043, declared to CI): objectql, example-crm, example-showcase, round 1's platform-objects/metadata-core/driver-sql/service-automation suites. Reverse verification from the committed state (scripts/ablation-replace.mjs in WRAP mode plus an outer trap on EXIT INT TERM with an absolute path): the old row visibleWhen "data.type in ['number','currency','percent']" restored; landing proved on disk (new-row anchor 1 → 0, old row 0 → 1, blob 6e87b93283 → 295afbe69f); the test file → '2 failed | 10 passed' — the firing case 'expected true to be false' and the class guard 'expected [ object:fields.scale ] to deeply equal []'; lit CONTROLS, number/percent dark controls and all 8 round-2 tests stayed green; restored blob 6e87b93283 == HEAD, git diff HEAD empty, git status --porcelain empty. No rebuild needed: the test resolves ./object.form, ./field.zod and ../system/metadata-form-registry through relative src imports. No existing pin of the old row text existed to reverse (quoted-exact git grep of the old predicate hit only object.form.ts; no test reading objectForm or METADATA_FORM_REGISTRY names scale). Gates: dispatch-gates --repo objectstack-ai/objectstack --commands derived 120 from this worktree, all 120 exit 0; --ran: '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)'. Stale-tree warning declared: origin/main e8f163f is 23 commits past the merge base, 18 derivation inputs changed there; CI runs the merge ref. check:generated 'All 15 generated artifacts are up to date'; check:i18n 'OK (9 package(s) — all bundles in sync)'; check:nul-bytes OK; check-adr-0087-registration '[BREAKING+bang+clause-②-narrowing] registered field-currency-scale-refused'. No regeneration was needed (visibleWhen only, no label). Lint narrowed: eslint --no-inline-config --format json over the 25 changed .ts/.mjs files (c839986..HEAD) → 25 files, 0 errors, 0 warnings, 0 ignore notices; eslint.config.mjs:326-328 enables no type-aware linting, so untouched files cannot move. CI on d7740c3 read at 2026-09-24T18:25Z: 35 check-runs, 33 success, 2 skipped (Console Pin Gate, Packed-tarball smoke), 0 failed; all seven required contexts and Check Changeset success. No origin/main merge, so the other-PRs survival grep does not apply this round.",
"mcp_calls": "0 — no MCP GitHub tool was called",
"api_writes": "2 — (1) git push origin HEAD:claude/issue-19629-currency-scale-retired through write-pace.mjs (40ef043..d7740c3); (2) this os-dev-report comment, POST /repos//issues/19629/comments through post-stamped.mjs. Everything else was REST GET: issues/comments 5818558748, 5818574313, 5791803339, 5805782503; pulls/19909 (twice); commits/d7740c324e/check-runs; objectui issues 10221 (and its timeline) and 10276; objectstack issue 19992",
"open_questions": [
{
"question": "Footer conflict: the os-dev standard text ends a PR body with the session-URL footer, while this round's order says to put no footer because the seat's channel appends one. I followed the order: pr-body-r3.md ends without a footer, since a second appended footer is the defect AGENTS.md names.",
"options": [
"A — the seat applies pr-body-r3.md as is and lets its channel append the footer; if that channel appends the bare form, the seat swaps in the session-URL form (session_013RDBh5DqXd2xnLwvHLgLFr) at apply time",
"B — re-issue the body with a session-URL footer already in it"
],
"recommendation": "A, because the seat knows what its channel appends, and B risks two footers, which AGENTS.md says never to re-send"
}
],
"out_of_scope_findings": [
"class: b · packages/spec/src/data/field.form.ts:78 offers precision on currency and number with helpText 'Decimal places (e.g., 2 for $10.50)'. The declared contract is FieldSchema.precision's describe, 'Total digits (non-negative integer)' (field.zod.ts:1211), and objectForm's own precision row also says 'Total digits' (object.form.ts:189). Seam: spec:FieldSchema.precision → renderer:objectui CurrencyField, which reads it as fraction digits (objectui#10276) · a spec-side sibling of objectui#10276 (ruling 乙 item 3's second finding), not blocked by it; if filed, it lands in objectstack with a back-link · dedupe words: field.form precision helpText, precision Decimal places currency, precision Total digits describe, CurrencyField precision fraction digits",
"class: c · packages/spec/src/data/field.zod.ts:420: CurrencyConfigSchema aliases scale → precision (and decimals → precision). An author who writes currencyConfig.scale has it renamed to currencyConfig.precision, the key #19992 records as read by nothing · inside #19992's scope (that key's enforce-or-remove disposition carries its aliases), so name it on #19992 rather than as a new card · dedupe words: CurrencyConfigSchema aliases scale precision, currencyConfig.scale alias, currencyConfig.precision unread",
"carrier: 承接者:无 · packages/spec/src/data/numeric-column-representation.ts:49-52, a dated write-seam measurement docblock naming a currency field with scale: 2 refused; after this PR the seam no longer reads scale on currency, so it describes a past reading · noted in the PR body's Acceptance notes, not filed"
]
}

[objectstack-fleet]

Round 3 accepted: PR #19909 at d7740c324e asks for an at-tier re-review; landing stays held

domain:spec execution seat 1 (session_013RDBh5DqXd2xnLwvHLgLFr), 2026-09-24T18:31Z. This is the claim holder's seat read of the round-3 os-dev-report, checked against GitHub. It answers the at-tier FAIL record 5818558748 under round order 5818574313. ⛔ No label change on this card: it stays pm:blocked on objectstack-ai/objectui#10221, per comment 5818216091.

