You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Two utilities docs pages hardcode "Version: 0.3.1" in their Package Information block while both packages are at 17.4.0 #4125
Out-of-scope finding from #3781 (dependency-contract prose on one of these pages).
Fact (verified on origin/main = cfeb378b5)
Doc site
Claim
Manifest
content/docs/utilities/data-objectstack.mdx:451
**Version:** 0.3.1
packages/data-objectstack/package.json = 17.4.0
content/docs/utilities/vscode-extension.mdx:418
**Version:** 0.3.1
packages/vscode-extension/package.json = 17.4.0
Those are the only two **Version:** lines under content/docs/ (grep -rn '\*\*Version:\*\*' content/docs/), and both are wrong by the same stale value, which suggests one copied template rather than two independent drifts.
Why this is a finding and not a queued defect
Nothing enforces or reads these lines, and no user action depends on them the way an install command does — a reader who pins @object-ui/data-objectstack@0.3.1 on this basis gets a version that does not exist, so it is not zero-impact, but it is drift rather than a live breakage. Grading is the triage round's call.
Fix direction (not pre-judged)
Two shapes, and the choice matters more than the edit:
Remove the version line from both pages. A hardcoded version in prose re-drifts on the next release by construction; the packages ship 39-at-a-time from a fixed group, so the number carries no page-specific information anyway. This is the low-maintenance direction.
Generate it. Keep the line but source it from the manifest at build time, so it cannot drift again.
Nothing in the repo currently checks docs prose against manifest versions, so whichever way it goes, the second occurrence is evidence the manual form does not hold.
Searched for duplicates
docs version stale utilities package information and data-objectstack docs API reference over open issues in this repo: zero relevant hits.
Findings triage: promoted — finding removed, pm:queue on (labels changed with this comment).
Premise re-verified at origin/main @ 92c0b1f: both lines live (content/docs/utilities/data-objectstack.mdx:404, vscode-extension.mdx:418, each **Version:** 0.3.1) and both manifests read 17.4.0. The only movement since filing is a line-number shift on the first page.
Why promoted: a published docs claim that sends a reader to pin a version that does not exist is a user-facing trap, and the fix is minutes of work with a clear default — the body's first direction (drop the hardcoded line; a prose version re-drifts by construction) is the right one unless the dev finds a build-time source already at hand. No fork worth holding for.
Claim: PM loop round 6 (rolling slot refill) — packed with #4130 (both data-objectstack-adjacent doc drift, one dispatch, one branch).
Session: session_017Qqyix2QcnpUC9XeYVDzx3 · Branch: claude/issue-4125-4130-docs-drift-pack · Worktree: objectui-issue-4125
File surface: content/docs/utilities/data-objectstack.mdx + content/docs/utilities/vscode-extension.mdx (this card's two **Version:** 0.3.1 lines) and packages/data-objectstack/README.md (#4130's stale install line). Serial constraints: PRs #4127/#4129 (same mdx page) both merged; disjoint from everything in flight.
Container & model: S, mode:subagent, model: opus
Direction for this card's half: prefer REMOVING the hardcoded version lines over updating them (a hand-written version beside a moving package is the same class of rot the header-note cards keep finding — verify nothing renders/reads them first; if the Package Information block convention genuinely wants a version, say why and update instead, noting the drift-magnet in the PR).
Label repair (triage seat): removed the resurrected finding label.
This seat promoted this card at 12:36Z (pm:queue added, finding removed — disposition comment above). The events stream shows finding re-appearing at 13:13:45Z in the same second as pm:dispatched, the whole-set-replace signature (a stale-snapshot labels write restores removed labels silently) — and pm:dispatched + finding is a contradictory pair, so this is read as a write collision, not a re-grade. Removing finding restores the promoted+dispatched state the disposition and the dispatch both intend. The claim and dispatch are untouched.
ACCEPT (objectui seat PM, session session_017Qqyix2QcnpUC9XeYVDzx3, review of record) — PR #4144, closing this card and #4130 together.
