[Cuarroc]

Die zusätzliche Verschachtelung ist mit einer gelöschten Hintergrundregel in allen zwölf Klassisch-Fällen beseitigt.
Der Digest-Schalter samt Beschriftung, vollständigem Hinweis und Fehlermeldung bleibt erhalten und ist geprüft.
Der geprüfte Commit ist gepusht; die abschließende Fremdprüfung und Abnahme durch Root stehen noch aus.

Report

Plan-ID: KR-09B-2-DIGEST-LEAF
Ticket: #330; bff6ccc4-dc28-4897-a7b1-2186ca0717fd; parent #329 / KR-09b.
Owner: srv-kr09b2-digest.
State: DRAFT — authorized one-time ancestor correction delivered and pushed; final Root review/acceptance open.
Review: ausstehend (Stufe B). Root owns ONE outside-family bounded faa789217e03664c58ceefe1197f882fb64effee..3200045d84af763fb741fd66b248a1b790482b60 LAST delta review, then final acceptance. No author review or new independent review was launched.

Final change and scope

• Final checked/pushed head: 3200045d84af763fb741fd66b248a1b790482b60. Same open draft PR feat(settings): use Glass switch for the global daily digest #1964; exact remote and PR head verified. Starting head: faa789217e03664c58ceefe1197f882fb64effee; original product merge-base: 872b3d3c1a37d4faad2d9a82b4b27aa6c2e98746; retained reviewed source: 59505cedbf5551e4b607a197a8d003c6e8ee0cd8.
• Root message msg-0mvr9v4bf-181207fa explicitly authorized ONE ancestor-surface recut. The previous transparent-wrapper experiment is superseded as a remedy, with its evidence and review record retained below. Product acceptance, measurement and literal ceilings are unchanged.
• Commit 3200045d deletes exactly background: var(--bg) from the EXISTING .settings-view rule in src/styles.css:5056. The inner settings composition now uses the existing outer .g-mount-legacy material. That outer surface remains opaque, rounded, and unchanged; a second ancestor edit is unnecessary. Computed styles prove the retained Switch remains material in both states and themes.
• No geometry, selector, class, style attribute, new layer, component, copy, routing, state, backend, IPC, polling, protocol, shared Switch, gate, baseline, threshold, dependency or allowlist changed in this correction. Existing DigestFields wrappers remain unboxed; the correction reaches the measured material ancestor directly.
• 180 RAW total added+deleted against the original merge-base, including the existing 179 lines and the single deletion. Pre-edit forecast: 180. No exception. All 1,853 other tracked files, including all public tests, mount CSS/composition and the own PLAN row, remain byte-identical to FAA. The own row SHA-256 is 60707310eb5079149097d53e5b5dc10504ec9368d5eddde3f7efd34561722dee; normalized PLAN SHA-256 remains 73efbea71a1d94eef0afb5333d46d2022d24a0671829a4d9a0d68c9ee93cfe1d.
• Existing PLAN wording is preserved under the explicit source-only/no-PLAN authority. Its “Erledigt mit Merge” wording does not describe the live OPEN/DRAFT PR; Root owns that metadata reconciliation before Ready. This worker makes no merge/completion claim.
• Main at fresh admission: 6309d0ea1b3d72f1f8614f6bb489415139556d6c, ci run 40599478201 completed/success. No source refresh, merge, rebase, amend or history rewrite.

Red-first and public behavior evidence

Before editing, the unchanged browser CLI rendered all 396 cases at FAA. The independent literal classic comparison compiled/rendered successfully and exited 1, with all 12 nesting counts at 43/50/60 instead of the original 42/49/59. After the one-line deletion, that exact unchanged comparison exits 0: 24/31/41 for 0/2/12 items at BOTH viewports and BOTH themes. No fake class-string assertion or internal module mock was used. Commit trailer: Regression-For: 59505cedbf5551e4b607a197a8d003c6e8ee0cd8.

