docs(spec): re-anchor the dead tracker citations in migrations/entries to the commits that decided them (stage 10)

[objectstack-fleet]

Part of #20234

Clause-②: no

Stage 10 of the dead-citation sweep: the migration registry's hand-written entries. Every tracker number in packages/spec/src/migrations/entries/** that no longer exists on the board now cites the commit that decided it, in ruling C+D form C (ruling 5749154545 on #19123). Where the number alone carried the meaning, the line now says what was decided. That is 131 sites over 21 numbers. All of them are comments; the entries' string literals carry no dead number. migrations/registry.ts moves only by gen:migration-registry. No entry id, literal, order, conversionIds or code token moves, and no live citation is removed.

Boundary

• In: all of entries/**. The gate's census reads 117 sites there, under the claim's ~120 slicing threshold, so there is no slice. My raw walk finds 14 more comment sites that the census does not count (see Census below), which gives 131 in total. They are the same dead card on the line after a counted site, plus one README line, so they are rewritten with it.
• Regenerated only: migrations/registry.ts, 128 lines. spec-changes.json and docs/protocol-upgrade-guide.md do not move, because entry comments are never projected into them. Both check: scripts pass with no regeneration.
• Out, per the claim: registry.ts's hand-written parts. They still hold 13 dead sites, listed under Acceptance notes. The six 18.*-unit-in-key.ts entries that PR chore(objectui): bump the console pin to db11afd4967c (carries objectui#11119, the objectui#11105 fix) #20706 edits carry no dead site, and I did not touch them. I did not touch conversions/ or other lanes' sites.

The anchors (one per number, reused from earlier stages where they anchored the same number)

number	sites	anchor	what the rewritten line says it decided
#13135	26 (13 are re-charter #13135)	commit 9e0ba21	retires the paper metadata-customization protocol; re-charter of #12057, which stays
#8495	23	commit 4bfe1a5 (PR #8666 kept)	the precedent: the first 17.x-line narrowing registered under protocol 18 (its own second commit says so)
#8715	16	commit 2c86fe3	the ApiKeySchema retirement; its message names the "route 3" kit (no carrier key, no tombstone, no D2)
#14691	11	commit b3a63d3	retires the ten inert RestServerConfig keys
#10724	11	commit be21955	retires the nine dead contributes members
#14369	10	commit a3d5724	the liveness census that recorded the 15 dead rows
#10485	6	commit 35ad101	retires the themes carrier and ThemeSchema
#11846	5	commit 0c2334f	retires preview mode; its own text carries the ruling record (#12428 kept)
#14676	5	commit 13c48c2	retires connector.errorMapping
#11332	3	commit dce5cd4	retires the manifest's three dead containers
#6361	2	commit 90bbf25	retires the notification-list cursor on both halves
#10627	2	commit be21955	that commit records the controlled monorepo census
#14365	2	commit f60ab90	the open z.partialRecord proposal that commit's changeset recorded; the retirement leaves no record to reshape
#14526	2	commit db16b94	the landing of the client envelope convergence (anchor block, and the README's measured case)
#6363	1	commit 17d0954	"ruled jointly with the unreadCount fix"
#6239	1	commit f549a0d	the ViewProtocol retirement sweep
#10726	1	commit bc56e18	retires contributes.routes (Option B)
#10812	1	commit be21955	"the cloud census", whose 2026-08-24 reading @5b5925a that commit's message records
#9041	1	commit d491625	the url-branch refinement (#9147, live, stays)
#14996	1	commit db16b94	the ADR-0087 registration, which landed in the same squash (its body: "Registration requested on #14996")
#14312	1	commit e944fdb	the oauth.* binding, which left applications.delete out as a behaviour change (PR #15445 kept)

Two lines change without a number, so the sentence still reads: patterns' follow-on line and the oauth anchor block's continuation line. In total 133 lines are removed and 133 added in 86 files, and every file is balanced.

