docs(skills): retirement playbook §0 judges each family by mainstream capability - #20364
Merged
objectstack-fleet[bot] merged 1 commit intoSep 28, 2026
Conversation
… capability §0 said benign display metadata (`description`, `tags`, `icon`) is never retired and that "no code reads it" is the evidence for retirement. Both contradict the maintainer's per-family criterion: ask whether mainstream platforms have the capability; yes ⇒ build the consumer, no ⇒ retire, never judged by whether anything in this repo reads the key. The docs-shaped keys (`hook.label`, `flow.description`) stay, now as the "yes" outcome of that one criterion (render, do not retire) rather than as a second rule. `authorWarn` marking defers to the liveness README. Net 0 lines, widest table row unchanged. Claude-Session: https://claude.ai/code/session_01MjvgiFAmjHqsxy1XLiVYfH Co-authored-by: Claude <noreply@anthropic.com>
github-actions
Bot
added
size/s
documentation
objectstack-fleet
Bot
added
the
skip-changeset
Contributor
Author
Contract reviewServed-tier: Face: ① Derived judgments
② Semver level
③ Boundary flagsNo Implemented-by: VERDICT: PASS Generated by Claude Code |
objectstack-fleet
Bot
deleted the
claude/issue-20354-retirement-playbook-criterion
branch
September 28, 2026 03:40
akarma-synetal
pushed a commit
to akarma-synetal/framework
that referenced
this pull request
Sep 28, 2026
…stem-* migration entries states each lesson in words, not tracker numbers (stage 3) (objectstack-ai#20384) Part of objectstack-ai#20233 Clause-②: no **Stage 3 of a staged card.** The card stays open for later stages; this PR carries no closing keyword. Text only: no entry id, `surface`, `from` / `to`, conversion or matching logic moves, and the chain rewrites exactly what it rewrote before. ## What this does `os migrate meta` prints every ADR-0087 semantic entry it crosses as one block: `⚠ [protocol N] SURFACE → REPLACEMENT`, then `why:` (the entry's `reason`) and `verify:` (its `acceptanceCriteria`). AGENTS.md's runtime-string rule applies to all of it: 「Runtime strings — refusal prose, prescriptions, anything an author is shown — carry no tracker number (`pnpm check:doc-authoring`): the lesson goes into the text.」 Form **D** of ruling C+D on the parent card sets the shape: the lesson in words, and no number, dead or alive. This stage covers the next three families by site count, `driver-`, `kernel-` and `system-`: **132 sites → 0** in the three prose fields. None of the 26 entries carries a tracker id in `surface` (ruling A of the stage-1 ACCEPT, `5858839916`, is checked and has nothing to do here). Each site now says what the cited ruling, measurement or fix decided. ADR ids stay. `registry.ts`, `spec-changes.json` and `docs/protocol-upgrade-guide.md` are regenerated from the entries (`gen:migration-registry`, `gen:spec-changes`, `gen:upgrade-guide`), never hand-edited. The stage-1 pin now holds `engine-`, `ui-`, `plugin-`, `driver-`, `kernel-` and `system-`. ## Census — tracker ids in the author-shown fields **Instrument.** The stage-2 AST instrument, unchanged: a TypeScript-AST walk over every `packages/spec/src/migrations/entries/**/*.ts`. For each `entry` object literal it evaluates the string value of `replacement`, `reason`, `acceptanceCriteria` and (counted separately) `surface`, joining string literals with `+`, then counts `#` followed by 4 or 5 digits at a word boundary. **Validated first** by reproducing the stage-1 readings on the stage-1 tree (`443b2f4fdc`, extracted with `git archive`): `driver-` 7 entries / 44 sites (0 / 44 / 0, 25 distinct), `kernel-` 9 / 44 (1 / 41 / 2, 11 distinct), `system-` 10 / 44 (0 / 42 / 2, 6 distinct), `engine-` 5 / 67, whole tree 266 entries / 1,016 sites / 9 `surface` sites — every figure equal to the stage-1 census. **Tree measured:** `objectstack-ai/objectstack` at `569d4d2dbf` (this branch's base). Unevaluable fields: 0. **Controls, same run.** - **Lit:** `17.aggregation-node-distinct-retired.ts` reads 7 sites (replacement 1, reason 6), the reading stages 1 and 2 took. - **Dark (comment lines):** 794 `//` lines in entry files carry a tracker id, and none is counted. Comment lines belong to the sibling card, and ⛔ this PR touches none (794 before and after). - **Dark (field boundary):** the 7 `surface` sites left in the tree (other families) count 0 in the three-field total and 7 in the `surface` column. **Re-measured on the base, matching the stage-1 census:** `driver-` 7 entries, **44** sites (replacement 0 / reason 44 / acceptanceCriteria 0), 25 distinct ids; `kernel-` 9 entries, **44** (1 / 41 / 2), 11 distinct; `system-` 10 entries, **44** (0 / 42 / 2), 6 distinct. 