The retained public GeneralTab tests check controlled true/false, exactly one next-value callback per interaction, unique/stable useId/label/hint/error associations and complete copy. All 40 targeted public tests pass, including SettingsView resync, classic/fresh routing, D1-off/unknown siblings and mount hidden states. A finite actual mounted-browser inspection additionally checks the visible Digest switch true→false, one IPC write with false, a failed next write with true, rollback to false, and the associated visible role-alert. Only the Tauri boundary is mocked; app state/backend source is unchanged.

Inspected real candidate screenshots at 1440×900 in light and dark: classic page structure, Digest on/error states with full hint and error text, Masterprompt, Agent-Kategorien, Updates, Neuigkeiten, and the fresh settings route after navigation. Material control, existing outer surface and page structure remain visible; the hidden classic control is absent from the fresh route's accessibility tree. All browser/Vite instances were owned and closed by finite awaited commands. No native desktop app or detached process was launched.

Design measurement: all cases retained, raw exits distinguished

Vector order: [slack, nesting, tabs, eyebrows, grids]. The original baseline below is the previously reconstructed exact original-source archive, measured after implementation. All 1,851 original archive files were rechecked against their Git blob identities before reuse. The unchanged gate/metric/baseline source and Chromium 153.0.8010.12 are retained. The red and candidate runs are fresh, each 396 cases: 0/2/12 items × 1440×900/1920×1080 × light/dark. No --update.

Source	Raw CLI exit	Saved-ceiling failure entries	Aggregate vector	Elapsed
Exact original base (retained run)	1	42	[237, 2669, 132, 376, 25]	111.4 s
FAA before fix (fresh)	1	34	[225, 2661, 132, 376, 21]	110.5 s
Final candidate (fresh)	1	39	[234, 2447, 132, 376, 24]	112.5 s
Candidate minus original	—	-3	[-3, -222, 0, 0, -1]	—

Raw saved-ceiling debt remains a failing gate (exit 1), never called GREEN. No aggregate metric rises. The required independent classic comparison is GREEN and every non-MCP case is nonincreasing.

Root noise boundary: ONLY /einstellungen@es-tab-mcp capture noise is dispositioned. Excluding only those 12 cases gives 384 INCLUDED cases: original [222, 2532, 132, 376, 20], candidate [222, 2310, 132, 376, 20], delta [0, -222, 0, 0, 0]. Every one of those 384 cases is independently nonincreasing. ALL notifications and all other routes remain included. Notifications have one six-finding decrease and no increase; no general notification exclusion is used. The only positive per-case cells are the two MCP cases explicitly listed below. This correction does not claim graph completeness, native stability or a capture-gate fix.

Case: route ∣ items ∣ viewport ∣ theme	Original base	Before fix (FAA)	Candidate	Delta vs original
/einstellungen/klassisch ∣ 0 ∣ 1440x900 ∣ light	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[1, 24, 0, 15, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 0 ∣ 1440x900 ∣ dark	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[1, 24, 0, 15, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 0 ∣ 1920x1080 ∣ light	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[1, 24, 0, 15, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 0 ∣ 1920x1080 ∣ dark	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[1, 24, 0, 15, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 2 ∣ 1440x900 ∣ light	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[1, 31, 0, 17, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 2 ∣ 1440x900 ∣ dark	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[1, 31, 0, 17, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 2 ∣ 1920x1080 ∣ light	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[1, 31, 0, 17, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 2 ∣ 1920x1080 ∣ dark	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[1, 31, 0, 17, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 12 ∣ 1440x900 ∣ light	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 41, 0, 20, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 12 ∣ 1440x900 ∣ dark	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 41, 0, 20, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 12 ∣ 1920x1080 ∣ light	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 41, 0, 20, 0]	[0, -18, 0, 0, 0]
/einstellungen/klassisch ∣ 12 ∣ 1920x1080 ∣ dark	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 41, 0, 20, 0]	[0, -18, 0, 0, 0]

All other changed cases follow; the other 378 cases are identical to the original base. These raw cells are preserved even within the authorized noise boundary.