Verification record (final head 1ee5841c09; base fbec216e2d)

Census (the gate's own check-issue-citations.mjs --census --json, board enumerated, 186 pages):

subtree	base (00:21Z, frontier #20740)	head (01:18Z, frontier #20743)
entries/retired-keys	73	0
entries/retired-defs	40	0
entries/semantic	4	0
registry.ts, generated regions	114	0
registry.ts, hand-written parts	2	2
chain.ts, types.ts, index.ts, spec-changes.ts, tests	0	0
migrations/	233	2
packages/spec/src	235	4

• Site-set difference: 254 sites are only at base. 231 are this diff's (117 plus 114 copies). The other 23 are plugin-audit's, from main's docs(plugin-audit): re-anchor the dead tracker citations to the commits that decided them #20737. 0 sites are only at head.
• Kind (every census site is a comment): a TypeScript 6.0.3 walker classes all 131 entry sites as comments. It also finds 13 dead string sites, all in registry.ts's hand-written rationale (see Acceptance notes).
• Probe: REST issues/N without following redirects, over all 302 distinct in-repo numbers of 100 or more in migrations/. 277 answer 200 and 25 answer 404. Controls were read at start, every 50 and end: lit packages/spec/src: 1,277 comment lines still cite 170 deleted tracker numbers (1,295 sites) — the staged remainder of ruling C+D on #19123, measured by PR #20226 #20234 200 (8/8), dead Does a DEACTIVATED sys_position still receive sharing-rule shares? expandPositionUsers never reads the catalogue row, so today it does #8710 404 (8/8).
• Blind spots: the census does not count 14 of the 131 entry sites.
  • 13 are re-charter #13135, because NON_CITATION_HEADS reads re-charter #N as an ordinal.
  • 1 is in entries/README.md, which is outside the gate's packages/**/src/**/*.ts surface.
• Numbers: no tracker number is added. The only numbers on added lines are the live ones kept from the same lines: feat(spec): refuse ${…} placeholder syntax in memory persistence.path / persistence.key at publish (#8495) #8666, [finding] spec ships a three-layer metadata-customization protocol (MetadataOverlaySchema, field-level change tracking) with zero consumers — published reference docs describe an unimplemented customization architecture #12057, additionalTypes on MetadataPluginConfig is authorable, documented as THE way a plugin declares a metadata type, and has no reader — a plugin cannot declare a kind at all #8586, [finding] HotReloadManager.startWatching watches nothing and logs "File watching started" at info; watchPatterns has no reader and watchHandles is never populated #12428, Composed-branch twin of #9041: external.credentialsRef bound + discrete mongo fields naming no username is the same silent no-op, unrefused #9147 and fix(client)!: bind the oauth.* family to the wire shapes better-auth sends #15445. The diff-scoped gate judges them: "every citation this change adds resolves".

Residue (scratch walker over the TypeScript parser, per file, base vs head, 86 .ts files):

• The file with every comment range cut out, everything else byte for byte, is IDENTICAL.
• The leaf-token stream is IDENTICAL: 38,417 tokens, aggregate 67c1f6db944e93ed.
• Controls mutate the head text in memory only, 259 of 259 as expected:
  • Expected identical: a comment insertion in every file.
  • Expected to differ: a string-literal edit, a template-literal edit, a regex-literal edit (where the file has one) and an appended declaration.

Regeneration (gen:migration-registry):

• check:migration-registry exits 1 before and 0 after.
• The registry diff is 128 lines out and 128 in. As (old, new) pairs they equal the entry diff's pairs, re-indented by four spaces.
• 0 changed lines fall outside the generated regions.
• 5 entry pairs are deliberately not carried: the README line, and the semantic "Anchors" header lines, which sit above a blank line and so are not in the carried comment run.
• check:spec-changes and check:upgrade-guide exit 0 with no regeneration.

Build and tests (under os-verify-lock):

• Build: turbo run build over ./packages/* and ./packages/*/*, 71 of 71, at 845e90fea4 and again at 1ee5841c09.
• check:generated: all 15 artifacts up to date, at both heads.
• spec local project: 576 files, 16,991 passed and 1 todo, at both heads.
• spec repo project: 41 of its 45 files, 666 passed, at both heads.
• spec typecheck: exit 0; 53 files / 251 errors / 138 pinned signatures held.

Gates: dispatch-gates --commands derives 82 gates, and the derivation is identical before and after the second merge. At 1ee5841c09 all 82 exit 0. --ran reconciles: 82 derived, 82 run, 0 NOT-MEASURED, a derived zero.

Lint (a proven narrowing):

• eslint --no-inline-config --format json over the 86 touched .ts files: 86 files, 0 errors, 0 warnings.
• isPathIgnored is false for all 86.
• eslint.config.mjs:327-328 states that type-aware linting is never enabled, so a comment edit cannot move an untouched file's verdict.

Changeset: patch.