37 distinct ids across the three (the families share `objectstack-ai#14478`, `objectstack-ai#15939`, `objectstack-ai#17635` and `objectstack-ai#3733`). `surface`: 0 in all three. Whole tree: 300 entries, **843** sites, 7 `surface` sites. **After this PR:** `driver-` 0, `kernel-` 0, `system-` 0; `engine-`, `ui-`, `plugin-` still 0; whole tree **843 → 711** sites; `surface` 7 (unchanged, other families). | entry | sites (replacement / reason / acceptanceCriteria) | |---|---| | `17.driver-aggregate-undeclared-key-aliases-removed` | 6 (0 / 6 / 0) | | `17.driver-capabilities-inert-bits-removed` | 4 (0 / 4 / 0) | | `18.driver-options-timeout-to-timeout-ms` | 1 (0 / 1 / 0) | | `17.driver-sql-distinct-bare-filter-typed` | 9 (0 / 9 / 0) | | `18.driver-sql-unresolvable-where-column-refused` | 14 (0 / 14 / 0) | | `18.driver-sql-upsert-cross-row-identity-merge-refused` | 9 (0 / 9 / 0) | | `18.driver-turso-config-local-path-wasm-retired` | 1 (0 / 1 / 0) | | `18.kernel-compatibility-matrix-estimated-migration-time-unit-in-key` | 5 (0 / 5 / 0) | | `18.kernel-context-preview-mode-retired` | 5 (1 / 4 / 0) | | `18.kernel-event-bus-retention-unit-in-key` | 3 (0 / 3 / 0) | | `18.kernel-health-check-and-hot-reload-durations-unit-in-key` | 7 (0 / 5 / 2) | | `18.kernel-package-lifecycle-durations-unit-in-key` | 3 (0 / 3 / 0) | | `18.kernel-plugin-health-report-durations-unit-in-key` | 3 (0 / 3 / 0) | | `18.kernel-plugin-security-durations-unit-in-key` | 4 (0 / 4 / 0) | | `18.kernel-runtime-config-timeout-unit-in-key` | 11 (0 / 11 / 0) | | `18.kernel-startup-orchestrator-durations-unit-in-key` | 3 (0 / 3 / 0) | | `18.system-cache-durations-unit-in-key` | 3 (0 / 3 / 0) | | `18.system-collaboration-durations-unit-in-key` | 4 (0 / 4 / 0) | | `18.system-failover-health-check-interval-unit-in-key` | 3 (0 / 3 / 0) | | `18.system-metrics-jsdoc-durations-unit-in-key` | 12 (0 / 12 / 0) | | `18.system-metrics-window-durations-unit-in-key` | 5 (0 / 3 / 2) | | `18.system-object-storage-durations-unit-in-key` | 3 (0 / 3 / 0) | | `18.system-registry-config-durations-unit-in-key` | 3 (0 / 3 / 0) | | `18.system-tracing-otel-exporter-durations-unit-in-key` | 5 (0 / 5 / 0) | | `18.system-tracing-span-duration-unit-in-key` | 3 (0 / 3 / 0) | | `18.system-worker-queue-rate-limit-duration-unit-in-key` | 3 (0 / 3 / 0) | | **total, 26 entries** | **132 (1 / 127 / 4)** | ## Every citation read, and what the text now says I read each cited issue or PR myself with single-card REST reads: the body, and the comments where a ruling or a measurement lives. Ids are in code spans so this body posts no cross-references. All 36 bare ids were resolved against this repository, because every sentence that cites one is about this repository's code; the one cross-repo id is `cloud#1651`. | cited | what it decided (read) | how the text now carries it | |---|---|---| | `objectstack-ai#3733` | The pruned `cached` field key: measured, the parse succeeded and the removed key was dropped without a word; the orphan schema was deleted. | "an earlier field-key prune measured exactly that — the parse succeeded and the removed key was dropped without a word" (health-check, OTel exporter) | | `objectstack-ai#3821` | The sharing-rule page: an unsortable query fell through to an empty page, and the driver fix made an unsortable query lose its ORDER BY, not its rows. | "the unknown-column recovery ladder (an unsortable query loses its ORDER BY, not its rows)"; "the ladder's own premise — rows matter more than their order"; "the ladder's recoveries" | | `objectstack-ai#4484` | `IDataDriver.findStream` removed: no production caller, two of three implementations buffered the whole set, and no tombstone because nothing parses a driver object. | "Retiring `IDataDriver.findStream` (it had no production caller, and two of its three implementations read the whole result set into memory …)"; "(`IDataDriver.findStream`, removed with no tombstone because nothing parses a driver object)"; by entry id in the `distinct` entry | | `objectstack-ai#4583` | The datasource ledger's dead keys removed; `capabilities.