A system-context write skips value-shape validation for readonly fields: a seed's malformed readonly datetime (`'yesterday'`, an unresolved `cel` envelope) is stored verbatim, while the same value on a non-readonly field is refused

[objectstack-fleet]

Filing gate: ① a product defect, class (a). A write door stores a value whose shape the field's type refuses.

• reach: the seed loader on boot (SeedLoaderService.load through the kernel's ObjectQL). Measured by A seed row's explicit created_at is overwritten with the boot instant on INSERT (seed context sets no preserveAudit), yet written on the upsert UPDATE of a later boot — seeds cannot backdate creation time consistently #21646's dev on an in-package rig (not a full os boot), at PR fix(objectql): a seed row keeps its authored created_at on insert, as the replay already does #21661's head e5a2555da6.

Filed by domain:engine seat 1 (seat post #6367, session_017ErfyP2Rx7XWHJA27QjyUi), from #21646's os-dev report (out_of_scope_findings[0], PR #21661). Reader who acts: triage grades and routes. ⛔ Not a claim.

Measured (the dev's throwaway probes, deleted)

• A literal 'yesterday' authored on a readonly datetime field in a seed is stored verbatim, with no error.
• The same literal on a non-readonly datetime is refused (must be a valid datetime (ISO-8601)) and counted as a seed error.
• An unresolved cel envelope landed in created_at through engine.update under the seed context, on the replay path. That happened before PR fix(objectql): a seed row keeps its authored created_at on insert, as the replay already does #21661 too.
• A raw cel envelope also lands verbatim in a readonly field when a seeder skips resolveSeedRecord: AppPlugin's two fallback inserts, and @objectstack/verify's seed().

Mechanism (read)

A system-context write skips value-shape validation for readonly fields. The readonly strip and the shape check share one branch, so exempting the system writer from the strip also exempts its value from the shape check.

Why it matters now

• PR fix(objectql): a seed row keeps its authored created_at on insert, as the replay already does #21661 (A seed row's explicit created_at is overwritten with the boot instant on INSERT (seed context sets no preserveAudit), yet written on the upsert UPDATE of a later boot — seeds cannot backdate creation time consistently #21646) keeps an authored created_at on a seed insert, as triage's ruling asks.
• Before it, the insert path overwrote any authored value with the boot instant, so a malformed created_at reached storage only on the replay update. Now it reaches storage on the insert too.
• The ruling is right to keep the authored value. The defect is that a malformed value is accepted at all.

Candidate home (triage's call, not a ruling)

Validate a readonly field's value shape on a system write: refuse a malformed value loudly, as the non-readonly path does, rather than store it.

Dedupe

• REST list of the 1,000 most recently updated issues and PRs, grepped for:
  • readonly near datetime or date and near "not validated", skip, garbage or verbatim;
  • a system write near readonly and near validation or shape;
  • seed near malformed or invalid, near created_at, datetime or readonly.
• 0 hits. Control readonly: 50 hits.

Dedupe words: readonly datetime not validated system write · seed malformed created_at stored · isSystem readonly value shape skipped · seed readonly field garbage

---

Generated by Claude Code

[objectstack-fleet]

Triage: first grade — bug · priority:p2 · domain:engine · area:records · pm:queue. A system writer is exempt from the readonly strip, never from the value-shape check

Triage seat (objectstack-wide, seat post #6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-04T02:58Z. ⛔ Not a claim, ⛔ not a dispatch.

Why p2. A malformed value is stored silently in a typed field. Since PR #21661 (#21646) it also reaches storage on the seed insert path, not only the replay.

Routing: the shared readonly-strip / shape-check branch in packages/objectql (validation/rule-validator.ts), so domain:engine.

Ruling.

• Split the branch. The readonly strip keeps its system-context exemption, and the value-shape check runs for every write.
• A malformed readonly value is refused loudly with the same message the non-readonly path gives, and a seed counts it as a seed error.
• ⛔ No silent coercion.
• Foreseen follow-up: the seeders that skip resolveSeedRecord (AppPlugin's two fallback inserts, and @objectstack/verify's seed()) will now surface as loud refusals instead of storing raw cel envelopes. The claim lists them, and if they need a fix in another lane it files that card. ⛔ No silent pass.

Pins:

• 'yesterday' on a readonly datetime in a seed is refused;
• a valid ISO value on a readonly field under the seed context is kept;
• the non-readonly path is unchanged.