• In the built dist, the new wording (for example "the precedent of commit 4bfe1a5, PR feat(spec): refuse ${…} placeholder syntax in memory persistence.path / persistence.key at publish (#8495) #8666") appears in dist/migrations/index.js and .mjs, and the old wording appears in 0 files.
• Control: an unchanged phrase appears in the same 2 files.
• So the published @objectstack/spec/migrations entry carries these comments.

Merges: origin/main was merged twice through scripts/pm/os-regen-merge.sh. Neither merge left a regeneration to commit.

• Since then, main has advanced to 97005aed04, and none of those commits touches packages/spec.
• Driver-free merge-tree probes (from a bare --shared scratch clone with no merge.* config) exit 0 against that main and against PR chore(objectui): bump the console pin to db11afd4967c (carries objectui#11119, the objectui#11105 fix) #20706's head d8bac3877d.

Acceptance notes

• The next stage for this card: 13 dead sites left in registry.ts's hand-written parts. They are outside this claim, which says registry.ts is regenerated only.
  • Comments (form C), 2:
    • :5940 [finding][spec] memory-driver persistence.path / persistence.key still accept unresolved ${…} placeholders — the #8336 shape one surface over, deliberately outside its connection-material class #8495, the step-18 doc comment on the persistence placeholder refusal.
    • :22819 [spec] View-Management 五方法(listViews/getView/createView/updateView/deleteView)是零实现零路由的声明面 —— 已导致 #5948 把它的响应 schema 误当成线上路由的契约 #6239, the RETIRED_DEFS_BY_MAJOR[17] doc comment.
  • Author-shown rationale strings (form D), 11:
    • Step 17's :344 cli/runtime: explicit driver(--database-driver / OS_DATABASE_DRIVER)在 CLI 与 standalone stack 之间仍有三处分叉(#6265 后续) #6345. It projects into docs/protocol-upgrade-guide.md:68, so that stage regenerates the guide.
    • Step 18's fragments: :5121 integration/ErrorMappingConfig + ErrorMappingRule are 11 authorable keys with zero consumers — and one of them is named userMessage, colliding with the live #9934 channel #14676, :5455 Narrow the per-option default key OUT of the form-view options vocabulary — the object-field face keeps it (enforced, #7246), the form-view face parses-and-ignores it (ruled on objectui#6263) #12868, :5526 spec(data): mapping lookup transform params (object / fromField / toField / autoCreate) are authorable and read by nothing #10329, :5544 [finding][spec] memory-driver persistence.path / persistence.key still accept unresolved ${…} placeholders — the #8336 shape one surface over, deliberately outside its connection-material class #8495, :5555 Retire the paper metadata-customization protocol with its full coupling set — spec module, authorable customizationPolicies/mergeStrategy keys, metadata API §5, optional IMetadataService overlay methods, unreachable packages/metadata limb (re-charter of #12057) #13135, :5742 Retire the nine dead members of the plugin-manifest contributes block (events/menus/themes/translations/actions/drivers/fieldTypes/functions/commands) — tombstone route, kinds survives #10724, :5745 Plugin-manifest contributes.themes declares a theme-contribution channel nothing reads — same declared-but-unenforced shape as the retired stack themes carrier #10627, :5752 contributes.routes has zero readers, yet a published skill, an ADR row and the dispatcher's own doc all present it as the way to serve a code-handler endpoint — enforce-or-remove fork #10726, :5799 defineStack({ themes }) is parsed, ingested and never applied — the last hop from a stored theme item to the theme engine does not exist #10485, :5816 i18n: component-translation submitLabel copy key lost its only declared carrier when element:form retired (#9249) — decide retire vs re-anchor #10926. These are not projected into either artifact yet.
• Census blind spots, measured here and not filed (a coverage boundary, not one of the three filing classes; new gates default to no):
  • NON_CITATION_HEADS skips re-charter #N even where N is a tracker card.
  • .md files under packages/**/src are outside the surface.
• NOT MEASURED locally, left to CI:
  • scripts/build-schemas-check-mode.test.ts hit the foreground cap alone (exit 124 after 560 s, no output). It parses registry.ts's source; the residue check above shows that source's code and string tokens are unchanged.
  • The other three heavy repo-project files (def-key-collisions, publish-smoke-boot-failure, publish-smoke-port-collision) were not run, as in stages 8 and 9.
• Hypothesis 6's baseline half is falsified by construction. scripts/doc-authoring-prose-id.baseline.json has no packages/spec row, because its leg excludes packages/spec. check:doc-authoring reads strings only, so a comment-only diff cannot move it. It read 808 sites across 229 files at 845e90fea4 and 794 across 227 at 1ee5841c09. The difference is main's own re-anchor commits; this diff touches no row.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/spec, touching 2 documentable anchor(s). ⚠️ 86 changed file(s) yielded no anchor (packages/spec/src/migrations/entries/README.md, packages/spec/src/migrations/entries/retired-defs/17.api__CreateViewRequest.ts, packages/spec/src/migrations/entries/retired-defs/18.api__CrudEndpointPattern.ts, …), so the pages documenting them are NOT COVERED by this run — this is not a clean bill of health for those files.

