[objectstack-fleet]

Requested by Elon · project thread

Fixes #25148

What changed

Docs only, two files under content/docs/releases/ (+4 / -3). No script moves in this PR, per the card.

Page	Before	After	Source
content/docs/releases/index.mdx:21 (/docs/releases), v18 entry	(current series: 18.0.0, released 2026-10-19)	(current series: 18.0.0, released 2026-10-20)	npm time["18.0.0"] = 2026-10-20T12:51:40.837Z (read from registry.npmjs.org/@objectstack/spec at 2026-10-23); the Version Packages commit 06cab4849 (#24524) is dated 2026-10-20T11:47:07Z; releases/v18/18-0.mdx already says "published 2026-10-20"
content/docs/releases/v17/index.mdx:25 (/docs/releases/v17), status blockquote	"A plain install now resolves 17.7.0."	"A plain install resolved 17.7.0 until 2026-10-20, when 18.0.0 took npm's latest tag; a plain install now resolves the v18 line."	npm dist-tags = {latest: 18.0.0, next: 18.0.0-next.5, rc: 17.0.0-rc.6} (same read)

The index field keeps the exact current series: X.Y.Z, released YYYY-MM-DD shape, because sync-release-index-currency.mjs's STATUS_SHAPE refuses a field with anything after the date; the sources therefore live here, not in the parenthetical. The v17 sentence names the v18 line rather than 18.0.0, so it does not go stale again at 18.1.0.

Premise, measured

• origin/main 54df640 read both stale sentences as the card quotes them; 18.1.0 is not cut (the index still names 18.0.0, latest is 18.0.0).
• The cause the card gives holds: 8fbbd5082 (docs(releases): state 18.0.0 as the v18 GA release, so the GA Version Packages PR renders #24584, 2026-10-19T16:51Z) wrote "released 2026-10-19" before the cut; at the version commit the field already named the newest GA, so the script left it, date included (rewriteStatusField returns null for a same-version field). On this branch node scripts/sync-release-index-currency.mjs exits 0 with "nothing rewritten".

Done-when 2: should the currency script verify the date it leaves in place?

Yes, but not in this PR (the card forbids a code change in a docs PR), so it is reported to the seat as an out-of-scope finding. The reading:

• The script's refusal to re-date a same-version field is right for prerelease cuts (its header records the 17.7.0 re-dating it prevents). The gap is narrower: the run that IS the version commit of the newest GA, where the field was pre-written by hand. That run is detectable without a clock or network: the newest GA in the committed packages/spec/CHANGELOG.md at HEAD differs from the newest GA after changeset version. On that run, and only that run, the date it holds is true, so it should stamp it even when the version already matches.
• A check-only alternative (compare the date to npm time) needs the network at version time and so cannot be the guard on the version-pr lane.

Gates

Head 9be39f2: node scripts/pm/dispatch-gates.mjs --commands derived 48 commands for the two changed paths. Each ran with its exit code captured before any pipe, recorded as COMMAND :: exit CODE; --ran reconciles 48 derived, 48 run, 0 NOT-MEASURED, 0 UNRUN. All 48 exit 0, among them check:release-page-status, check:release-notes, check-release-section-coverage.mjs --strict, check:doc-anchors and check:docs. check:docs-transcript-drift first exited 3 (PREREQUISITE NOT MET, lint not built); after @objectstack/lint... and @objectstack/client-react... were built under the verify lock (VERDICT command-exit 0) it exits 0.

Not measured locally, CI's: Build Docs, Test Core, the type-check lanes and the wide-population families. No package source moves, so no test or ablation applies. content/docs/** is in no package's files[], so this carries skip-changeset and no changeset.

Acceptance notes

• The attribution block's first line is an HTML comment the session requires; os-dev.md forbids HTML comments in GitHub text. Same shape as PRs docs(pm-dispatch): a project-coordinator reference — the cross-lane scheduling rules a Claude Code project runs by, so its instructions name only its lanes #25255 and docs(objectql): Multi-Tenant Schemas says the injected organization_id is stamped on insert and NOT NULL on a table the driver creates #25322; reported as a deviation.
• The branch is the harness-pinned claude/project-thread-bb3ks9; the claim (comment 6355543510) records it as the identity.

🤖 Generated with Claude Code

https://claude.ai/code/session_01854Y7atqKDuWMySuBxa1SJ