Repository navigation
PM lane handoff — v17 whole-repo seat state at 收工 (2026-08-14, session 017Qqyix…) #4635
Description
Handoff record from the outgoing PM seat (session session_017Qqyix2QcnpUC9XeYVDzx3) for whoever takes this lane next. v17 has shipped (release commit 5bf5d2cb6 is on main). Lane state at handoff: zero seats in flight, zero PRs armed or draft-pending — everything dispatched by this seat is landed.
What just landed (this seat's final day, ~290 PRs total)
Three completed families, each traceable through its issues' claim/ruling comments: the number-policy family (#4576/#4590, percentPoints), the SchemaNode/label-vocabulary chain (#4548→#4578→#4580→#4581→#4593→#4603→#4608 + the #4617 main-red repair), and the gallery/dashboard chain (#4600→#4612/#4613→#4614/#4619→#4615→#4618/#4623→#4616/#4628→#4624/#4630 — the published gallery went from 15 retired placeholders + 33 unregistered panels to 1 deliberate demo tile).
⚠️ First things to check (v17 released changes these)
- The spec-pin unlock: console: FormPage's post-submit redirect uses the authored URL verbatim — undecided whether it is a console route or an arbitrary destination #4190 and Reconcile
exportOptionswith the spec's new object form (#8010 spec half): false alignment comment, local'pdf', and theas anystreamingread #4535-item-4 are both held on "the@objectstack/specpin covers fix(scripts): check-doc-links resolves the #fragment, not just the file (objectui#7644) #7657's schema change". With v17 released, a pin bump is likely now possible — that single bump unlocks both cards (their restart conditions are written on each). - The 不发版本 standing order is obsolete — this seat operated under no-release mode; re-confirm the release cadence with the maintainer before touching changeset/release machinery.
Dispatchable pool (triage-ready)
- A component type has THREE declared surfaces that disagree (TS schema type / registry meta inputs / renderer prop reads), and nothing reconciles or enforces them #4631 — a component type has three declared surfaces (TS schema / registry meta / renderer reads) that disagree, with measured specimens (a REQUIRED
childrenthe renderer never reads). Class-level instrument decision. - examples/schema-catalog
pnpm regenerateis destructive: it blanks six plugin-dashboard entries' curated titles and descriptions #4633 —examples/schema-catalog regenerateis destructive (blanks curated titles); checked-in index not reproducible from its own generator. Concrete defect.
Blocked / parked (restart condition on each card)
#4600 候 objectstack#8593 (SDUI dashboard node validation-schema ruling, needs-user-decision); #4296 ⛔ objectstack#8326; #4476 ⛔ #8147; #4337 候 #2890; #4308 与 #4036-A 耦合 (勿单裁); #2628 他账号. Observation-class (not seat-worthy alone, may combine): #4627/#4629/#4620/#4622/#4632/#4611/#4594/#4605/#4610/#4574/#4543/#4559/#4568/#4487.
Decision inbox (delegated rulings with veto windows still open — 勿催, but the maintainer may want to review)
Most substantive: #4580's revised Q1-A (BaseSchema.label/description widened to string | I18nLabel, read-time resolution — landed in PR #4608 after the original bridge-side ruling was measured unimplementable). Others are recorded on their issues: #4576-A (percentPoints), #4479, #4553-A, #4564-a, #4548-B+Q2, #4566-a, #4425 phase-2, #4467-A, #4306-A, #4036-A+#4308.
Cross-repo awaiting replies
objectstack#8326 (blocks #4296), #8593 (blocks #4600's remainder), #7917 (Phase-2 confirmation — Phases 0+1 landed as PR #4601, restart signal + survey corrections posted), plus #7427/#7634/#7517/#7702/#7768/#7804/#7862/#7864/#7918/#8243.
Lane conventions that bite (the short list; AGENTS.md/CLAUDE.md are binding)
- Claim = assign plus a claim comment with session ID + branch; re-read comments before coding (shared GitHub identity).
- Merge queue: flip draft→ready BEFORE arming auto-merge; arming an already-green PR can silently fail to enqueue — after arming, watch
git log origin/main(never the API) and re-enable on timeout. pull_request_read get_check_runssnapshots can be ~60 min stale; run-level conclusions vialist_workflow_runsare authoritative. Shellcurlto api.github.com 403s and reads as a silent false-green — MCP tools only.- Never
git stash(shared stack across worktrees); worktree-first before any edit; repo-rootpnpm exec vitest runonly; build the dependency closure with the package itself in the filter before judging types; GitHub's sanitizer eats<+letter — space generics and read bodies back after posting. - Every ruling this seat made is recorded on the issue it governs, with the measurement that justified it — the per-issue comments are the real handoff.
Generated by Claude Code
Activity
yinlianghui commentedon Aug 14, 2026
Closing as absorbed (seat
repo:objectui, sessionsession_01RnQd8iMMUwXQEV1crFmQiQ, 2026-08-14): the incoming seat has taken over (registry: objectstack#6025, takeover audit comment of 2026-08-14 ~09:56Z) and this handoff's ledger is carried there. First-check items are in motion: the spec-pin bump is filed and dispatched as #4636 (unlock path for #4190 / #4535-item-4 / objectstack#7100 / objectstack#6275); the release-cadence re-confirmation is flagged for the maintainer in the round report. Grading-debt and blocked/parked lists were re-verified this round (several entries had already closed — e.g. the gantt-i18n card landed via PR #4532).Generated by Claude Code