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

• content/docs/kernel/cluster.mdx (via RETIRED_DEFS_BY_MAJOR (symbol, a top-level const object))

What this run could not see

• 86 changed file(s) yielded no anchor (packages/spec/src/migrations/entries/README.md, packages/spec/src/migrations/entries/retired-defs/17.api__CreateViewRequest.ts, packages/spec/src/migrations/entries/retired-defs/18.api__CrudEndpointPattern.ts, …) — pages documenting those are invisible to this run
• 1 name(s) were too generic to anchor anything (single lowercase words)
• 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 97005aed04a069abc2d2fa82fd6c1594aebd7cd1 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 3d26556ab74c39abd96cd3056f0605693de27cd3 — the merge of head 1ee5841c092cdc7d2c446ab1e7cf933b940e3e25 into base 97005aed04a069abc2d2fa82fd6c1594aebd7cd1, 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 3d26556ab74c39abd96cd3056f0605693de27cd3 && git checkout 3d26556ab74c39abd96cd3056f0605693de27cd3
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 97005aed04a069abc2d2fa82fd6c1594aebd7cd1 1ee5841c092cdc7d2c446ab1e7cf933b940e3e25 && git checkout -B drift-repro 97005aed04a069abc2d2fa82fd6c1594aebd7cd1 && git merge --no-ff 1ee5841c092cdc7d2c446ab1e7cf933b940e3e25

node scripts/docs-audit/affected-docs.mjs --json 97005aed04a069abc2d2fa82fd6c1594aebd7cd1

⚠️ 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 97005aed04a069abc2d2fa82fd6c1594aebd7cd1 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: 1ee5841c092cdc7d2c446ab1e7cf933b940e3e25
Local-runs: probe — one scratch script over the repo's git objects, tokenising the merge-base and head blobs of the 86 changed .ts files with a TypeScript scanner (comments and whitespace dropped) to settle comment-only; nothing built, tested or re-run.

Read: card #20234, body and all 41 comments (the C+D path decision 5856637615, the pointer 5858331362, stage 8's landing 5896689039, stage 9's ACCEPT 5899135835 and landing 5899597998, the stage-10 claim 5901544565, the dev report 5902687602); PR #20750, its body, its one comment, its 88-file list and the net diff against main at the merge base 01e78dceef; the 21 anchor commits by git show and git log; the check-runs on the head, read once.

① Derived judgments

The claim says no accept-set or public-surface move. Each judgment below is from the diff and the git objects, not the dev's prose.

