릴리스 노트의 각 칸은 Markdown 인데 줄을 쪼개 `- ` 만 떼는 것이 전부였다.
그래서 작성자가 소제목으로 쓴 `## 작성` 이 항목 하나로 섞였고, `:::note`
와 `---`, `` 도 글자 그대로 목록에 나왔다. 그 노트의 내용이
하필 "본문의 표·코드·callout·그림이 렌더링됩니다" 였다.
소제목·수평선·블록 울타리는 항목이 아니므로 버린다. 소제목까지 살리려면
화면이 절을 나눠야 하는데 이 칸은 「무엇이 바뀌었나」 아래의 한 목록이다.
번호 목록도 항목으로 읽는다. 백틱은 남겨 두고 화면이 코드로 그린다.
robots.txt 가 404 였다. 파일은 이미지에 들어 있었지만 nginx 설정이 서빙할
파일을 하나씩 명시하는 구조라 등록되지 않은 것은 SPA 폴백으로 떨어진다 —
크롤러가 index.html 을 규칙으로 읽을 수는 없으므로 규칙이 없는 것과 같았다.
설명(meta)에서 백틱을 지운다. 본문에서는 코드가 되지만 검색 결과에는 그
문자가 그대로 나간다.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011nAjrqNHGPNngN4aUbcyvJ
에디터 칸 이름을 공개 화면과 맞춘다. 목적→이 기준을 쓰는 이유, 규칙→판단
기준, 적용 조건→적용할 때, 예외→예외와 주의, 사실→확인한 사실,
미지수→남은 미지수, 선택지→검토한 선택지. 쓰는 사람이 지금 채우는 칸이
공개 화면 어디로 가는지 외우지 않아도 된다.
대시보드의 세 칸이 같은 목록을 세 번 반복했다. 「검증과 미리보기」와
「게시 준비」가 「이어서 작성」을 걸러 낸 부분집합인데 「이어서 작성」은
전체를 그대로 그렸기 때문이다 — 모든 문서가 같은 단계면 세 칸이 겹친다.
각 문서는 지금 서 있는 구간 한 곳에만 나온다.
릴리스에서 「상태 공개 · 공개일 2026-08-22」 옆에 「아직 공개하지 않음」이
같이 나왔다. 경로가 비었다고 미공개는 아니다 — 공개 여부는 상태가 말한다.
행에서 눈에 띄는 단 하나가 되돌릴 수 없는 동작이었다. 편집을 같은 격의
버튼으로 올리고, 삭제는 danger 로 낮추고, 격자에 자리를 만들어 다음 줄로
떨어지지 않게 한다. 게시 기록이 이미 그 모양이었다.
주제·프로젝트는 「다음: 내용 편집」을 안내하면서 그 줄의 버튼은 게시 취소와
삭제뿐이었다. 공개 프로젝트 화면이 비어 있던 이유가 여기다.
죽은 코드였던 PublicDocumentHeader 를 걷어낸다. 화면을 실제로 그리는 것은
PublicRecordRenderer 이고 그 컴포넌트는 아무도 부르지 않았다 — 어제 이
컴포넌트에 줄바꿈을 고쳐 넣었지만 화면에는 닿지 않았다. 그 자리를 지키던
가드도 진짜로 그려지는 쪽으로 옮긴다.
탭 제목과 설명에 문서를 반영한다. 공개 문서가 전부 "Reference · Tech Log"
였고 설명은 모든 화면이 "Tech Log frontend" 하나였다.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011nAjrqNHGPNngN4aUbcyvJ
Case 만 블록 파서를 타고 나머지 유형은 문자열을 그대로 내보내고 있었다.
그래서 두 가지가 깨졌다.
작성자가 백틱으로 적은 식별자가 백틱째 화면에 나왔다. 어떤 Reference 는
한 문서에 백틱이 32개였고, 그 원문이 홈 카드·탐색 카드·검색 결과까지
그대로 퍼졌다. 판단 기준은 `<p>{rule.body}</p>` 로, 적용할 때·예외·예시는
`<li>{item.text}</li>` 로 아무것도 거치지 않고 나가고 있었다.
빈 줄로 나눈 문단이 문단으로 보이지 않았다. 어제 넣은 PlainText 는 모든
줄바꿈을 <br> 로 만들었기 때문에 빈 줄도 <br><br> 로 나갔고, 문단 사이
여백이 없었다. Reference 7개 전부에 빈 줄이 9~16개 있다.
ProseText 로 바꾼다 — 빈 줄은 문단을 가르고, 문단 안 줄바꿈은 그 자리에
지키고, 백틱은 코드로 읽는다. 백틱 수가 홀수면 원문을 그대로 둔다:
작성자가 백틱 자체를 쓴 것일 수 있고, 반쯤 해석해 한쪽만 코드로 만드는
쪽이 더 나쁘다.
블록 파서(parseCaseContent)를 쓰지 않는 이유는 그쪽이 Case 전용 문법과
도식 참조까지 물고 있어 이 칸들에 없는 해석기를 끌고 오기 때문이다.