*` went as a whole block (11 of 11 unread). | "was retired separately, as a whole block nothing read" | | `objectstack-ai#4634` | Audit of all 34 `DriverCapabilities` bits: 3 live, 31 dead and tombstoned. | the entry already states the audit ("the follow-up audit checked every bit"); the trailing id is dropped | | `objectstack-ai#4914` | Maintainer, 2026-08-04: remove `manifest.loading` and `PluginHotReloadSchema`; keep `HotReloadConfigSchema`, the side with an implementation (`HotReloadManager`), as the start point. | "kept twice: as the hot-reload vocabulary that had an implementation when the manifest-side copy was removed, …" | | `objectstack-ai#4984` | An org-axis red-line gate read only aliases the schema rejects while its own fixtures spelt them: tests green, rule dead. | "the family of the org-axis red-line gate that read only rejected aliases while its own fixtures spelt them, so its tests stayed green and the rule stayed dead" | | `objectstack-ai#5181` | Narrow the query parameter of `IDataDriver`'s methods (`DriverQuery`, no redundant `object`). | "neither the narrowing of `IDataDriver`'s query parameters to `DriverQuery` nor the follow-through …" | | `objectstack-ai#5499` | Maintainer, 2026-08-05: freeze investment in `driver-memory` / `driver-mongodb`; fully lifted 2026-08-11 (comments `5249019855`, `5252526378`). | "the maintainer's 2026-08-05 investment freeze on driver-memory, which was lifted on 2026-08-11" | | `objectstack-ai#5540` | Remove `IStorageService.list(prefix)`: zero consumers, and the two adapters answered differently and both incompletely. | "(the zero-consumer `IStorageService.list`, whose two adapters answered differently and both incompletely)" | | `objectstack-ai#6011` | Maintainer: close the `ctx.user` `roles` alias now. | "(the `ctx.user` `roles` alias, closed at once on the maintainer's word rather than given a window)" | | `objectstack-ai#6075` | **404** — see Acceptance notes. | "the follow-through that brought five drivers' implementations in line" | | `objectstack-ai#6320` | `distinct`'s third argument meant different things on memory and sql; the sql half was dispatched, the memory half held under the freeze. | "(the measurement that found the two drivers reading this argument differently split the fix: the sql half is this entry, and the memory half was held back by that freeze)" | | `objectstack-ai#6321` | `query.aggregate` / `agg.func` are undeclared aliases whose only writers are driver fixtures; order: re-spell the fixtures, delete the aliases, then narrow the signature. | "The removal ran in a fixed order — the fixtures re-spelt first, the two alias branches deleted second, the parameter narrowed to `DriverQuery` last — because the reverse order yields red nobody can explain." | | `objectstack-ai#6404` (PR) | Executed that order and narrowed `aggregate`'s query parameter to `DriverQuery`. | the same sentence | | `objectstack-ai#7929` | Maintainer, 2026-08-12, ruling B: `driver-sql`'s filter refusal stops echoing `$field` operands, for every caller; the full diagnostic goes to the server log. | "the same predicate-text disclosure shape the driver's field-reference filter refusals had already been made to stop echoing (the full diagnostic goes to the server log, never the response)"; "that disclosure shape closed on the last dialect" | | `objectstack-ai#8371` | Ruled option 2: a dotted filter key whose head is a relation, a formula or a scalar is refused at both doors; a structured head stays unjudged. | "the axis owned by the dotted-filter verdict, which refuses a dotted key whose head is a relation, a formula or a plain column at the protocol and engine doors" | | `objectstack-ai#8592` | Measured on live MySQL: knex compiles the named conflict target away. | stated by the entry ("knex drops the named keys before the statement leaves the process"); the trailing id is dropped | | `objectstack-ai#8621` | Option A: a pre-flight refusal when no unique index backs the caller-named conflict target. | "Two earlier pre-flight refusals closed the half where no unique index backed a caller-named target …" | | `objectstack-ai#8622` | `id` becomes insert-only on the merge path: a merge on a non-primary conflict key was measured rewriting the existing row's primary key. | "`id` is insert-only on the merge path (made so once a merge on a non-primary conflict key was measured rewriting the existing row's primary key)" | | `objectstack-ai#8755` | Ruling option A: a pre-flight refusal when a second unique key could absorb a backed, caller-named target. | "… and the half where a rival unique key could absorb a caller-named one" | | `objectstack-ai#8790` | Maintainer, 2026-08-15: refuse both halves with `INVALID_FILTER` / 400, naming the column. | "Ruled by the maintainer on 2026-08-15: refuse BOTH halves …"; "Recover-both was excluded by the ruling's own argument" | | `objectstack-ai#8807` | Maintainer, 2026-08-15: an upsert must never modify a row whose identity the caller did not supply and whose conflict key it did not name; enforcement delegated, blanket refusal excluded. | "Ruled by the maintainer on 2026-08-15, as a contract principle …" (the principle itself was already quoted verbatim) | | `objectstack-ai#8926` | Maintainer, 2026-08-16, option A: MySQL's spelling joins the one shared predicate (envelope and recoveries together). | "Addendum 2026-08-16." — the paragraph already states option