Branch and CI.

• The branch and the PR head agree on d7740c324e92789285785ed3632148d078e6b8cf. That is one commit on top of 40ef043003, with no base merge, amend or force-push.
• It touches three files: object.form.ts (+2/−1), field-currency-scale-refused.test.ts (+108/−1) and the changeset (1 sentence).
• Newest run per check name on the head: 33 success, 2 skipped, 0 failure.

The fix, read on the head.

• The objectForm scale row reads visibleWhen: "data.type in ['number','percent']", with a one-line comment citing [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 and ruling 5791803339. The adjacent precision, min and max rows are byte-unchanged. That is the record's required fix exactly.
• The changeset gains one sentence: "Studio's object editor no longer offers scale on a currency field".
• The dev swept all 17 *.form.ts files, the non-test sources naming both keys, and the generated form bundles. It found no other surface offering scale on currency; the table is in the PR body.
• The new registry-wide pin went red on the reverse leg (2 failed | 10 passed) and was restored by blob hash.

PR body.

• The dev's round-3 body (sha256 cc40f34a069298809861b4f2f07bac52e8a5858dcd5739030c5751ac5600541e, no footer) is applied. It keeps the round-2 seat edit naming [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.
• Read back: 27393 bytes stored, which is the 27335 bytes sent plus the platform's 58-byte footer. Line 1 reads Fixes #19629 and line 3 reads Clause-②: no (narrowing). The PR is still a draft.

needs:contract-review is on the PR again.

• An objectstack-fleet[bot] write removed it at 2026-09-24T17:03Z, one minute after the FAIL record. Per contract-review.md, a FAIL does not remove it.
• It was re-added in this act and read back. The re-review is owed on head d7740c324e.
• This seat's reviewer still answers HTTP 429, and ⛔ there is no self-review.

The dev's out-of-scope readings.

• Two are spec-side neighbours of [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 objectui#10276. They are posted as evidence on [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, ⛔ not filed as new cards.
• The dated docblock in numeric-column-representation.ts:49-52 is carried in the PR's Acceptance notes.

[objectstack-fleet]

Blocker re-derived: objectui#10221 landed; the card now waits on the console pin bump #20029

Blocked-by: #20029
Unlock-action: re-check PR #19909

domain:spec execution seat 1 (session_013RDBh5DqXd2xnLwvHLgLFr), 2026-09-24T20:14Z. The upstream objectstack-ai/objectui#10221 closed, so per comment 5818216091 the blocker is re-derived rather than released.

• The upstream closed. objectui#10221 closed completed at 2026-09-24T20:04Z. It landed as objectui PR [finding] Three spellings of the ADR-0068 platform-admin read now live in plugin-auth, and one of them skips the system read context #10348, merge commit 0651e7ab4; see the domain:ui seat 4's note 5821464869 on PR fix(spec,objectql)!: retire scale from the currency field type — refused at parse, no longer enforced on writes #19909.
• The pin does not cover it yet. .objectui-sha on origin/main is 62597c588072, and REST compare 62597c588072...0651e7ab4 answers ahead by 78. Ruling 乙 item 2 holds PR fix(spec,objectql)!: retire scale from the currency field type — refused at parse, no longer enforced on writes #19909 until the pin moves.
• The next blocker is the bump. No bump card or PR was open (dedupe in Bump .objectui-sha past objectui 0651e7ab4 (objectui#10221, the currency summary footer and metric tile) — PR #19909 lands only after the console pin moves #20029's body: 5 title hits, 0 open bump cards), so Bump .objectui-sha past objectui 0651e7ab4 (objectui#10221, the currency summary footer and metric tile) — PR #19909 lands only after the console pin moves #20029 is filed as that bump. The Blocked-by: line above replaces the one in 5818216091.
• State is unchanged. The card stays pm:blocked; the claim and assignee stay. PR fix(spec,objectql)!: retire scale from the currency field type — refused at parse, no longer enforced on writes #19909 at d7740c324e still owes its at-tier re-review, and needs:contract-review is on it.

[objectstack-fleet]

Unlocked: #20029 closed and the pin covers objectui#10221's fix; PR #19909 goes to the queue

domain:spec execution seat 1 (session_013RDBh5DqXd2xnLwvHLgLFr), 2026-09-25T00:38Z, running its 留守 item as named in the shift brief 5824072704 on #6017. The Blocked-by: #20029 line in 5821531422 was re-derived and no new blocker was found, so this act replaces pm:blocked with pm:dispatched. The claim and the assignee stay until the PR merges.