머리말 가드는 <br> 개수로 재고 있었다. 이제 기준이 문단이므로 문단 수와
문단 안 줄바꿈을 함께 확인하도록 고친다 — 가드가 낡은 것이지 화면이
틀린 것이 아니다.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011nAjrqNHGPNngN4aUbcyvJ
600 은 「아래 저장 경로가 이미 600 이니 거기 맞추자」는 판단이었지 실제로
쓰는 글의 길이를 본 것이 아니었다. 요약은 여러 문단이다.
길어진 요약이 드러낸 두 가지를 함께 고친다.
목록: 최근 기록·탐색·검색·홈 초점은 요약을 통째로 그렸다. 2000자면 한
항목이 화면을 덮어 목록이 목록이 아니게 된다. 세 줄에서 자른다 — 전문은
문서를 열면 나온다.
머리말: `public-document-header` 만 `PlainText` 를 거치지 않고 요약을
그대로 넣고 있었다. 작성자가 나눈 문단이 전부 한 덩어리로 이어졌다.
판단 결정 요약에서 고쳤던 것과 같은 결함이고, 이 자리를 빠뜨렸다.
되돌리면 <br> 가 3개에서 0개가 되어 가드가 깨지는 것을 확인했다.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011nAjrqNHGPNngN4aUbcyvJ
The earlier fix covered the Markdown body and stopped there, so the preview
still ran lines together — which is exactly what it looked like from the
outside: nothing had changed.
Summary, problem, conclusion, environment and the rest are plain text. They
never pass through the Markdown parser, so their newlines sit in a text node
and HTML collapses them, and the renderer that now emits <br> for the body was
never asked about them.
They render through the same rule now. One helper, one behaviour: a line the
author broke stays broken, wherever they typed it.
Text written across several lines rendered as one run-on line. Markdown reads a
single newline as a space that joins a paragraph, the parser leaves that
newline inside the text node, and HTML then collapses it — so the break the
author pressed Enter for disappeared at the last step.
This was never a preview artifact: the Studio preview and the public page go
through the same renderer, so a published record ran its lines together too.
Newlines inside a paragraph now render as <br>. Paragraphs separated by a blank
line are already two paragraphs by the time they reach here, so this only
affects the breaks an author put inside one.
`PublicContentQueries` returned arrays, not promises. That signature is only
implementable by something already in memory, so the port could hold exactly
one adapter — the bundled fixture — and no amount of configuration could put
the public site on the backend. Turning it async is the change that makes a
second adapter possible; the adapter itself follows.
The markup is untouched. Every page reads a value and hands it to a
presentational component, so the shape those components receive is mapped at
the adapter boundary and nothing below the page changes.
Screens load through one query, not one per read. Several pages read in a loop
— the home timeline walks every project for its activity, the explore filter
walks search results to resolve titles — and a hook per read would mean a
variable number of hooks per render, which React forbids. `usePublicContent`
takes the whole screen's reads as one loader, where a loop is a loop and
`Promise.all` is available; the loops that used to be N sequential lookups now
issue together.
Two places deliberately do not show the loading surface. The explore filter
sits inside a page that already renders one, so a second skeleton would move
the layout under it — it keeps its structure and fills its options in when they
arrive. The search dialog is a type-ahead: re-querying per keystroke would
replace the results with a skeleton on every key, so it loads the catalog once
and applies the same predicate locally.
`usePublicContent` requires an object because `undefined` is how the query
layer says "no result yet". A loader returning the record itself would make a
missing slug indistinguishable from a request in flight, and the page would sit
on a skeleton instead of rendering its not-found route.
Studio's `resolvePublishedLabel` stays synchronous. It is called from inside
the public renderer, so making it async would push awaits through the render
tree; the shell loads the catalog once and the callback remains a lookup.
The component tests now assemble the query providers the running app assembles.
Without them the render throws "No QueryClient set" — not a harness quirk, but
the same failure the app would produce if it were mounted without its query
layer.
CLS was 0.192 on every public route, and one element accounted for all of
it: the footer moved at t≈482ms, right when the lazily loaded route chunk
arrived. The site frame laid the footer out in normal flow, so before the
content existed it sat at the bottom edge of the viewport — visible — and
then dropped out of view when the page grew to 2400px.
The frame is a flex column now with the footer pinned by `margin-top:
auto`, and the content slot holds a viewport of height so the footer
starts below the fold and only ever moves further out of sight. The slot
is `main` once the route renders and `section.ui-page` while Suspense is
pending; covering only the first left the shift in place, which is what
the intermediate measurements showed.