A | | `objectstack-ai#9061` (PR) | Implemented that option A. | the same | | `objectstack-ai#11825` | Maintainer, 2026-08-25: retire the declarative `AdvancedPluginLifecycleConfig` container; the classes stay a host-driven library. | "… and as a host-driven library when the declarative lifecycle config container was retired" | | `objectstack-ai#11846` | **404** — see Acceptance notes. | "maintainer ruling 2026-08-27 (Option A: remove)"; "(as the removal ruling recorded)" | | `objectstack-ai#14478` | Ruling B, 2026-09-02: a no-baseline gate plus an ADR-0087 rename of every offender (`DriverOptions.timeout` among the seven named); ruling B again, 2026-09-05: the population is every authored and every runtime-emitted duration, minus exemptions declared on the schema. | "Maintainer ruling B on duration units (2026-09-02, its population widened on 2026-09-05 to every authored and every runtime-emitted duration, bar the exemptions a schema declares on the key itself)"; "the duration-unit rule (…)" | | `objectstack-ai#14519` | The two tenant timeouts published a describe naming no unit (the unit sat in the JSDoc only); folded into the rename. | "the unit-nowhere shape (no unit in the name or in the published describe, first measured on two tenant timeouts)" | | `objectstack-ai#15626` (PR) | Landed the gate and the seven founding renames, the tenant `idleTimeout` → `idleTimeoutSeconds` among them. | "The tenant half was already renamed, in the same change that landed the duration gate itself" | | `objectstack-ai#15678` | `kernel/`: the 14 remaining duration keys carry their unit in the key name. | "the kernel-directory duration renames" / "the kernel-directory round"; trailing ids dropped | | `objectstack-ai#15679` | `system/`: the 15 remaining duration keys carry their unit in the key name; `size` got an honest name. | "the system-directory duration round"; trailing ids dropped | | `objectstack-ai#15939` | The gate did not read JSDoc. Ruled 2026-09-07: refuse the JSDoc / describe divergence. Ruled A 2026-09-11: remediate the population per file first, land the widened gate last. | "Director-seat ruling A of 2026-09-11 on the JSDoc-channel finding … a duration key whose JSDoc names a unit its describe does not is refused, and the keys in that shape are remediated per file before that refusal lands" | | `objectstack-ai#16024` | Maintainer, 2026-09-06, per key: forward `timeout`, remove `localPath` and `wasm`. | "ruled per key by the maintainer on 2026-09-06, once all three of this package's unread config keys had been measured" | | `objectstack-ai#17635` (PR) | The widened gate: refuse a duration key whose JSDoc names a unit its describe does not; landed last. | "lands that widened gate last, into a tree already clean" | | `objectstack-ai#18669` | Ruling A, 2026-09-17: rename `FileValue.duration` and `estimatedMigrationTime`, each with an ADR-0087 entry; no new closed type, no narrowing of stored data. | "Maintainer ruling A of 2026-09-17 on the last two duration keys no closed duration type could express …" | | `cloud#1651` | **Not readable from this session** — see Acceptance notes. | "cloud — a census closed 2026-08-26: OS_PREVIEW_MODE there is a routing-only switch …" | No call-shaped token moves: a `name(` census over `registry.ts` is identical before and after (297 distinct tokens), so textual call-spelling ratchets read the same. ## Pin — `packages/cli/test/migrate-meta-engine-guidance.test.ts`, widened `COVERED_PREFIXES` is now `engine-`, `ui-`, `plugin-`, `driver-`, `kernel-`, `system-`. The pin still spawns the real CLI (`os migrate meta --from 16 --to 18`) once, locates each covered block **verbatim** in stdout, and asserts the printed block — `surface` included — carries no `#` plus 4 or 5 digits. Anti-vacuity: - the `REWRITTEN` floor rises from 29 to **55** ids: the 26 entries of this stage (7 `driver-`, 9 `kernel-`, 10 `system-`) are added, and every covered prefix must still select at least one entry; - presence in stdout is asserted before cleanliness (the `driver-sql-unresolvable-where-column-refused` reason carries two blank-line paragraph breaks, and its block is found verbatim); - the detector is exercised on both sides first (lit on 4 and 5 digits, dark on 3, 6 and `ADR-0112`). The file keeps its stage-1 name; the header lists the six covered families. ## Ablation — the widened pin can fail on a `driver-` block From committed state, HEAD `1c0dc7ad54`, with `scripts/ablation-replace.mjs` in wrap mode (it owns the restore trap) and `scripts/ablation-dist-preflight.mjs` gating each leg. The bundle is built from the generated `registry.ts`, so that is the file mutated. - **Mutation.