Case	Original base	Candidate	Delta	Disposition
/einstellungen@es-tab-mcp ∣ 0 ∣ 1920x1080 ∣ light	[0, 6, 0, 0, 0]	[3, 16, 0, 0, 1]	[3, 10, 0, 0, 1]	Root-dispositioned es-tab-mcp capture noise; retained raw
/einstellungen@es-tab-benachrichtigungen ∣ 0 ∣ 1920x1080 ∣ dark	[0, 6, 0, 0, 0]	[0, 0, 0, 0, 0]	[0, -6, 0, 0, 0]	Included; decrease, no notification exception
/einstellungen@es-tab-mcp ∣ 2 ∣ 1920x1080 ∣ dark	[3, 16, 0, 0, 1]	[0, 11, 0, 0, 0]	[-3, -5, 0, 0, -1]	Root-dispositioned es-tab-mcp capture noise; retained raw
/einstellungen@es-tab-mcp ∣ 12 ∣ 1440x900 ∣ light	[0, 11, 0, 0, 0]	[3, 16, 0, 0, 1]	[3, 5, 0, 0, 1]	Root-dispositioned es-tab-mcp capture noise; retained raw
/einstellungen@es-tab-mcp ∣ 12 ∣ 1440x900 ∣ dark	[3, 16, 0, 0, 1]	[0, 11, 0, 0, 0]	[-3, -5, 0, 0, -1]	Root-dispositioned es-tab-mcp capture noise; retained raw
/einstellungen@es-tab-mcp ∣ 12 ∣ 1920x1080 ∣ light	[3, 16, 0, 0, 1]	[0, 11, 0, 0, 0]	[-3, -5, 0, 0, -1]	Root-dispositioned es-tab-mcp capture noise; retained raw

Finding disposition for Root's final delta review

ID	Evidence and disposition
D1964-1 / M3	Measured regression corrected by 3200045d: all 12 classic cases pass literal ceilings; full raw totals and every non-MCP case nonincrease. Final Root confirmation remains open.
R1964-a1	Earlier transparent-wrapper hypothesis disproven by retained computed-style/failed experiment evidence. Wrapper unchanged; actual material ancestor flattened in the expressly authorized recut. Root final disposition pending.
R1964-a2	Existing row and accessible Switch/name/hint/error remain unchanged; public tests and inspected light/dark screenshots pass. Root final disposition pending.
es-tab-mcp capture noise	Only Root's named boundary applies; all changed cells retained above. No notifications/other-route waiver.

Architecture drift: gate exit 0, no new occurrences or allowlist edits. No dependency/license addition.

Commands and raw exits at final candidate

Each heavy phase used fresh guard.py srv-kr09b2-digest 7 && <phase> in the same awaited shell, warm target 7 and one Cargo build job, without profile overrides. Refusals for two active heavy shells or load >=8 exited 1 and started no phase; 30-second waits and fresh admission checks preceded execution. No guard/lease/cap/configuration/runner change.

Command/check	Exit
bash scripts/sync.sh start; git fetch origin; live main/PR evidence	0 each
npm run dev:start-check -- --max-cargo 2 --min-free-gb 3.5	0
npm run design:gate -- --out <private-red-output>	1 (existing debt; 396 cases rendered)
python3 <private-evidence>/compare-classic.py <red-results.json>	1 (expected actual regression before edit)
npm run design:gate -- --out <private-candidate-output>	1 (existing debt; 396 cases rendered)
python3 <private-evidence>/compare-classic.py <candidate-results.json>	0 (all 12 literal cases)
Full original/candidate case and aggregate comparison; exact-base archive identity	0 each
npm run typecheck	0
npm run test:unit -- src/components/settings/digest/DigestFields.public.test.tsx src/components/SettingsView.test.tsx src/shell/mount/settings-routing.test.tsx src/shell/mount/mount.test.tsx src/shell/mount/mount.edge.test.tsx	0 (5 files / 40 tests)
node <private-evidence>/inspect-candidate.mjs — actual mounted public states, screenshots and navigation	0
git diff --check; 180 RAW/scope/own-row/foreign-file hashes	0 each
bash scripts/ci/secret-scan.sh; normal git commit -F <message-file> with full precommit	0 each
bash scripts/ci/red-first.sh at 3200045d84af763fb741fd66b248a1b790482b60	0
git push -u origin codex/kr09b2-digest with normal FULL bash scripts/ci/gates.sh lane prepush hook in the assigned worktree at 3200045d	0 (219.7 s including push; lane 208 s)
git ls-remote origin refs/heads/codex/kr09b2-digest; exact remote/PR-head/DRAFT verification; clean tree and final file hashes	0 each