/ /explore /projects /releases /search /profile 0.1924 → 0.0001
360 / 768 / 1440 across four routes: no horizontal overflow
Separately, the route gate stopped the Studio page but not the Studio
shell, so a signed-out visitor who typed /studio still got the whole
workspace navigation — 작업본, 게시 기록, 새 문서, by name. No data crosses
an href, but "비로그인 사용자가 Studio 화면을 볼 수 없다" is not satisfied by
hiding the contents of a screen while showing the screen. The header is
drawn only for an authenticated session; `children` is already the
router's sign-in surface, which is the whole of what such a visitor gets.
signed out h1 "세션이 필요합니다." 0 nav, 0 studio links
signed in h1 "작업 흐름" 2 nav, 8 links, sign-out present
Three test updates follow from behaviour that changed rather than broke:
the frozen route inventory now derives `access` from each route's own
layoutGroup instead of asserting "public" for all 28 — so a Studio route
added without a gate fails that table too — and the router and shell
harnesses supply the session the Studio surface now reads. The router
suite also gains a test that a signed-out visitor gets neither the Studio
heading nor its navigation, which is the regression the gate exists for.
test:all is 1811 passed; the eight remaining failures are the three
load-dependent flake families (ci-artifact-contract,
provider-guardian-transaction, security-followup), all of which pass in
isolation and reference none of the changed files.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Adds TECH_LOG_STUDIO_SOURCE (MOCK | HTTP, default MOCK) to the V2 runtime
config schema so a build can switch createTechLogFeatureInstalledInput
between the mock and HTTP Studio gateways without a rebuild. V1 documents
predate the key and always normalize to MOCK. The HTTP gateway is
constructed with only { operations } per Task 4's actual signature -
no CSRF provider is wired here; that lands with attachCredentials at a
later composition-root task.
Updates every existing Studio test call site to the new required
createTechLogFeatureInstalledInput(context) signature via a shared
tests/helpers/studio-install-context.ts MOCK fixture, so the whole
existing Studio suite keeps exercising the mock adapter unchanged.
The template's demonstration screens exist to explain the template. A product
replaces them with its domain, so deleting them is the expected end state — but
two gates were coupled to them, and the merge accommodated that coupling instead
of fixing it.
product-features.test.ts derives its own scope now: a registry must gate on the
manifest when it imports a module belonging to a manifest-declared feature. The
previous fix put installed-feature-runtimes.tsx in the exempt list, which
silenced the guard for that file permanently. Verified by removing the manifest
reference from installed-feature-adapters.ts and watching the guard fail; a
counter asserts the sweep still watches at least one file.
product-feature-switch.test.tsx exercises the kill switch end to end again. The
mechanism is the ownership lookup plus isFeatureActive, which has nothing to do
with which screens ship, so the ownership map is the fixture: one real
registered route attributed to a real installed feature, with the product's own
router, components and codecs.
That restoration exposed a real gap. The navigation-withdrawal half is not
implemented here: it lives in the template's PrimaryNavigation and this product
does not render the template's AppShell at all — the public header is a
hand-written list of paths. A disabled feature's route is refused by the router
but its link would still be advertised. Harmless only while no feature-owned
route is navigable, which is now asserted so the gap cannot ship silently.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Integrates the frontend template sync (a0fbafb → 5434760) into the TechLog UI
migration. Merged in this direction so every conflict is resolved and proved in
the worktree; main is only fast-forwarded afterwards and never holds a state
that was not verified here.
15 conflicts. The rule throughout: keep the template's mechanism, keep the
product's content, and never invent a third state neither branch would accept.
The template's product manifest and its runtime feature kill switch are adopted.
The route registry is deliberately not composed from contract.routes: the
reference feature still declares screens this product deleted, and reducing over
them would register paths with no component behind them. ROUTE_FEATURE_OWNER is
narrowed to registered routes for the same reason. The first resolution did
compose from contract.routes and was rejected by product-features.test.ts.
Three files pinned counts and a digest describing the gate contract. Neither
side's numbers describe the merged config/ci/gates.json, so they were recomputed
from it rather than chosen: 27 gates, 82 commands, 94 command references, 107
evidence references, 128 artifacts, shape sha256 5063586d.
README.md and docs/accessibility/manual-checklist.md now enumerate this
product's 27 routes, which the template's own verify:documentation requires.
product-feature-switch.test.tsx was rewritten around the invariant that still
applies here — no registered route without a component — rather than deleted
with the screens it used to exercise.
docs/operations/template-merge-2026-08-17.md records every decision, the gate
results, and the three follow-ups this merge deliberately did not decide.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Carries eight template commits: the provider sandbox actually running, release
admission to a named environment, the product feature manifest with its runtime
kill switch, architecture and documentation rules that match what is enforced,
the removability fixtures, and the browser, visual and performance evidence.
Product identity is unchanged. `package.json` keeps `tech-log-frontend` and the
catalog keeps the Tech Log naming; the home page was not in the delta. The
visual baselines are this product's own — the template's were excluded from the
transplant and these were regenerated here, where the only difference is the
platform overview's new product-feature section.
What this repository gains operationally: `config/runtime/{local,development,
staging,production}.json` with `FE-GATE-027` refusing an artifact whose runtime
document does not match the environment it is being admitted to, and
`FEATURE_OVERRIDES` for taking an installed feature out of service without a
rebuild.
Verified here: eight gates green, build green, visual 5/5, and 1,858 of 1,859
tests in the suites that do not need a sandbox — the one failure passes in
isolation and is a jsdom lazy-chunk timeout under parallel load. The provider
suites cannot run on this machine at all: `kernel.apparmor_restrict_unprivileged
_userns=1` makes `bwrap --unshare-net` fail, reproducible without any code from
either repository.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>