• Comment-only, proven. The token probe finds the non-trivia token streams of all 86 changed .ts files (85 entries plus registry.ts) identical between merge base and head, and its control (one string literal mutated in memory) is detected. So no entry id, verdict, order, conversionIds, string, template, regex or code token moved. Independently, every one of the 133 changed entry lines (132 .ts pairs plus one README line) and every one of the 128 changed registry lines begins with a // marker. RIGHT.
• registry.ts moves only in its generated regions. The os-generated markers stand at the same twelve lines at base and head. Mapping each of the 256 changed registry lines to a region: semantic:18 two, retired-key:17 four, retired-key:18 one hundred and fifty, retired-def:17 two, retired-def:18 ninety-eight, outside any region zero. The 128 registry (old, new) pairs equal the entry .ts pairs re-indented by four spaces, with zero registry pairs lacking an entry source; the five entry pairs not carried are the README line and the four semantic anchor-block header lines the generator does not copy. No hand-written part is edited. RIGHT.
• Scope. 88 files: one new changeset plus 87 under packages/spec/src/migrations/ (the entries README, 85 entries, registry.ts). None of the six entries/semantic/18.*-unit-in-key.ts files that PR chore(objectui): bump the console pin to db11afd4967c (carries objectui#11119, the objectui#11105 fix) #20706 edits is touched. spec-changes.json and docs/protocol-upgrade-guide.md do not move. Nothing under conversions/. RIGHT.
• The anchors are true. Each of the 21 dead numbers answers 404 today (controls: packages/spec/src: 1,277 comment lines still cite 170 deleted tracker numbers (1,295 sites) — the staged remainder of ruling C+D on #19123, measured by PR #20226 #20234 200, Does a DEACTIVATED sys_position still receive sharing-rule shares? expandPositionUsers never reads the catalogue row, so today it does #8710 404), and each now names a commit whose own content names that number (git show hit counts from 2 to 85 per pair). Checked against the commits themselves:
  • The four NEW anchors. db16b9424 is the squash of PR feat(client)!: converge analytics.query / meta / explain and automation.trigger on unwrapResponse — one SDK calling convention (#13079) #14526 (its subject), and its sixth sub-commit registers the ADR-0087 entry with "Registration requested on spec: register the ADR-0087 semantic entry PR #14526 needs — the @objectstack/client envelope convergence has a truthful registered disposition and two exact precedents, not a missing category #14996; the landing PR is feat(client)!: converge analytics.query / meta / explain and automation.trigger on unwrapResponse — one SDK calling convention (#13079) #14526", so the feat(client)!: converge analytics.query / meta / explain and automation.trigger on unwrapResponse — one SDK calling convention (#13079) #14526 and spec: register the ADR-0087 semantic entry PR #14526 needs — the @objectstack/client envelope convergence has a truthful registered disposition and two exact precedents, not a missing category #14996 lines are both true. e944fdb24 is PR fix(client)!: bind the oauth.* family to the wire shapes better-auth sends #15445's squash, whose body says oauth.applications.delete "is deliberately NOT bound" because "binding it needs a behaviour change", which is the rewritten wording. f60ab90ae's changeset carries "CrudEndpointsConfigSchema.patterns is an exhaustive enum-keyed z.record — the input type demands all five operations and a partial parses to explicit undefined entries (wants z.partialRecord) #14365 proposes z.partialRecord for this record", so "the open z.partialRecord question (commit f60ab90)" points at the one in-repo record of that question rather than claiming the commit decided it, the right rung when the card is gone. 4bfe1a539 is PR feat(spec): refuse ${…} placeholder syntax in memory persistence.path / persistence.key at publish (#8495) #8666, and its second sub-commit reads "register the [finding][spec] memory-driver persistence.path / persistence.key still accept unresolved ${…} placeholders — the #8336 shape one surface over, deliberately outside its connection-material class #8495 refusal as an ADR-0087 semantic entry under protocol 18", so "the precedent of commit 4bfe1a5, PR feat(spec): refuse ${…} placeholder syntax in memory persistence.path / persistence.key at publish (#8495) #8666" is the precedent it was.