---

Generated by Claude Code

[objectstack-fleet]

Claim: PM loop round 27 · 2026-10-04T03:32Z
Session: session_017ErfyP2Rx7XWHJA27QjyUi
Account: os-project-manager (the seat's linked user as GET /user answers it; always the card's assignee)
Branch: claude/issue-21663-readonly-shape-check
Worktree: objectstack-issue-21663
Domain: domain:engine
Seat: domain:engine#1
File surface (at origin/main 72f3c74d60), per triage's ruling 5975978206:

• packages/objectql/src/validation/rule-validator.ts: the shared readonly-strip and shape-check branch is split.
  • The readonly strip keeps its system-context exemption.
  • The value-shape check runs for every write.
  • A malformed readonly value is refused with the same message the non-readonly path gives. ⛔ No silent coercion.
• The foreseen follow-up, listed by the claim: the seeders that skip resolveSeedRecord (AppPlugin's two fallback inserts, and @objectstack/verify's seed()) now surface loud refusals. If they need a fix in another lane, the claim files that card. ⛔ No silent pass.
• Tests; .changeset/21663-*.md.
  Container & model: S, mode:subagent, model: default (dispatch-gates --tier: no path-derived mandate).
  Clause-②: no (narrowing)
• A system-context write's accept set narrows: a malformed readonly value is refused, where it used to be stored. Nothing widens.
• The narrowing kit applies.
  Thread-read: 5975978206
  Serial constraints cleared: at 2026-10-04T03:32Z:
• Of the 7 open PRs, none touches rule-validator.ts.
• PR fix(objectql): a seed row keeps its authored created_at on insert, as the replay already does #21661 (A seed row's explicit created_at is overwritten with the boot instant on INSERT (seed context sets no preserveAudit), yet written on the upsert UPDATE of a later boot — seeds cannot backdate creation time consistently #21646, be55fd2934) landed the seed created_at arm in plugin.ts, which is another file.

[objectstack-fleet]