** In `registry.ts`, the reason of `driver-sql-upsert-cross-row-identity-merge-refused`: anchor `pre-flight refusals closed the half` → `pre-flight refusals (objectstack-ai#8621) closed the half`. The tool read anchor 1 → 0 and replacement 0 → 1, blob `b41e1d44` → `8f8d226e`. - **Mutate leg** (one lock turn: build, preflight, pin). Spec build exit 0. Preflight: marker present in 4 built files. Pin: **red**, `1 failed | 2 passed` — `driver-sql-upsert-cross-row-identity-merge-refused: the printed guidance cites a tracker id: expected 'objectstack-ai#8621' to be undefined`. - **Restore.** Tool-proven: blob `b41e1d44` == HEAD, `git diff HEAD` empty. - **Restore leg.** One lock turn, taken on the second try (the first waited out its 540 s budget, exit 99, NOT MEASURED, the tree already restored). Spec build exit 0. The `--absent` preflight found the marker in none of 222 built files, with the working tree clean against HEAD. Pin: **green**, `3 passed`. ## Verification Final head **`1c0dc7ad54`** for every line below; each heavy run went through `scripts/pm/os-verify-lock.sh` (one turn, `VERDICT command-exit 0`, per-step exits recorded separately). - **Build:** `pnpm exec turbo run build --concurrency=2 --filter='@objectstack/cli^...'` gives `Tasks: 55 successful, 55 total`. - **Pin and its neighbour:** `pnpm --filter @objectstack/cli exec vitest run --project integration --maxWorkers=2 test/migrate-meta-engine-guidance.test.ts test/migrate-meta-default-range.test.ts` gives `Test Files 2 passed`, `Tests 10 passed | 1 skipped` (the skip is the default-range file's own pre-existing `skipIf`). - **Spec tests that read these entries or the registry:** `pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 src/migrations src/kernel/preview-mode-retirement.test.ts scripts/build-schemas-check-mode.test.ts` plus the 19 other spec test files that read `MIGRATIONS_BY_MAJOR`, the registry or an entry file: `Test Files 24 passed`, `Tests 696 passed`. - **CLI unit:** `test/vitest-tiers-partition.test.ts` and `src/utils/spec-release-changes.test.ts`: `Test Files 2 passed`, `Tests 28 passed`. - **The call-spelling census that reads `registry.ts`:** `pnpm --filter @objectstack/driver-sql exec vitest run --maxWorkers=2 src/sql-driver-query-signature.test.ts` gives 15 passed. - **Typecheck:** `pnpm --filter @objectstack/spec typecheck` exits 0 (test layer: 53 files / 255 errors held in its ledger); `pnpm --filter @objectstack/cli typecheck` exits 0 (test layer: 3 files / 28 errors held, unchanged). - **Gate families:** `node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands` derives **89** families at `1c0dc7ad54` (after `git fetch origin main`). `--ran` over the recorded exit codes reads **89 derived, 89 run, 0 NOT-MEASURED, 0 UNRUN**, all exit 0. They include `check:doc-authoring` ("16466 customer-facing string(s) across 1135 spec sources clean"), `check:issue-citations`, `check:migration-registry` ("registry.ts is current (300 semantic, 219 retired-key, 199 retired-def)"), `check:spec-changes`, `check:upgrade-guide`, `check:generated` ("All 15 generated artifacts are up to date"), `check:duration-unit-keys`, `check:nul-bytes`, `check:adr-0087-registration` and `check:changeset-no-major`. - `check:dual-build-cjs-loads` refused first with `PREREQUISITE NOT MET` (exit 3: twelve packages outside the CLI closure had no `dist/`). Those `dist/` directories were written later in the same pass (04:48–04:49Z, inside the `check:type-check-debt` run, whose re-measure builds them); re-run at the same head it exits 0 (104 entries / 66 packages / 659 CJS files). The reconciled list takes that latest run. - **Lint (a proven narrowing, not the repo-wide run, which is CI's):** `eslint --no-inline-config --format json` over the 28 changed `.ts` files reports 28 files, 0 errors, 0 warnings. - The population is read from `eslint.config.mjs`: `**/*.{ts,tsx,mts,cts,js,jsx,mjs,cjs}` minus `NEVER_LINTED`, and all 28 are in it (no file-ignored warning). - Invariance: the config enables no type-aware linting (no `parserOptions.project`, no typed rules), so a text edit cannot move the verdict on a file it does not touch. - **Mergeability:** see Acceptance notes (driver-free `merge-tree` against `862b6ce869` exits 0). ## Acceptance notes - **Two dead ids, rewritten from the code on `main`.