  • The anchors whose subject does not name the number. 9e0ba21a1 (Retire the paper metadata-customization protocol with its full coupling set — spec module, authorable customizationPolicies/mergeStrategy keys, metadata API §5, optional IMetadataService overlay methods, unreachable packages/metadata limb (re-charter of #12057) #13135) added all thirteen entry files that carried the #13135 header and retired the paper metadata-customization module whole, so "re-chartered, that commit executes the widened surface" is true. b3a63d32c (Enforce-or-remove the 15 RestServerConfig rows the liveness ledger now records as dead (#14369's verdicts) — routes.*, crud.patterns, crud.objectParamStyle, metadata.cacheTtl, metadata.endpoints.schema, batch.defaultAtomic, batch.operations.upsertMany #14691) added all eleven RestServerConfig entries and its body names the [finding] Ten declared RestServerConfig keys are normalized by RestServer and read by nothing — routes.* entirely, crud.patterns / objectParamStyle, metadata.cacheTtl / endpoints.schema, batch.defaultAtomic / operations.upsertMany (ADR-0049 enforce-or-remove candidates) #14369 census. dce5cd4f0 (Three plugin-manifest blocks — capabilities, configuration, extensions — have zero reads of the container itself, so all 8 keys beneath them are inert (configuration.properties.secret promises encryption/masking) #11332) added the three manifest-container entries. 0c2334f6c (RuntimeMode: 'preview' and the whole KernelContext.previewMode block are declared, documented as an auth bypass, and enforced by nothing #11846) added the three preview-mode entries and its changeset carries the ruling wording. bc56e1881 (contributes.routes has zero readers, yet a published skill, an ADR row and the dispatcher's own doc all present it as the way to serve a code-handler endpoint — enforce-or-remove fork #10726) added contributes.routes, which be21955ba names as the open fork. a3d5724c8 ([finding] Ten declared RestServerConfig keys are normalized by RestServer and read by nothing — routes.* entirely, crud.patterns / objectParamStyle, metadata.cacheTtl / endpoints.schema, batch.defaultAtomic / operations.upsertMany (ADR-0049 enforce-or-remove candidates) #14369) is the ledger commit whose rows read "seeded 2026-09-02 ([finding] Ten declared RestServerConfig keys are normalized by RestServer and read by nothing — routes.* entirely, crud.patterns / objectParamStyle, metadata.cacheTtl / endpoints.schema, batch.defaultAtomic / operations.upsertMany (ADR-0049 enforce-or-remove candidates) #14369)" and whose body counts "15 are dead", so "the liveness census commit a3d5724 recorded" is true.
  • The reused anchors. 2c86fe3ea ([finding] The ApiKey reference table documents better-auth's apiKey-plugin schema — a plugin this platform does not load and a shape sys_api_key does not have #8715; "Route 3 retirement kit" in its body), be21955ba (Retire the nine dead members of the plugin-manifest contributes block (events/menus/themes/translations/actions/drivers/fieldTypes/functions/commands) — tombstone route, kinds survives #10724 by subject; the Plugin-manifest contributes.themes declares a theme-contribution channel nothing reads — same declared-but-unenforced shape as the retired stack themes carrier #10627 reading and the 2026-08-24 cloud reading @5b5925a both recorded in its body, so "that commit's census" and "the cloud census" are true), 35ad101bc (defineStack({ themes }) is parsed, ingested and never applied — the last hop from a stored theme item to the theme engine does not exist #10485), 13c48c2a5 (integration/ErrorMappingConfig + ErrorMappingRule are 11 authorable keys with zero consumers — and one of them is named userMessage, colliding with the live #9934 channel #14676), 90bbf2510 (GET /api/v1/notifications 从不解析它声明的请求 schema —— cursor 被静默丢弃(SDK 分页永远第一页),limit 默认 20 声明 vs 50 实现 #6361; "ruled jointly with notification 响应侧:unreadCount 声明「总未读数」实测只数 limit 窗口内;响应 cursor 从无 producer 发出 #6363" in its body), 17d095413 (notification 响应侧:unreadCount 声明「总未读数」实测只数 limit 窗口内;响应 cursor 从无 producer 发出 #6363, the unreadCount fix, so "the unreadCount fix, commit 17d0954" is true), f549a0d4a ([spec] View-Management 五方法(listViews/getView/createView/updateView/deleteView)是零实现零路由的声明面 —— 已导致 #5948 把它的响应 schema 误当成线上路由的契约 #6239, the ViewProtocol sweep), d491625c1 (Nothing refuses the contradictory pair "external.credentialsRef bound + a config.url naming no user" — the binding is a silent no-op at connect #9041, the url-branch refinement whose composed twin 00777a0a6a is Composed-branch twin of #9041: external.credentialsRef bound + discrete mongo fields naming no username is the same silent no-op, unrefused #9147's).
  • The two 17.* entry files were split out of registry.ts by f293d454b7, not by their anchors; the anchors are the commits that first wrote the decision into the registry (git log -S puts 90bbf2510 and f549a0d4a first), which is the right reading of "the commit that decided it". RIGHT.
• The words. Where a line lost its number, the sentence is true to the commit it names: the README's "Measured while landing commit db16b94" describes the census-test incident that squash's own sub-commits fix ("slice census methods on masked source"); "(the additionalTypes on MetadataPluginConfig is authorable, documented as THE way a plugin declares a metadata type, and has no reader — a plugin cannot declare a kind at all #8586 / commit 2c86fe3 precedent)", "(the commit 0c2334f / [finding] HotReloadManager.startWatching watches nothing and logs "File watching started" at info; watchPatterns has no reader and watchHandles is never populated #12428 grading)" and "(the precedent of commit 4bfe1a5, PR feat(spec): refuse ${…} placeholder syntax in memory persistence.path / persistence.key at publish (#8495) #8666)" keep the live number beside the commit and the PR as a link only, ruling C's form. RIGHT.
• Nothing lit removed, nothing added. The numbers on removed lines that do not reappear on added lines are exactly the 21 dead numbers. The numbers on added lines are additionalTypes on MetadataPluginConfig is authorable, documented as THE way a plugin declares a metadata type, and has no reader — a plugin cannot declare a kind at all #8586, feat(spec): refuse ${…} placeholder syntax in memory persistence.path / persistence.key at publish (#8495) #8666, Composed-branch twin of #9041: external.credentialsRef bound + discrete mongo fields naming no username is the same silent no-op, unrefused #9147, [finding] spec ships a three-layer metadata-customization protocol (MetadataOverlaySchema, field-level change tracking) with zero consumers — published reference docs describe an unimplemented customization architecture #12057, [finding] HotReloadManager.startWatching watches nothing and logs "File watching started" at info; watchPatterns has no reader and watchHandles is never populated #12428 and fix(client)!: bind the oauth.* family to the wire shapes better-auth sends #15445, all answering 200, all already on the same lines at base. No number appears only at head. RIGHT.
• Census. Read from the dev's report and the diff, not re-run: packages/spec/src 235 to 4, migrations/ 233 to 2, entries 117 to 0, registry generated regions 114 to 0, hand-written 2 to 2; 254 base-only sites, of which 231 are this diff's (117 plus 114 copies) and 23 are main's own plugin-audit re-anchor (4dfff176b9, PR docs(plugin-audit): re-anchor the dead tracker citations to the commits that decided them #20737, on main's log); 0 head-only. The arithmetic matches the diff: 133 changed lines, 131 dead-number sites, the two lines without a number being the patterns follow-on and the oauth continuation. RIGHT.
• No accept-set or public-surface move. Follows from comment-only; Spec property liveness and the three concluded Type Check legs are green on this head. RIGHT.