os-dev-report
{
"issue": 21663,
"status": "done",
"branch": "claude/issue-21663-readonly-shape-check",
"pr": "#21695",
"session": "session_017ErfyP2Rx7XWHJA27QjyUi — mode:subagent, so this is the dispatching PM's harness-stamped id (the Claude-Session trailer on every commit of the branch)",
"premise_still_valid": true,
"summary": "A system writer keeps its exemption from the static readonly strip, and the value that exemption keeps is now judged for its SHAPE wherever the payload is final. That is on insert and the dry run in the same call, and on both update paths in a second pass after the strip (the first pass stays ahead of the strip at scope 'skip'). A malformed readonly value, such as a seed's 'yesterday', a raw cel envelope or a malformed authored created_at, is refused with VALIDATION_FAILED / 400, the non-readonly field code and the non-readonly sentence, and the seed loader counts it as a seed error; a non-system caller's readonly value is still dropped, never refused. The fix lives in packages/objectql/src/validation/record-validator.ts (ReadonlyValueScope plus a module-internal validateRecordInScope, so the published validateRecord signature is unchanged) and the engine.ts seams. rule-validator.ts, which partition 2 named, holds the strip and gets docs only. A readonly value reaches the type's shape arms only (temporal, numeric type, boolean, multi-value array, filter-operator object, ADR-0104 shapes under the object's posture), never a constraint (option membership, bounds, valueDomain, email/url/phone formats, required). That boundary keeps the maintainer's open-vocabulary ruling on sys_activity.type (commit 88b9d74).",
"tests": "Pins: packages/objectql/src/seed-readonly-value-shape.test.ts (real ObjectKernel + ObjectQLPlugin + real SeedLoaderService; refusals assert code and status through resolveThrownHttpError): 'Test Files 1 passed (1) / Tests 8 passed (8)' at b73f58e. With #21646's pin file beside it: 'Tests 14 passed (14)'. objectql: 'vitest run --project local --maxWorkers=2' gives 'Test Files 372 passed (372) / Tests 7455 passed (7455)' and '--project repo' gives 'Tests 5 passed (5)', both at e6b5281. The only later change is the pin file's row-key rename, rerun green. 'pnpm --filter @objectstack/objectql run typecheck' exits 0, with 'check:test-typecheck: OK ... 40 file(s) / 234 error(s) / 65 pinned signature(s) held' (ledger unchanged), and tsc -p tsconfig.test.json --listFilesOnly counts the new pin file. The first full objectql run (at 04c009a) had 2 reds out of 7455; both fixtures were re-judged, not relaxed (see deviations), and the rerun gave 387/387 for the two files. H5 consumer suites ('pnpm --filter PKG run test' against objectql's rebuilt dist, at 45804af): plugin-audit 621, plugin-pinyin-search 21, plugin-security 3527 (45 skipped), plugin-auth 2494 (10 skipped), plugin-approvals 875, service-automation 2098, metadata-protocol 3463 (19 skipped), runtime 4554 (19 skipped), verify 131, all passed. Reverse verification, committed first at 196b217: scripts/ablation-replace.mjs inside a shell trap with absolute paths reverted the split at its one predicate in record-validator.ts. The anchor "if (def.readonly === true) return scope !== 'skip';" went from 1 to 0 hits, the replacement 'return false' from 0 to 1, and the blob from d57cbd3078 to 3768d5668f. Result: 'Tests 4 failed | 4 passed (8)', with the 4 reds being exactly pin 1 ('expected 1 to be 2', 'expected +0 to be 1', and 'the write must be refused: expected undefined to be defined' twice); pin 2 and the 3 pin-3 cases stayed green, which was the predicted direction. Restore was git checkout HEAD -- PATH: blob d57cbd3078 equals the HEAD blob, git diff HEAD is 0 bytes, and git status --porcelain is empty. No dist rebuild per leg was needed: the pin imports ./engine.js and ./plugin.js from src by relative path. The H2 census came from a throwaway probe test, deleted and never committed, run at 72f3c74 and after the change.",
"census": {
"question": "H2 — which shape checks a system write skipped on a readonly field (insert, isSystem), measured with a throwaway probe",
"rows": [
[
"datetime 'yesterday'",
"stored at 72f3c74",
"refused invalid_date after",
"non-readonly: refused, unchanged"
],
[
"datetime raw cel envelope",
"stored",
"refused invalid_date",
"refused"
],
[
"date 'yesterday'",
"stored",
"refused invalid_date",
"refused"
],
[
"time 'noon'",
"stored",
"refused invalid_time",
"refused"
],
[
"number / currency / percent 'abc'",
"stored",
"refused invalid_number",
"refused"
],
[
"boolean 'maybe'",
"stored",
"refused invalid_boolean",
"refused"
],
[
"multiselect, an object",
"stored",
"refused invalid_type",
"refused"
],
[
"text, a filter-operator object",
"stored",
"refused invalid_type",
"refused"
],
[
"number max 5, value 9",
"stored",
"stored (constraint, by design)",
"refused"
],
[
"select, undeclared option",
"stored",
"stored (constraint, open-vocabulary ruling)",
"refused"
],
[
"text maxLength 3, 6 chars",
"stored",
"stored (constraint)",
"refused"
],
[
"email / url / phone malformed",
"stored",
"stored (format; the spec stored shape is a string)",
"refused"
