Split (b) of #12237, created by the triage seat on the domain:devx seat's dispatch-time routing measurement (comment 5535922656). #12237 keeps split (a) — the authored pages and the navTitle adoption.
Blocked-by: #12238
History: this line read Blocked-by: #12237 until 2026-09-27. #12237 closed completed (PR #20170). #12238 (p1, pm:dispatched) is in flight on the adjacent description emission in packages/spec/scripts/build-docs.ts (:479 / :912, next to this card's title emission at :478) and regenerates all of content/docs/references/**, so two concurrent PRs would collide on every generated page (the domain:devx seat 2 serial note, 5857122314). Serial after it; rebase and regenerate with bash scripts/pm/os-regen-merge.sh (triage seat, session_01W89enF2dYV7K4N2Fbfj33f).
Why this is its own card
#12237 asks for one title rule applied across the docs site. Measured on origin/main at 2026-09-04T14:40Z by the triage seat, ⛔ not inherited from the card:
content/docs/**/*.mdx 405
with a generated / do-not-edit marker in the first 20 lines 225
That reproduces the domain:devx seat's reading exactly, and it is the fact that splits the card: a title rewrite on those 225 pages is a change to the generator, not to the page. Hand-editing them is overwritten on the next generator run and goes red on the drift gate. So the edits land in two places with two owners, even though the rule is one artefact.
The generator sites the devx seat names are packages/spec/scripts/build-docs.ts and build-skill-docs.ts, with check-generated.ts as the drift gate. ⚠️ Those are the seat's citation, ⛔ not re-verified by triage — locate the title emission by symbol at dispatch time rather than trusting the path, and re-derive the 225 against the tree you get.
Why it is blocked rather than queued
The rule itself is settled on #12237, whose acceptance criteria put the written rule and the full before/after table in a PR body that stays open for maintainer review — the card's own words are that this is maintainer-voice territory. Emitting a rule from the generator before that rule is approved buys rework on 225 pages.
⇒ Unblocks when #12237's rule is accepted. The dispatchable shape here is: take the accepted rule, express it in the emitter, regenerate, and show the before/after for the generated set.
What the split does not decide
⛔ Triage does not rule that the rule must live on #12237 rather than here. The devx seat offered a third framing — that the whole thing rides domain:spec because the generator's rule is the rule — and that stays available: if the executing seats find that expressing the rule in the emitter first and letting the authored pages follow is the cheaper order, say so on #12237 and the two cards swap roles. The split exists so that ⛔ nobody dispatches a 403-page edit of which 225 would be reverted, which was the devx seat's actual refusal, and it is correct.
⚠️ navTitle (apps/docs/source.config.ts:44, landed by #12311) had 0 uses in content/docs at the devx seat's reading. If (a) adopts it for authored pages, generated pages get short sidebar labels only if this card emits it too — so read (a)'s landed shape before choosing whether to emit navTitle here.
Standing trap carried in from #12236's report: it names 76 demoted body H1s whose wording differs from the frontmatter title. Those are inputs to the mapping table on whichever half covers the page.
Split (b) of #12237, created by the triage seat on the
domain:devxseat's dispatch-time routing measurement (comment5535922656). #12237 keeps split (a) — the authored pages and thenavTitleadoption.Blocked-by: #12238
History: this line read
Blocked-by: #12237until 2026-09-27. #12237 closedcompleted(PR #20170). #12238 (p1,pm:dispatched) is in flight on the adjacent description emission inpackages/spec/scripts/build-docs.ts(:479/:912, next to this card's title emission at:478) and regenerates all ofcontent/docs/references/**, so two concurrent PRs would collide on every generated page (thedomain:devxseat 2 serial note, 5857122314). Serial after it; rebase and regenerate withbash scripts/pm/os-regen-merge.sh(triage seat, session_01W89enF2dYV7K4N2Fbfj33f).Why this is its own card
#12237 asks for one title rule applied across the docs site. Measured on
origin/mainat 2026-09-04T14:40Z by the triage seat, ⛔ not inherited from the card:That reproduces the
domain:devxseat's reading exactly, and it is the fact that splits the card: a title rewrite on those 225 pages is a change to the generator, not to the page. Hand-editing them is overwritten on the next generator run and goes red on the drift gate. So the edits land in two places with two owners, even though the rule is one artefact.domain:devx— the ~180 authored pages: frontmattertitle:, plus adoptingnavTitlefor short sidebar labels.domain:spec— the same rule expressed inside the emitters, so generated pages carry it too.The generator sites the devx seat names are⚠️ Those are the seat's citation, ⛔ not re-verified by triage — locate the title emission by symbol at dispatch time rather than trusting the path, and re-derive the 225 against the tree you get.
packages/spec/scripts/build-docs.tsandbuild-skill-docs.ts, withcheck-generated.tsas the drift gate.Why it is blocked rather than queued
The rule itself is settled on #12237, whose acceptance criteria put the written rule and the full before/after table in a PR body that stays open for maintainer review — the card's own words are that this is maintainer-voice territory. Emitting a rule from the generator before that rule is approved buys rework on 225 pages.
⇒ Unblocks when #12237's rule is accepted. The dispatchable shape here is: take the accepted rule, express it in the emitter, regenerate, and show the before/after for the generated set.
What the split does not decide
⛔ Triage does not rule that the rule must live on #12237 rather than here. The devx seat offered a third framing — that the whole thing rides
domain:specbecause the generator's rule is the rule — and that stays available: if the executing seats find that expressing the rule in the emitter first and letting the authored pages follow is the cheaper order, say so on #12237 and the two cards swap roles. The split exists so that ⛔ nobody dispatches a 403-page edit of which 225 would be reverted, which was the devx seat's actual refusal, and it is correct.navTitle(apps/docs/source.config.ts:44, landed by #12311) had 0 uses incontent/docsat the devx seat's reading. If (a) adopts it for authored pages, generated pages get short sidebar labels only if this card emits it too — so read (a)'s landed shape before choosing whether to emitnavTitlehere.Standing trap carried in from #12236's report: it names 76 demoted body H1s whose wording differs from the frontmatter title. Those are inputs to the mapping table on whichever half covers the page.