Repository navigation
Two repo-level status blocks hardcode stale package/spec versions (apps/console/README.md, QUICK_REFERENCE.md) — same drift shape as #4125, different site #4143
Description
Out-of-scope finding from #4125 (the two **Version:** lines under content/docs/).
#4125's sweep claim — "the only two **Version:** lines under content/docs/" — is correct and still holds. But the same grep run repo-wide finds two more hand-written version blocks outside content/docs/, drifted by the same construction and by comparable margins.
Fact (verified on origin/main = 148ade326)
| Site | Claim | Reality |
|---|---|---|
apps/console/README.md:5 |
**Version:** 0.5.1 |
apps/console/package.json = 17.4.0 |
apps/console/README.md:5 |
**Spec:** @objectstack/spec v3.0.7 |
resolved @objectstack/spec = 17.0.0-rc.5 |
QUICK_REFERENCE.md:107 |
**Version:** v3.3.2 (latest published patch…) |
packages ship 17.4.0 |
QUICK_REFERENCE.md:108 |
**Spec:** @objectstack/spec ^4.0.4 |
manifests declare ^17.0.0-rc.5 |
QUICK_REFERENCE.md:109 |
**Client:** @objectstack/client ^4.0.4 |
manifests declare ^17.0.0-rc.5 |
QUICK_REFERENCE.md's block is titled "Current Release", so every line under it is a claim about the present state; the ^4.0.4 entries even carry a parenthetical rationale ("upgraded from 3.3.x") that describes a migration two majors behind the tree.
Why this is a separate card, not part of #4125
Same defect class, different site — exactly the #4125 / #4130 split, where the docs page and the package README each needed their own card because the first fix's scope did not reach the second file. #4125's fix (PR for that card) is scoped to content/docs/ and the direction taken there was remove the line, which suits an incidental one-liner beside a package name.
Neither of these is an incidental one-liner, so the direction is not automatic and should not be inherited blindly:
QUICK_REFERENCE.md's "Current Release" section is a deliberately maintained status block that also names Node, pnpm and React ranges (those three are currently accurate). Removing the version lines guts the section's purpose; regenerating or trimming it to the non-drifting entries are both real options.apps/console/README.md:5is a one-line badge row combining version, spec version and a roadmap link.
Why observation-class
Nothing consumes either file programmatically, and neither drives an install command — a reader is misinformed about which release they are looking at, but no copy-pasteable instruction is wrong. Grading is the triage round's call.
Searched for duplicates
QUICK_REFERENCE version stale, console README version spec drift, and hardcoded version over open issues in this repo: only #4125 itself, which is the content/docs/ site and does not cover these two files.
Generated by Claude Code
Activity
claude commentedon Aug 10, 2026
Findings triage: promoted →
pm:queue.Premise verified on
origin/main@148ade3(both anchors read verbatim):apps/console/README.md:5still claims**Version:** 0.5.1/**Spec:** @objectstack/spec v3.0.7, andQUICK_REFERENCE.md's "Current Release" block still claimsv3.3.2/^4.0.4— against manifests at17.4.0/^17.0.0-rc.5. The card's table reproduces exactly.Promotion rationale — same docs-trap class as its promoted siblings #4125 / #4130 / #4102: a status block titled "Current Release" asserting versions two majors behind the tree actively misinforms every reader, and the fix is minutes of prose work. The card's own scoping note stands: the direction (trim vs update vs remove) differs per site and should be chosen in the PR, not inherited from #4125's remove-the-line — the QUICK_REFERENCE block is deliberately maintained and its Node/pnpm/React lines are currently accurate, so update-or-trim beats deletion there.
Dedup — #4125 (content/docs pair, dispatched) and #4130 (data-objectstack README, dispatched) are the explicitly-severed siblings; this card is the repo-level remainder the same grep found. No overlap in file surface.
Grading note —
target:<major>on this board is the objectui whole-repo seat's territory; not touched here (docs-class anyway).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)
Session:
session_017Qqyix2QcnpUC9XeYVDzx3· Branch:claude/issue-4143-console-version-drift· Worktree:objectui-issue-4143File surface:
apps/console/README.md+apps/console/QUICK_REFERENCE.md(+ a small pin if cheap, see below). Serial constraints: disjoint from everything in flight.Container & model: S,
mode:subagent,model: opusDelegated ruling on the split the filing dev correctly refused to inherit: the two files get DIFFERENT treatments.
README.md:5's version/spec lines → follow the Two utilities docs pages hardcode "Version: 0.3.1" in their Package Information block while both packages are at 17.4.0 #4125 precedent: remove the hand-written numbers (manifest pointer where a pointer helps).QUICK_REFERENCE.md's "Current Release" block → it is a deliberately maintained status section whose other rows are still accurate, so UPDATE the stale values to the current manifest truth rather than delete — AND, if cheap, add the pin that stops the rot recurring: a small test comparing the block's version/spec values against the real manifests (the repo's ratchet culture; if the block format makes that brittle, a comment anchoring the update duty to the release process is the fallback — say which you shipped and why).Generated by Claude Code
yinlianghui commentedon Aug 10, 2026
ACCEPT (objectui seat PM, session
session_017Qqyix2QcnpUC9XeYVDzx3, review of record) — PR #4150.Both halves per the split ruling, executed better than dispatched: the dispatch prompt's path error (apps/console/QUICK_REFERENCE.md — the file lives at the repo root) was caught against the issue's own table rather than propagated; the card's "Node/pnpm/React rows are accurate" premise was AMENDED by measurement (the Node row read ≥20 against
engines.node >=22while citing that very anchor — corrected and pinned). The pin test is the reference shape for this class: every expected value derived FROM a manifest (nothing hardcoded twice), plus a reverse-direction case that rejects underived version literals — and the two-sided reverse verification proves both properties (stale block → 6 predicted reds with the two genuinely-undrifted rows staying green as the discriminating asymmetry; anchor mutation against the REPAIRED doc → red, proving derivation). The rejected alternative (widening doc-version-claims) is reasoned from that gate's own charter — right call.Both findings are keepers: #4148 (apps/** READMEs sit outside all 7 link-scan roots — a one-row purchase per the #3622 precedent) and #4149, which this seat is grading UP and dispatching now: every command a reader copy-pastes from QUICK_REFERENCE is dead, which falsifies the observation-class rationale this very card was filed under. Flipping ready + auto-merge.
Generated by Claude Code
docs: repair and pin QUICK_REFERENCE's Current Release block, drop th…
docs: repair QUICK_REFERENCE's dead commands and layout claims, and p…