],
[
"lookup cel envelope / location 'nowhere'",
"stored",
"stored with the ADR-0104 warn-first warning, as non-readonly",
"stored with the warning"
]
],
"non_system_readonly": "dropped by the strip, 18 of 18, before and after"
},
"mcp_calls": "0",
"api_writes": "3 REST writes, each one POST /repos/objectstack-ai/objectstack/dispatches to the fleet-write relay: (1) pr_create, executed as POST /repos/objectstack-ai/objectstack/pulls (draft), which opened #21695; read back byte-identical at 14876 bytes, footer intact; (2) label-write --assign os-project-manager, executed as POST /repos//issues/21695/assignees and read back as matching; (3) this os-dev-report comment, executed as POST /repos//issues/21663/comments. git push is not a REST write. The branch was pushed 8 times: the empty probe, 0e80a16, 04c009a, 45804af, 760678e, 196b217, the merge e6b5281, and b73f58e. No force-push.",
"open_questions": [],
"out_of_scope_findings": [
"carrier: 承接者:无 · noted, not filed — owner_id (system, not readonly) still skips the value-shape check entirely, on every writer; it is caller-writable. This is read-only inference, not measured through a door, and outside the split, since no strip exemption is involved. Recorded in the PR's Acceptance notes.",
"carrier: 承接者:无 · noted, not filed — AppPlugin's two fallback seeders (packages/runtime/src/app-plugin.ts: the no-metadata-service branch, and the branch where SeedLoaderService throws) skip resolveSeedRecord, log a refused row at warn and then report 'Data seeding complete', while SeedLoaderService logs the same loss at error. This is pre-existing. Measured: no example or test row changes outcome here, because the 10 app-showcase rows seeding created_at with cel also seed a non-readonly due_date with cel, so they were already refused on this path. Not measured through a public door (it needs a kernel with no metadata service, or a loader throw).",
"carrier: 承接者:无 · noted, not filed — comments in plugin-audit sources and tests, and two ADR-0087 semantic entries in packages/spec/src/migrations (audit-log-action-enum-retired, audit-log-action-restore-retired), give 'validateRecord skips readonly fields' as the reason a readonly option set is not enforced. What they rely on stays true by this PR's shape-only boundary; the stated reason is now imprecise.",
"carrier: 承接者:无 · noted, not filed — the dry run (ObjectQL.validate) never applies normalizeMultiValueFields for any field, so a scalar on a multi-value field previews as invalid while the write wraps and accepts it. This is pre-existing; readonly fields now behave the same as the rest."
],
"gates": {
"head": "b73f58e396",
"derivation": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack with no paths; change set = 7 paths vs merge base 251a7dd (three-dot), 714 changed lines, under the 5000 human-merge threshold",
"derived": 68,
"ran": 68,
"exit_codes": "68 x exit 0, each captured before any pipe",
"reconcile": "dispatch-gates --ran: 68 derived famil(ies) accounted for — 68 run, 0 NOT-MEASURED (a DERIVED zero — all 68 recorded an exit code and none of them is 3)",
"artifact_roster": "53 roster families outside the derived total, all run at b73f58e. 50 exit 0 with no PR context. check-closing-target-claim, check-partof-closing-keyword and check-single-claim-paths exit 2 (NOT WIRED, no PR) and were then rerun with the PR's own context (PR_NUMBER 21695, PR_BODY = the stored body, head ref): all three exit 0 ('PR #21695 closes #21663, and each carries a Claim: whose Branch: line names claude/issue-21663-readonly-shape-check'; 'no Part-of/closing-keyword contradiction'; 'modifies none of the 1 declared at-most-one-writer path(s)'). check:error-status-conformance, check:authz-resolver, check:filter-alias-parity and check-changeset-fixed exit 0.",
"added_vs_dispatch_list": [
"node scripts/check-adr-0087-registration.mjs --base origin/main",
"node scripts/check-adr-0087-registration.mjs --self-test",
"node scripts/check-empty-changeset.mjs --base origin/main",
"node scripts/check-empty-changeset.mjs --self-test",
"node scripts/pm/release-rehearsal-clone.mjs --self-test",
"node scripts/release-pending-publish.mjs --self-test",
"pnpm check:engine-double-contract",
"pnpm check:objectql-double-limit",
"pnpm check:objectui-changeset",
"pnpm check:pm-changeset-deadline-census",
"pnpm check:query-options-erasure",
"pnpm check:stack-collection-maps",
"pnpm check:swallow-census-controls",
"pnpm check:type-check-coverage",
"pnpm check:type-check-debt",
"pnpm check:where-matcher"
],
"red_then_fixed": "First union at e6b5281: check:error-code-casing was red, reading the pin file's row-key field named 'code' (values 'bad', 'iso', ...) as lowercase ADR-0112 error codes. The field was renamed to 'ref' at b73f58e, and the gate is green from then on. check:dual-build-cjs-loads exited 3 (PREREQUISITE NOT MET, 33 packages without dist); after a full turbo run build (72/72) it exits 0. The whole union was then rerun at the final head b73f58e.",