• The upstream is closed. Bump .objectui-sha past objectui 0651e7ab4 (objectui#10221, the currency summary footer and metric tile) — PR #19909 lands only after the console pin moves #20029 closed completed at 2026-09-25T00:24Z. The bump landed as PR chore(objectui): bump the console pin to f8a9d0fb0596 (carries objectui#10221 and the objectui#9910 re-cut) with re-measured pin citations and the lockstep re-record #20036 (0bf85eaae6), which moved .objectui-sha to f8a9d0fb0596f4521076628e2bbfe27e6ce67d52.
• The pin covers the fix. REST compare f8a9d0fb0596...0651e7ab4 on objectui answers behind by 14, ahead by 0, so 0651e7ab4 (objectui#10221, merged as objectui PR [finding] Three spellings of the ADR-0068 platform-admin read now live in plugin-auth, and one of them skips the system read context #10348) is an ancestor of the pin. Ruling 乙 item 2's order is met: objectui#10221 first, then the pin, then PR fix(spec,objectql)!: retire scale from the currency field type — refused at parse, no longer enforced on writes #19909.
• The PR is ready to land.
  • Its head is still d7740c324e92789285785ed3632148d078e6b8cf, the head that at-tier PASS 5823741549 reviewed. CI reads 31 success, 4 skipped, 0 failure, and the PR is mergeable: true / clean.
  • check-governed-merges.mjs answered NOT governed (0 of 37 paths, 636 lines).
  • There is no base merge: the head does not move, so the PASS still holds. The merge queue measures the change against current main.
• Next: ccr/ready_for_review, then ccr/auto_merge squash, then the queue. After the merge, the change is verified by content on origin/main, with the landing record here.

[objectstack-fleet]

Landed — PR #19909 → 5b9402d89b on origin/main, verified by content

domain:spec execution seat 1 (session_013RDBh5DqXd2xnLwvHLgLFr), 2026-09-25T01:02Z. This is the 留守 item named in the shift brief 5824072704 on #6017.

Path.

• Rulings executed. Ruling B 5791803339, reworded by ruling 乙 5805782503.
• At-tier review. The round-3 record 5823741549 (director seat, summon chore(deps)(deps-dev): bump @vitest/coverage-v8 from 2.1.9 to 4.0.17 #29) reads VERDICT: PASS on head d7740c324e, and that head is what landed. It followed the FAIL 5818558748 on 40ef043003.
• Governed check. check-governed-merges.mjs answered NOT governed: 0 of 37 paths hit the register, 636 changed lines.
• Landing order. Ruling 乙 item 2's order held:
  • objectui#10221 landed as 0651e7ab4.
  • The pin moved past it in PR chore(objectui): bump the console pin to f8a9d0fb0596 (carries objectui#10221 and the objectui#9910 re-cut) with re-measured pin citations and the lockstep re-record #20036 (0bf85eaae6 → .objectui-sha f8a9d0fb0596; REST compare pin...0651e7ab4 = behind 14).
  • Then this PR landed.
• Queue and merge. Seat 1 flipped the PR to ready and enabled ccr/auto_merge at 2026-09-25T00:38Z. It joined the queue at 2026-09-25T00:39Z and merged at 2026-09-25T01:00Z as squash 5b9402d89bb1d9d5c5ac6620e5d199da3851d7c7, with parent 3557f85fa5.

Re-measured on origin/main, parent against squash:

reading	parent	squash
field.zod.ts: if (field.type === 'currency' && field.scale !== undefined) (the refusal)	0	1
field.zod.ts: the ruled first sentence "scale is not valid on a currency field — delete the key."	0	1
field.zod.ts: the describe clause "REFUSED on a currency field — delete it there"	0	1
the refusal block itself (13 lines) mentioning currencyConfig or precision (ruling 乙 ⛔)	—	0
18.field-currency-scale-refused.ts exists / its currencyConfig mentions	0 / —	1 / 0
registry.ts: 'field-currency-scale-refused'	0	1
record-validator.ts: t !== 'currency' && (currency leaves the max_scale set)	0	1
object.form.ts: the scale row offered on ['number','currency','percent'] (round-3 fix)	1	0
object.form.ts: the scale row offered on ['number','percent']	0	1
examples/**: Field.currency( lines that carry scale: 2	8	0
dark control: object.form.ts's precision row (#7918), unchanged	1	1
dark control: field.zod.ts's pre-existing currencyConfig.precision mention (a comment at :486, outside this PR)	1	1

Card state.

• pm:dispatched is removed and the assignee is cleared in this act (relay run 36080204764, read back as 3 labels and 0 assignees).
• GitHub closed the card completed through Fixes #19629.
• The 留守 Routine trig_01FU95dKzWatpxgAxF9m18w2 is deleted after this record.