** `objectstack-ai#6075` and `objectstack-ai#11846` answer 404 on both the issues and the pulls endpoint, re-probed with a 200 control (`objectstack-ai#14478`). - `objectstack-ai#6075` (`distinct`'s "never reached it" sentence): `packages/drivers/driver-sql/CHANGELOG.md` (commit `d367f03`) and `sql-driver-query-signature.test.ts` record what it did — the five drivers' implementations followed `IDataDriver`'s `DriverQuery` narrowing. The sentence now says exactly that. - `objectstack-ai#11846` (preview mode): `packages/spec/CHANGELOG.md` (commit `0c2334f`), `packages/spec/src/kernel/context.zod.ts` and `preview-mode-retirement.test.ts` record the 2026-08-27 ruling (Option A: remove), the three-repo zero-consumer measurement and the re-declare-fresh condition. Dropped because `main` does not state them: "decision-inbox batch 2" and "all four decision facets pointed the same way". "The objectstack-ai#11846 card records the measurement" (objectui leg) now reads "zero consumers, measured when the removal was ruled", which is what the changelog and the test header say. - **One cross-repo id this session cannot read.** `cloud#1651` answers 403 here: `objectstack-ai/cloud` is not attached to this session (`add_repo` refused: no access). It is neither confirmed nor refuted, so its sentence was rewritten from what `main` records about it — `packages/spec/CHANGELOG.md` (`0c2334f`: closed 2026-08-26 with positive controls, `RuntimeMode` zero hits, `ArtifactKernelFactory` 20+ hits and never touching `previewMode`) and `context.zod.ts` (`OS_PREVIEW_MODE` there is routing-only). The cloud-side detail `main` does not state — `previewMode` "only as a local variable" whose effect is adding wildcards to "CSRF" trusted origins — is dropped; the parenthesis that replaces it describes this repository's own `serve.ts` (the one reader of `OS_PREVIEW_MODE` here only widens better-auth's trusted origins to preview-domain wildcards), which is measured on `main`. - **Decision-batch numbers went too (invisible to the regex).** Nineteen sites cited a decision batch as `#` plus two or three digits (`objectstack-ai#43` ×13, `objectstack-ai#115` ×4, `objectstack-ai#151` ×1, `objectstack-ai#158` ×1). They are numbers an author is shown and cannot follow, so each is dropped. One consequence worth naming: 13 entries said `Maintainer ruling B on objectstack-ai#14478 (2026-09-02, decision batch objectstack-ai#43)`, which fused two rulings on the same card — B of 2026-09-02 (the gate and the no-baseline rename) and B of 2026-09-05, decided in that batch (the population: every authored and every runtime-emitted duration, minus schema-declared exemptions). The sentence now names both dates. The `objectstack-ai#158` sentence (the agreement shape ruled an offence on 2026-09-18) is corroborated by `.changeset/18075-agreement-shape-is-an-offence.md` on `main`. - **"issue NNNN" / "PR NNNN" spellings, checked by hand.** No bare-number spelling exists in these 26 entries' author-shown text; the two `PR` citations (`PR objectstack-ai#6404`, `PR objectstack-ai#9061`) were `#`-spelled, so the instrument saw them and they are gone. The only `#` left in these 26 files is on `//` comment lines (sibling card's surface), including a `Prime Directive objectstack-ai#13` reference. - **A citation whose page says something narrower than the text.** `kernel-health-check-and-hot-reload-durations-unit-in-key` called `shutdownTimeout`'s shape "the objectstack-ai#14519 unit-nowhere shape". `objectstack-ai#14519`'s keys carried their unit in the JSDoc; "unit nowhere" is the gate's name for it (`check-duration-unit-keys.ts` header: "no unit ANYWHERE (the objectstack-ai#14519 shape)"), because the gate did not read JSDoc. The sentence now says what the shape is — no unit in the name or in the published describe — and that it was first measured on two tenant timeouts. - **A comment that my text edit makes slightly stale.** `18.system-metrics-window-durations-unit-in-key.ts` carries a `//` comment saying its acceptanceCriteria sentence "is objectstack-ai#15679's, left word for word". That sentence now says "that JSDoc-channel gap is filed as a finding of its own" where it said "is objectstack-ai#15939": same content, no number. The comment is the sibling card's surface (comment lines), so it is untouched here. - **Three "card" references re-anchored.** Removing an id left "the same card" in the Turso entry pointing at nothing; it now says "the same measurement". The kernel entries' "renamed by this same card" carry no number and were not otherwise rewritten, so they are left. - **Cross-PR check: no open PR adds or edits a `driver-`, `kernel-` or `system-` semantic entry.** Read at 2026-09-28T04:0xZ: the 18 open PRs' file lists (`GET /pulls/{n}/files`) carry 0 files matching `migrations/entries/semantic/NN.