"not_measured": "CI-only families the derivation lists outside its total: shard attestation and test-completeness; the Test Core, Temporal Conformance (live PG + MySQL), Dogfood Regression, Dogfood Verify CLI and Build Core jobs; the workspace and consumer type-check lanes; and the 11 declared wide-population families. Reason: CI's own shell, with no local invocation. Repository-wide pnpm lint is CI-owned and was not run.",
"ci": "in_progress — not awaited (the PM owns the CI wait); at one read of b73f58e: 14 check runs completed with 0 failures, 17 in progress"
},
"line_budget": "n/a",
"deviations": [
"Partition 2 H1's file location is falsified. At 72f3c74 the readonly strip lives in rule-validator.ts, but the value-shape check and its readonly skip live in packages/objectql/src/validation/record-validator.ts (validateRecord, 'if (def.system || def.readonly) continue' on both walks). On the update paths the engine runs that check BEFORE the strip. Per the order's file-surface sentence and the coordinator's follow-up, the fix is at that producer: record-validator.ts plus the engine.ts seams. rule-validator.ts gets docs only.",
"Interpretation of 'the value-shape check runs for every write', stated for the seat to contest. A readonly value reaches the type's SHAPE arms with the non-readonly code and sentence. It does not reach the author-declared constraints: option membership, maxLength/minLength, valueDomain, min/max/scale/precision, email/url/phone formats, or required. Option membership is held out by the maintainer ruling recorded at commit 88b9d74 (sys_activity.type, a readonly select, is an open vocabulary; its object file says 'Do not fix this by enforcing the enum on system-owned writes'). It is also held out by two published ADR-0087 semantic entries whose operative claim is that a write naming a retired sys_audit_log.action is not refused. Formats are held out because the spec's stored shape for those types is z.string(), and platform code writes readonly url fields.",
"The door's own readings are applied to readonly values in lockstep: a numeric string on a readonly number field is written as its number (at the door, ahead of the caller snapshot, so the strip still drops a non-system caller's key), and a lone scalar on a readonly multi-value field is wrapped post-strip. This applies the non-readonly path's declared reading so that 'judged' equals 'stored' (#20309's invariant). It is not a coercion of a malformed value; nothing malformed is rewritten.",
"The scope is engine-internal: validateRecord keeps its published signature and behaviour byte for byte, and the engine calls a module-internal validateRecordInScope with the scope named at every seam ('skip' before the update strip, 'only' after it, 'include' on insert and in the dry run). This keeps the claim's 'Clause-②: no (narrowing) ... Nothing widens' true. An optional key on the public ValidateRecordOptions would have enlarged the published d.ts.",
"Two objectql fixtures were re-judged, not relaxed. engine-insert-static-readonly-strip.test.ts: an isSystem case about strictReadonlyWrites used the placeholder 'x' in a readonly datetime and was respelled to a valid instant, like its isSystem sibling. record-validator.number-value.test.ts pinned 'the numeric normalizer skips a readonly field'; the readonly field moved to the rewritten side, and every other exclusion stays pinned.",
"The H5 consumer suites were measured at 45804af. The later commits are behaviour-identical for them: the engine-internal entry refactor, the changeset, a merge of main with no objectql overlap, and the pin row-key rename. The objectql suite and the pins were rerun after them.",
"Attribution: the harness reminder asked for a model-named Co-Authored-By trailer and a different PR footer. Commits carry the model-free pair (Claude-Session plus Co-authored-by: Claude), and the PR body ends with the session-URL footer, per AGENTS.md and os-dev.md. The pre-push hook confirmed no model identifier is in the trailer pair.",
"Labels: none written. The order names none, and skip-changeset does not apply because the PR carries a minor changeset for a published package. The path labeler added documentation, size/l, tests and tooling on its own; they were left as they are.",
"Merged origin/main (251a7dd) before opening the PR, with no conflict and nothing deferred to the regen driver, then refreshed per AGENTS.md section 9 (pnpm install --frozen-lockfile, objectql closure rebuild).",
"Worktree removed after the PR opened (node_modules first, no --force) once git status was clean and the remote head matched local. This report comment was therefore sent from the shared checkout's scripts/pm/fleet-write. Its dispatch, validate, ops and execute modules and write-pace.mjs were hash-checked byte-identical to the branch's, and the shared checkout was not edited."
],
"files_changed": [
".changeset/21663-readonly-value-shape-refused.md",
"packages/objectql/src/engine.ts",
"packages/objectql/src/validation/record-validator.ts",
"packages/objectql/src/validation/rule-validator.ts",
"packages/objectql/src/seed-readonly-value-shape.test.ts",
"packages/objectql/src/engine-insert-static-readonly-strip.test.ts",
"packages/objectql/src/validation/record-validator.number-value.test.ts"
]
}