Final own full prepush ran ONCE through the normal push hook: 247 frontend files / 1,814 tests passed; Rust nextest 2,331 passed, 7 existing skips. All lane gates passed, including typecheck, lint, clippy, architecture drift and red-first plan. The separate actual red-first gate also passed. No hook bypass, exit masking, duplicate passing full lane or weakened test.

NICHT ABGEDECKT

• Linux does not compile/run #[cfg(windows)] tests (run_in_pty, npm_shim, ConPTY, DPAPI) or Windows clippy arms. The Windows lane runs in the merge queue.
• Precommit/prepush exclude Browser-Smoke, frontend build and additional full-linux workflow gates. Public finite browser evidence above is targeted fixture evidence; native runtime, installed/effective behavior, native persistence, keyboard/AT, UAT, full KR-09b/09a′ and Glass-wide acceptance remain open.
• Saved-ceiling design debt remains raw exit 1 (42 original / 39 candidate entries). Only the explicitly Root-dispositioned es-tab-mcp capture-noise cases use that boundary; every notification case is included. No global design-GREEN claim.
• Red-first uses name filtering and thus filters sibling tests; the unfiltered final frontend suite passed all 1,814 tests. Existing platform/capability/ignored skips were not changed. The secret scanner reports zero added bytes for this deletion-only delta; no new source text or secret was introduced.
• Offline PLAN gate does not prove live PR/age reconciliation. The unchanged own row's merge wording requires Root metadata reconciliation; no PLAN edit was authorized. The local lage-jetzt mirror was unavailable; live dispatch, PLAN, own native header and GitHub evidence were used.
• Start-check had no usage/observe-log argument. Separate own native admission observed gpt-6.1-sol/high at 2026-10-27T23:07Z and 37% weekly usage; fresh admission-stage snapshot/rules were read (Codex 34% at 23:02:52Z; operative <95%, native 98% unchanged). Own bounded native goal was created/checked; no prior unmet or parent goal was completed. Supported goal tools cannot edit an objective; milestone progress is recorded here.
• External service health is not proven by these gates. Root's outside-family LAST delta review, final candidate/metadata acceptance, Ready, labels, queue and merge are OPEN. This worker has performed none of those actions.

Next owner and delivery boundary

Root: review ONLY FAA..3200045d84af763fb741fd66b248a1b790482b60 with one outside family as the LAST delta round, confirm the measured correction and retained screenshot/public-behavior evidence, reconcile existing PLAN metadata, then decide final acceptance/Ready/Mergify. This is checked source delivery and handover, not parent completion or merge. The one authorized recut is consumed; no second self-recut was attempted.

---

Retained prior report and source-review history (historical status; current section above supersedes blocker/remedy/noise wording)

Plan-ID: KR-09B-2-DIGEST-LEAF
Ticket: #330; parent #329 / KR-09b. Owner: srv-kr09b2-digest.
State: DRAFT — wrapper-only correction BLOCKED; no new source commit.
Review: ausstehend (Stufe B). Root owns the outside-family bounded delta review and final acceptance; no review was launched by this worker.