(driver|kernel|system)-*`. The Version Packages PR (`objectstack-ai#17076`) lists more than 1,000 files; the 1,100 rows read carry no entry file, and it is the bot-generated release PR. Nothing in flight will be held by the widened pin on arrival. - **`main` moved 4 commits past the base** (`862b6ce869`: `objectstack-ai#20364`, `objectstack-ai#20341`, `objectstack-ai#20366`, `objectstack-ai#20352`); none touches `packages/spec/src/migrations/` or the pin. A driver-free bare-clone `merge-tree --write-tree` of this head against `862b6ce869` exits 0 with no conflicted path, so `registry.ts` needs no merge, and `main` was not merged in. - **Generated projections** (`spec-changes.json`, `docs/protocol-upgrade-guide.md`) are regenerated, as in stages 1 and 2; their `--check` legs are green. Only the three `driver-` entries registered at protocol 17 appear in them, which is why those diffs are small. - **No other test pins these entries' text.** A `git grep` of test files for the 26 entry ids finds one (`preview-mode-retirement.test.ts`), which names the entry in a comment and reads no prose; a grep of tests for the 37 cited numbers finds only comment lines. So no test needed re-pinning this stage (stage 2's `migrations.test.ts` case has no counterpart here). ## Line budget Entry files: **352 changed lines** (+228 / −124) across 26 files, against the stage-1 ≈400 budget. The whole diff is **776 lines** (+516 / −260) in 31 files. Of the rest, `registry.ts` is 352, the two projections are 18 (`spec-changes.json` 12, the upgrade guide 6), the widened pin is 33 and the changeset is 21. --- _Generated by [Claude Code](https://claude.ai/code/session_01QcAS3qiYYZNezaxZxaUdMV)_ --------- Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #20354
Clause-②: no
What changed
One rule text in
.claude/skills/spec-property-retirement/SKILL.md§0. No new gate, no code, net 0 lines.description、tags、icon)永远谈不上「误导」;不要标authorWarn,也不要退役它。」 It is replaced by one bullet that states the per-family criterion in the maintainer's own words and derives the documentation-shaped case from it:5727134555;hook.label,flow.description) record intent for the next reader (ADR-0033), and the text now calls them an instance of the 「有」 branch: render, do not retire. The verdict still goes into the ledgernoteso the next audit does not re-open it. It is one criterion's outcome, not a second rule;description,tagsandiconare no longer exempt by kind. The row-level-policytagskey is named as the 「没有」 example (mainstream platforms have no such attribute, so it retires);authorWarnnow defers to the liveness README's author-warning rule, which is unchanged. The playbook no longer carries a second copy of it.dead」. An unread key isdead, anddeadis what the criterion is asked about. It does not answer the criterion. Same defect class, same section, one line, 8 bytes shorter. It sits inside the claimed file surface (§0) and is named here so review can take or drop it separately.Before (lines 43–47 at
8af914a3):After:
Why
The maintainer's ruling on #18900 (comment
5727134555, item ④, A′), verbatim: 「18900 同意 每族该问的是:主流平台有没有这个能力 —— 有 ⇒ 补消费端(一次做对);没有 ⇒ 退役,而不是看仓里有没有人读。」 The ruling record's own gloss is "a family is judged by whether mainstream platforms in the domain have the capability, ⛔ not by whether anything in this repo reads the key". Triage applies it to every family card. The playbook said the opposite fordescription/tags/icon, and a dev on the RLS family already hit the fork: #20353 retires atagskey the old text forbade. Triage direction (5862082626): "the ruling wins, and the text follows it". The documentation-shaped exemption is kept as the case where the mainstream carries the capability, which is how #20299 was graded (render, do not retire).Inventory: other texts that restate the old exemption (triage note 2)
Searched at
8af914a3withgit grepover.claude/**,skills/**,AGENTS.md,CLAUDE.md,docs/**(ADRs included),content/docs/**,packages/spec/liveness/**,packages/spec/README.mdandpackages/lint/src/lint-liveness-properties.ts. Patterns: 良性 / 展示元数据 / benign display / display metadata / 永远谈不上 / never retire / 不要退役 / 文档形状 / docs-shaped / 有意保留 /authorWarn/ 主流平台 / mainstream / 零消费者..claude/skills/spec-property-retirement/SKILL.md§0.claude/skills/**,.claude/agents/**authorWarnhas exactly 1 hit in.claude/**, the §0 line)skills/**,AGENTS.md,CLAUDE.mdskills/objectstack-ai/SKILL.md:225"Display metadata" is a table cell describinglabel/description, not an exemption)docs/adr/**(ADR-0033, ADR-0049 read)packages/spec/liveness/README.md"Author warnings", rule 1 (:562–565)authorWarnhalf: benign display metadata is not warned. It never says "never retire"packages/**). The new §0 points at it rather than restating itpackages/lint/src/lint-liveness-properties.tsheader (:21)authorWarnhalf onlyThe ruling does not address warnings, and the README's warning rule is consistent with it. A key the mainstream carries is a missing consumer, and warning its author would call it misleading. A key the mainstream lacks retires. So the family to fix is this one file.