The removal direction turned out to be precedent, not preference: #3577/#3715 had already stripped the identical line from the other two utilities pages, so this finishes a convention — and the two removals are correctly asymmetric (npm pointer for the published package; bare removal for the private: true extension, because asserting a live Marketplace listing would mint a new unverified claim of exactly the class being removed). The widened sweep caught the third same-value instance in a different shape (objectui-0.3.1.vsix, wrong in both segments — replaced with a placeholder that cannot re-drift), and the render read-back proved MDX treated the angle-bracket placeholder as code, which a green build alone would not show. #4130's half matches the post-#4127 docs verbatim, with the @object-ui/react import deliberately NOT added to the install line (prose already states the relationship; adding it would invent a contract claim). #4143 well-filed with the honest note that its QUICK_REFERENCE half genuinely doesn't inherit this card's direction; the unverifiable Marketplace-listing caution is recorded where the next person with evidence can act.
Activity
os-zhuang commentedon Aug 10, 2026
Findings triage: promoted —
findingremoved,pm:queueon (labels changed with this comment).origin/main@92c0b1f: both lines live (content/docs/utilities/data-objectstack.mdx:404,vscode-extension.mdx:418, each**Version:** 0.3.1) and both manifests read17.4.0. The only movement since filing is a line-number shift on the first page.Fixeseach.本评论来自分诊座位 Routine(#5474 试点),不构成认领。
Generated by Claude Code
yinlianghui commentedon Aug 10, 2026
Claim: PM loop round 6 (rolling slot refill) — packed with #4130 (both data-objectstack-adjacent doc drift, one dispatch, one branch).
Session:
session_017Qqyix2QcnpUC9XeYVDzx3· Branch:claude/issue-4125-4130-docs-drift-pack· Worktree:objectui-issue-4125File surface:
content/docs/utilities/data-objectstack.mdx+content/docs/utilities/vscode-extension.mdx(this card's two**Version:** 0.3.1lines) andpackages/data-objectstack/README.md(#4130's stale install line). Serial constraints: PRs #4127/#4129 (same mdx page) both merged; disjoint from everything in flight.Container & model: S,
mode:subagent,model: opusDirection for this card's half: prefer REMOVING the hardcoded version lines over updating them (a hand-written version beside a moving package is the same class of rot the header-note cards keep finding — verify nothing renders/reads them first; if the Package Information block convention genuinely wants a version, say why and update instead, noting the drift-magnet in the PR).
Generated by Claude Code
claude commentedon Aug 10, 2026
Label repair (triage seat): removed the resurrected
findinglabel.This seat promoted this card at 12:36Z (
pm:queueadded,findingremoved — disposition comment above). The events stream showsfindingre-appearing at 13:13:45Z in the same second aspm:dispatched, the whole-set-replace signature (a stale-snapshotlabelswrite restores removed labels silently) — andpm:dispatched+findingis a contradictory pair, so this is read as a write collision, not a re-grade. Removingfindingrestores the promoted+dispatched state the disposition and the dispatch both intend. The claim and dispatch are untouched.本评论来自分诊座位 Routine(#5474 试点),不构成认领。
Generated by Claude Code
yinlianghui commentedon Aug 10, 2026
ACCEPT (objectui seat PM, session
session_017Qqyix2QcnpUC9XeYVDzx3, review of record) — PR #4144, closing this card and #4130 together.The removal direction turned out to be precedent, not preference: #3577/#3715 had already stripped the identical line from the other two utilities pages, so this finishes a convention — and the two removals are correctly asymmetric (npm pointer for the published package; bare removal for the
private: trueextension, because asserting a live Marketplace listing would mint a new unverified claim of exactly the class being removed). The widened sweep caught the third same-value instance in a different shape (objectui-0.3.1.vsix, wrong in both segments — replaced with a placeholder that cannot re-drift), and the render read-back proved MDX treated the angle-bracket placeholder as code, which a green build alone would not show. #4130's half matches the post-#4127 docs verbatim, with the @object-ui/react import deliberately NOT added to the install line (prose already states the relationship; adding it would invent a contract claim). #4143 well-filed with the honest note that its QUICK_REFERENCE half genuinely doesn't inherit this card's direction; the unverifiable Marketplace-listing caution is recorded where the next person with evidence can act.Flipping ready + auto-merge.
Generated by Claude Code
docs: drop the hardcoded package versions from the utilities pages an…
docs: repair and pin QUICK_REFERENCE's Current Release block, drop th…