Repository navigation
sweep(spec-surface): 5-item description-surface truth sweep — one claim, one PR, per-item checklist review (#6085 #6081 #6137 #6065 #5676) #6243
Description
First bundled trivia sweep (maintainer 2026-08-07: PM coordinates same-class trivia into one card — economy of a drive-by fix WITHOUT losing the state machine). One dev, one claim, one PR closing all seven via multiple Fixes lines; review is a per-item checklist.
Membership rule (why exactly these seven)
Same class: spec description-surface text that lies or drifted — JSDoc/describe/ledger prose corrections with ZERO open semantic forks (each is either truth-restoration or already ruled). No behavior changes; regen + existing pin conventions carry the verification.
The seven (fix direction per item — each already carries an audit/triage comment)
| # | file | fix |
|---|---|---|
| #6085 ★ | expression.zod.ts dialect table (renders verbatim to the published ref page) |
remove retired js row (says "mapping"), add missing template; align with the enum 12 lines below; regen; pin per retirement-docs convention |
| #6081 ★ | stack.zod.ts:396 skills JSDoc |
drop "trigger phrases for intent matching" copy — triggerPhrases retired in #3896; point authors at the current spelling; keep the existing pin test honest |
| #6161 | strictness ledger prose (.md, hand-written half) |
the two no gate exemplars were re-verdicted (#5020/#5068); refresh the classification-table exemplars to current population (tree-wide no gate = 0) |
| #6137 | theme.zod.ts / chart.zod.ts |
delete the two obsolete "no doc block above this" // warnings (#5059 made them wrong) |
| #6065 | ApiEndpointSchema/ApiMappingSchema .describe() ×3 |
ruled 2026-08-07 direction B: pure describe copy — stop advertising whole-key rejection the publish gate doesn't do; vocabulary stays frozen (#5040) |
| #5676 | DiscoverySchema.environment (3) vs EnvironmentTypeSchema (7) |
ruled remaining scope: docs cross-reference both ways + a subset pin test; explicitly NO contract change |
| #5886 | packages/spec/src/data/hook.test.ts:422 |
re-spell the insert fixture input: { doc: … } to the engine's real shape (spec half of #5671) |
Excluded on same-file serialization: #6094 (flow.zod.ts tombstone) collides with in-flight #4415 (flow schema family) — it rides the NEXT sweep after #4415 lands.
Discipline
- ⛔ Nothing beyond the seven listed fixes — a sweep with riders defeats its own review model.
- Spec regen: the sweep touches TSDoc that feeds generated reference pages — expect
content/docs/references/**regen. Landing order: in-flight Design: should FlowNodeSchema parse its ADR-0031 regions, instead of a post-parse pass callers must remember to run? #4415 also regenerates spec artifacts; whichever PR lands second follows the generated-artifacts resync four-step (merge main → checkout generated from main → commit the merge → full regen → assert sibling entries survived). - PR body: per-item checklist mirroring the table, each with before/after one-liner.
- All seven members are assigned + marked "riding this sweep" — they close via the PR's
Fixeslines.
Activity
hotlong commentedon Aug 7, 2026
Handoff note (maintainer 2026-08-07): dispatch belongs to the spec seat. This card was assembled by the coordinating PM session (
session_01BickTBKm2JYSNnrtPT8ysa) as the first bundled-trivia sweep, and is deliberately left UNCLAIMED in the lane queue — the spec seat picks it up through its own batch selection (it sits on the v17 board, so it ranks high). For the claimer:Fixeslines.fallbackNodeIdtombstone prescribescondition: 'fault'— the migration it hands authors does not draw a fault edge #6094 stays excluded until Design: should FlowNodeSchema parse its ADR-0031 regions, instead of a post-parse pass callers must remember to run? #4415 (in flight, same file) lands.Generated by Claude Code
hotlong commentedon Aug 7, 2026
Re-routed to
domain:spec-surface+ membership updated (triage pass for the #6298 split, maintainer-instructed 2026-08-07). This card was assembled before the split existed; every member was screened against #6298's admission test in that pass, with two adjustments:no gate两个范例都已改判,但表里仍写成该类的现役人口(全树 no gate = 0) #6161 drops out →domain:spec-tooling. It is the strictness ledger's classification prose on a hand-written audit page, which [PM seat] domain:spec-surface — 🔀 merged into #6017 #6298's second red line assigns to the tooling seat. Not a sweep member any more; do not close it via this PR'sFixeslines.input: { doc: … }—— 按引擎真值 re-spell(#5671 摘除死 alias 的 spec 半边) #5886 drops out → staysdomain:spec. It passes the invariant but a test fixture is not in the seat's enumerated face; the spec seat may hand it back later.Sweep membership is now five: #6085 ★, #6081 ★, #6137, #6065, #5676. Still excluded and rejoining once in-flight #4415 frees
flow.zod.ts: #6094 ★.Unclaimed and waiting for the
domain:spec-surfaceseat (⏳ vacant as of this comment) — #6298 cites this card as its sweep-first pilot, so it is the natural opening card.Generated by Claude Code
[-]sweep(spec): 7-item description-surface truth sweep — one claim, one PR, per-item checklist review (#6085 #6081 #6161 #6137 #6065 #5676 #5886)[/-][+]sweep(spec-surface): 5-item description-surface truth sweep — one claim, one PR, per-item checklist review (#6085 #6081 #6137 #6065 #5676)[/+]Merge origin/main into claude/issue-6225-reference-cell-width
os-project-manager commentedon Aug 9, 2026
只读观察,来自 spec-surface 席(座位贴 #6298,session
session_018ffcE95NaMJcL9XJ9VDYgk)。⛔ 不构成认领,未改标签,未关闭 —— 本卡 assignee 是qq9340100,不是本席身份。巡检时发现本卡状态与实际落地情况不一致,证据如下,交由本卡持有者或分诊席处置。
这次 sweep 已经落地了。
git log origin/main:抽查成员卡 #6085:
state: closed/state_reason: completed/closed_at: 2026-08-07T14:20:38Z/closed_by: qq9340100—— 与 PR #6288 的七条Fixes一致。但容器卡本身仍是 open,且仍带
pm:queue+target:v17,最后更新停在 2026-08-07T13:49:46Z(逾 35 小时)。两处后果值得单说:
target:v17开着 = 发版板上的一条假条目。 工作已经发了,板上还挂着,清板时会被当成未完成项重新盘一遍。这是本卡最值得处理的一点。pm:queue开着 = 队列查询会把它捞出来。本席今晚巡检就是这样撞上它的 —— 它看起来像一张可派的卡,实际不是。顺带一个同类残留(同样只记录,不代改):已关闭的成员卡也留着
pm:queue/target:v17(#6085 实测)。这与本席今天在自己四张卡上清掉的是同一个机制 ——issue_write的labels是整组替换,dev 认领时会把读到的旧标签集连同pm:queue一起写回,所以收卡后需要回读一次清残留。建议处置(非约束):若七条
Fixes确实全部兑现,关闭本卡并摘pm:queue/target:v17;若有成员未随 #6288 落地,则相反 —— 本卡应当留着,但该说明还差哪一条。本席没有逐条核完七张,只抽查了 #6085,所以不下结论。本席座位贴 #6298 曾把本卡记作「已归档」。就工作是否完成而言那条记录是对的,就卡的状态而言是错的,下次全量刷新时改正。
Generated by Claude Code