② Semver level

• patch is right, and its stated reason holds. The diff is comment text, but the comments ship: at this head packages/spec/tsup.config.ts lists src/migrations/index.ts as a bundle entry and package.json exports ./migrations to dist/migrations/index.mjs and index.js (the shared checkout's HEAD, two days older, has neither, which is why the dev's dist measurement is not reproducible there). That the bundler keeps // comments is verified on the published @objectstack/spec@17.5.0 tarball: its dist/index.mjs carries 9,901 comment lines, the old "PR feat(spec): refuse ${…} placeholder syntax in memory persistence.path / persistence.key at publish (#8495) #8666 precedent" phrase 23 times, and the RETIRED_KEYS_BY_MAJOR table with its entry comments. So the compiled @objectstack/spec/migrations entry carries these comments and the changeset's rationale sentence is true. No export, schema, prescription or projected text moves, so nothing above patch applies. Check Changeset is green.
• Clause-②: no is right, in the PR body and as a standalone line in the changeset: no accept or reject moves, by the comment-only proof.

③ Boundary flags

• Scope judgment (the gate's census reads 117, under the ~120 slice threshold; the raw walk 131). The claim names the gate's own census as the instrument, so 117 governs; taking all of entries/** without a slice is right. Answered.
• The dev's open question: 13 dead sites left in registry.ts's hand-written parts (two comments, :5940 [finding][spec] memory-driver persistence.path / persistence.key still accept unresolved ${…} placeholders — the #8336 shape one surface over, deliberately outside its connection-material class #8495 and :22819 [spec] View-Management 五方法(listViews/getView/createView/updateView/deleteView)是零实现零路由的声明面 —— 已导致 #5948 把它的响应 schema 误当成线上路由的契约 #6239; eleven author-shown rationale strings, including step 17's :344 cli/runtime: explicit driver(--database-driver / OS_DATABASE_DRIVER)在 CLI 与 standalone stack 之间仍有三处分叉(#6265 后续) #6345 behind guide :68). Confirmed the claim forbade touching them: registry.ts appears only under "Regenerated only" with "Nothing is hand-edited", and both residual comment sites sit outside the os-generated regions (:5940 in step 18's doc comment above the semantic:18 opener; :22819 between the retired-key:18 closer and the retired-def:17 opener). The claim's own guide-:68 expectation conflicts with that ⛔; the dev honoured the ⛔ and reported, the right side of "stop on breach". Leaving them is not a gap in this stage: they are a different edit kind (form D wording, reviewed sentence by sentence as stage 9 was), and this PR's comment-only proof stays exact because of it. Option A, a next stage of packages/spec/src: 1,277 comment lines still cite 170 deleted tracker numbers (1,295 sites) — the staged remainder of ruling C+D on #19123, measured by PR #20226 #20234 for registry.ts's hand-written parts plus gen:upgrade-guide, fits. Answered.
• Two lines change without a number (the patterns follow-on, the oauth continuation): sentence continuity only. Answered.
• Merges. main has since moved to 9c8f113c64, three commits past the second merge base 01e78dceef, none touching packages/spec; the PR's recorded base 97005aed04 is one behind. No spec artifact can be jointly stale, and the queue rebuilds regardless. Answered.
• NOT MEASURED locally (build-schemas-check-mode.test.ts, def-key-collisions, the two publish-smoke-* files): left to Test Core, which had not concluded at the read below. Named, not presumed.
• Attribution. The three authored commits carry the model-free trailer pair; the two merge commits carry none. Fine.
• Labels untouched, scratch-directory reuse, the background gate batch, the kept worktree, hypothesis 6 falsified by construction: no action owed.
• Out-of-scope finding: the census reads re-charter #N as an ordinal. NON_CITATION_HEADS carries re-charter as a deliberate row ("re-charter round ordinals"), so a five-digit tracker number after it is skipped by the heuristic's design, not by accident; still, thirteen dead sites at base were invisible to the gate's count, which is a wrong census result anyone running --census reaches. My read: card-worthy (a bounded fix, for instance an ordinal is at most two digits), filed unassigned; not a blocker here, since this diff rewrites the sites anyway. ESCALATED to the seat as a filing decision.
• Out-of-scope finding: .md under packages/**/src is outside the census. The gate's surface is the declared glob packages/**/src/**/*.ts, and its header says a deferred surface is a decision, not an omission. Acceptance note. Answered.
• Out-of-scope finding: packages/spec/liveness/*.json carries Enforce-or-remove the 15 RestServerConfig rows the liveness ledger now records as dead (#14369's verdicts) — routes.*, crud.patterns, crud.objectParamStyle, metadata.cacheTtl, metadata.endpoints.schema, batch.defaultAtomic, batch.operations.upsertMany #14691 and CrudEndpointsConfigSchema.patterns is an exhaustive enum-keyed z.record — the input type demands all five operations and a partial parses to explicit undefined entries (wants z.partialRecord) #14365 in REMOVED notes. The liveness-JSON class is already on this card (pointer 5858331362, [worklist] Migrate liveness line citations to symbol anchors — census: 117-173 of 298 live line-cited pairs fail key-proximity at line granularity today #13003 in permission.json) and outside every citation instrument's declared surface, so it is a staging note on packages/spec/src: 1,277 comment lines still cite 170 deleted tracker numbers (1,295 sites) — the staged remainder of ruling C+D on #19123, measured by PR #20226 #20234, not a new card and not a wrong result of any instrument. Answered.

Check-runs on the head, read once in the first CI wave (the newest concluded run at the read was Dogfood Regression Gate (2/3)), not re-read and not polled. 32 runs: 17 success, 3 skipped, 12 in progress, 0 failed. Green: Check Changeset, Check PR Size, Governed Surface Queue Guard, Spec property liveness, Type Check · source gates, Type Check · consumer gates, Type Check · debt ledger, the four claim, closing-keyword and single-writer guards, Check Documentation Links, Flag docs affected by code changes, Auto Label, filter, Dogfood Verify CLI, Dogfood Regression Gate (2/3). Roster skips: Build Docs, Console Pin Gate, Packed-tarball smoke (opt-in). NOT CONCLUDED at the read, so not presumed green: Lint & Repo Gates (the run carrying check:issue-citations, check:migration-registry, check:spec-changes, check:upgrade-guide, check:doc-authoring and check:generated, the gates closest to this diff), Type Check · workspace, Build Core, Test Core (all six shards), Dogfood Regression Gate (1/3 and 3/3), Temporal Conformance (live PG + MySQL). The derived families the concluded green runs answer: changeset shape and deadline, PR size, governed surface, liveness, the three type-check legs, card and branch parity. The rest is the seat's landing rule; nothing red was seen.

Implemented-by: claude/issue-20234-migrations-citations
Reviewed-by: session_014EJ1ED8X4MMrT18BhVx4tx

VERDICT: PASS

---

Generated by Claude Code