PM mechanism hypotheses, as measured
8af914a3(67047171plus one unrelated docs commit). The bullet spans:43–47and the two target sentences:45–47.check:pm-skill-id-lint: it does not scan this file. Its scan set is the pm-dispatch tree,os-dev.mdandAGENTS.md(34 files, green), and this playbook carries 20 existing tracker-number citations. The bare comment id does not match its#pattern either. So the ruling record id stays in the rule text, as the triage direction asked, and no tracker number is added to the file.Line budget
SKILL.mdlines (ceiling 337)ceil(bytes/4)(the token ratchet's convention).claude/skills/spec-property-retirement/, 1 file)Paid by deletion in the same bullet. Removed: 「没有运行时消费者,但被有意保留 —— 它们」, the
hook.descriptionexample, 「按 … 常是模型」, and the whole 「良性展示元数据 … 也不要退役它」 sentence. No re-wrap of untouched text. New lines are 116 / 115 / 115 / 119 / 102 bytes, and line 61 goes from 120 to 112. Every line break falls between two wide characters or at a real space, so none renders as a stray space. No ceiling was raised and nothing moved to another file.Verification (HEAD
085a5026)The derived list comes from
node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands(18 commands, re-derived aftergit fetch origin main, same list). Every command ran. Exit codes were captured before any pipe.node scripts/check-ci-filter-parity.mjsnode scripts/check-closing-keyword-parity.mjs(and--self-test)node scripts/check-comment-mask-corpus.mjsnode scripts/pm/check-harness-current.mjs --self-testpnpm --filter @objectstack/lint run check:doc-formula-expressionspnpm check:agent-test-spellingpnpm check:cross-package-test-inputspnpm check:doc-authoringpnpm check:driver-memory-censuspnpm check:gitlink-declaredpnpm check:nul-bytespnpm check:pm-governed-mergespnpm check:pm-skill-ratchetpnpm check:refd-timer-probepnpm check:required-contextspnpm check:skill-frame-syncpnpm check:watch-hint-literalReconciliation:
dispatch-gates --ranwith coded exits reports "18 derived famil(ies) accounted for — 18 run, 0 NOT-MEASURED (a DERIVED zero)".Also run:
packages/spec/src/shared/retired-key-migrate-sentence.test.tsreads this playbook.pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 src/shared/retired-key-migrate-sentence.test.tspassed: 1 file, 14 tests, exit 0. The rest of the shard is left to CI..claude/skills, which the derivation marks as not placeable. Each exited 0:check:pm-settings-deny-roster,check-skills-token-ratchet.mjs,check:pm-skill-id-lint,check:pm-governed-prose,check:skill-top-level-keysandcheck-published-list-mirrors.mjs.Changeset
skip-changeset. The only path is under.claude/**(fast lane). As a measurement, the root package that holds.claude/isprivate: true, no workspacepackage.jsonnames.claude, and the new text 「主流平台有没有这个能力」 has exactly one hit in the tree, this file. Positive control: the liveness README sentence "Only mark genuinely" is found underpackages/spec/liveness/, which@objectstack/specships throughfiles[](liveness).Acceptance notes
hook.descriptionleft the example list, and 「按 ADR-0033,常是模型」 is shortened to 「(ADR-0033)」, to pay for the criterion inside the same five lines.job/datasourcelabel rows and the RLS policy label row. They are left as they are. Their keep / do-not-warn outcomes agree with the criterion for keys the mainstream carries. studio: show the authoredlabel/descriptionon flows, hooks, app areas, RLS policies and the view container (7 keys) #20299 carries the render for seven such keys, and family grades re-grade rows as they land (feat(spec)!: retire rowLevelSecurity[].tags — no mainstream platform tags a row-level policy (ADR-0049) #20353 retires a tags key on the RLS family)..claude/**. Draft. It lands through the skills seat's contract review. This PR does not flip it to ready, arm auto-merge or merge it.维护者速读(草稿)
spec-property-retirement)§0 那句「良性展示元数据(description、tags、icon)永远不退役」,换成您在 [Decision] Route declared≠enforced work by the SEAM, not the layer — aSeam:line on filing, vertical dispatch by default in the spec lane, automatic parent + sub-issues for spec↔objectui seams, Journey as a filter, bulk retirement per spec family, Console Pin Gate back to required (the maintainer's 「同意」 on the five-line batch, 2026-09-18) #18900 定的判据原话:每族问主流平台有没有这个能力,有就补消费端,没有就退役,不看仓里有没有人读。hook.label、flow.description 这类文档型键仍然保留,但理由改成「主流平台也有」,是同一条判据的结果,不再是另一条豁免规则。同节另一句「没有代码读它才是退役的证据」一并改成「才证 dead」。Generated by Claude Code