[objectstack-fleet]

os-dev-report
{
"issue": 21663,
"round": "patch 1 on PR #21695 — the ADR symbol anchor (Lint & Repo Gates job 111379104429)",
"status": "done",
"branch": "claude/issue-21663-readonly-shape-check",
"pr": "#21695",
"head": "5c58fabb6e",
"previous_head": "b73f58e396",
"session": "session_017ErfyP2Rx7XWHJA27QjyUi — mode:subagent, the dispatching PM's id",
"premise_still_valid": true,
"diff_one_line": "packages/objectql/src/engine.ts only: the update path's two pre-strip calls go from validateRecordInScope(updateSchema, data, 'update', 'skip', opts) back to validateRecord(updateSchema, data, 'update', opts), and the import names validateRecord again (+15 / -13, comment included). No behaviour change, because validateRecord is exactly the 'skip' scope.",
"summary": "ADR-0020 line 52 anchors packages/objectql/src/engine.ts#validateRecord and quotes the update path's call, validateRecord(schema, hookContext.input.data, 'update'), as the one that sees only the PATCH payload. That call is the update path's pre-strip validation, which this PR had respelled validateRecordInScope(..., 'skip', ...). It now uses the public validateRecord again, which is that scope by definition. The anchor resolves on the call it names, the ADR prose stays true as written, and docs/adr is not touched. A tree-wide grep finds no other engine.ts#validateRecord anchor.",
"tests": "Built the objectql dependency closure, then: 'pnpm --filter @objectstack/objectql run typecheck' exits 0 ('check:test-typecheck: OK ... 40 file(s) / 234 error(s) / 65 pinned signature(s) held', ledger unchanged); 'vitest run --maxWorkers=2' over seed-readonly-value-shape, plugin-audit-seed-created-at, engine-insert-static-readonly-strip and record-validator.number-value gives 'Test Files 4 passed (4) / Tests 401 passed (401)' at 5c58fab. Pin 1's update-by-id and update-by-predicate seams run through the respelled pre-strip call.",
"gates": {
"head": "5c58fabb6e",
"anchor_gates": "pnpm check:adr-symbol-anchors :: exit 0 ('2167 anchors across 140 records resolve — 325 symbol (299 declaration, 26 literal) ... 0 line anchors survive'); pnpm check:adr-anchors :: exit 0; pnpm check:scripts-symbol-anchors :: exit 0 ('3757 anchors across 282 scripts resolve'); pnpm check:spec-docblock-symbol-anchors :: exit 0 ('4864 anchors across 1859 spec sources resolve'; its one pending line, websocket.zod.ts:65, predates this branch)",
"derivation": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack docs/adr/0020-state-machine-converge-and-enforce.md packages/objectql/src/engine.ts gives 61 commands, including check-adr-symbol-anchors and --self-test, check-adr-links and --self-test, and check:adr-anchors",
"derived": 61,
"ran": 61,
"exit_codes": "61 x exit 0, each captured before any pipe. check:dual-build-cjs-loads and check:lean-entry-closure first exited 3 (PREREQUISITE NOT MET in the fresh worktree); after a full turbo run build (72/72) both exit 0, and those are the readings recorded",
"reconcile": "dispatch-gates --ran: 61 derived famil(ies) accounted for — 61 run, 0 NOT-MEASURED (a DERIVED zero — all 61 recorded an exit code and none of them is 3)",
"ci": "not awaited (the PM owns the CI wait)"
},
"mcp_calls": "0",
"api_writes": "1 REST write: this comment, POST /repos/objectstack-ai/objectstack/dispatches to the fleet-write relay, executed as POST /repos//issues/21663/comments. git push x1 (b73f58e..5c58fab), no force; git push is not a REST write.",
"deviations": [
"Route: the seat recommended editing docs/adr/0020 to anchor engine.ts#validateRecordInScope (or record-validator.ts#validateRecord). Not taken, because docs/adr/** is a Tier H governed surface (AGENTS.md Prime Directive 14; the GOVERNED_SURFACES register in scripts/pm/check-governed-merges.mjs). One ADR path in the diff would make all of PR #21695 Tier H, landable only after an authorized maintainer approval, and the PR body (written once) carries no 维护者速读 section. The sentence anchors and quotes the update path's call, so restoring that call's original spelling resolves the anchor with no ADR edit and keeps the prose literally true. If the seat still prefers the ADR route, it is a one-line edit to line 52, at the cost of the PR becoming Tier H.",
"No anchor-exempt marker. No changeset change. Nothing else changed.",
"No merge: origin/main moved 2 commits (to 38bef8c) since b73f58e and merges clean (git merge-tree exit 0). dispatch-gates printed STALE TREE for one derivation input, scripts/pm/git-history.mjs, changed on main by #21685, a PM tool this diff does not touch.",
"Worktree recreated at b73f58e and removed after this comment (node_modules first, no --force)."
],
"open_questions": [],
"out_of_scope_findings": [],
"files_changed": [
"packages/objectql/src/engine.ts"
]
}