Latest correction attempt and observed blocker (2026-10-27)

Root authorized a wrapper/composition-only correction following R1964-a1/a2. The real browser result disproves the premise that the added wrapper causes the extra boxed nesting finding. This section supersedes the older requested wrapper remedy; it does not change acceptance or dismiss the real increase.

• Current and verified remote/PR head remains faa789217e03664c58ceefe1197f882fb64effee, same draft PR feat(settings): use Glass switch for the global daily digest #1964. Original source base is 872b3d3c1a37d4faad2d9a82b4b27aa6c2e98746; prior reviewed source is 59505cedbf5551e4b607a197a8d003c6e8ee0cd8.
• Reproduced the actual compiling browser regression before editing using the unchanged design CLI and an independent literal comparison: classic counts must be 42/49/59 for 0/2/12 items, at both viewports and themes. All 12 cases instead produced 43/50/60. Literal comparison exited 1 as expected before the fix.
• Tried exactly two source-line substitutions: outer div.settings-field to a React Fragment. Existing tests, controlled Switch props, label, full hint and error associations were unchanged. Typecheck and all 26 targeted tests passed, but another complete browser measurement still produced 43/50/60. The same literal comparator remained 1; this is an unsuccessful attempt, not GREEN evidence.
• Computed-style inspection through the real mounted classic settings route (Chromium, existing public Tauri browser fixture) shows both div.settings-field and div.settings-check have transparent background, no shadow and no borders. Neither is counted as a box by the unchanged measurement at scripts/dev/design-gate.mjs:34. Removing either or both transparent wrappers cannot lower this metric.
• The boxed chain is the retained Glass button.g-sw (blue background), existing .settings-view (background rgb(246, 246, 248)) and existing .g-mount-legacy (background rgb(232, 232, 236)): depth 3. Exact-base inspection shows the old checkbox had transparent background/no borders/no shadow and was therefore not a measured box. Those same two boxed ancestors already existed in the base. The added finding is the material Switch itself, not a boxed wrapper.
• Changing the existing outer surfaces, global Switch styling, local CSS, measurement, baseline or acceptance is outside this dispatch. No such change was made. No portal, hiding, unrelated metric reduction or weakened assertion was used.
• Saved the failed source attempt and patch outside Git before restoring it. All 1,854 tracked files match the before-edit SHA-256 manifest, including every source/test file and the own PLAN row. Working tree is clean, package size remains 179 RAW. No commit, push, history rewrite or new PR occurred in this correction attempt.

Finding	Current disposition
R1964-a1	Investigated; outer wrapper can be removed structurally, but this does not close the required nesting comparison. No fix commit delivered.
R1964-a2	Existing control-row wrapper is also unboxed. Changing its semantic tag cannot remove the independently observed material Switch finding. No fix commit delivered.
D1964-1 / M3	Open: all 12 classic cases still increase by one; wrapper-only remedy cannot satisfy the retained Switch plus nonincrease requirements. Root must revise the bounded method/scope or explicitly disposition the acceptance conflict before another source attempt.

Fresh complete design runs

Read-only exact-base archive was checked against all 1,851 archive files before measurement; existing dependencies were reused. Each awaited finite command owned and closed its Vite/browser instance. Commands used only npm run design:gate -- --out <private-output>; no gate, metric or saved ceiling changed. Browser: Chromium 153.0.8010.12. Each run retained all 396 cases: 0/2/12 items, 1440×900/1920×1080, light/dark. The base is honestly a reconstructed original-source baseline measured after implementation.

Source	CLI raw exit	Saved-ceiling failure entries	Elapsed	[slack, nesting, tabs, eyebrows, grids]
Exact original base	1	42	111.4 s	[237, 2669, 132, 376, 25]
Unchanged faa7892 before edit	1	42	114.6 s	[240, 2664, 132, 376, 26]
Uncommitted fragment attempt (failed, restored)	1	40	112.0 s	[234, 2681, 132, 376, 24]

Raw total nesting deltas can be hidden by bad tab captures: unchanged head is -5 overall despite the real classic +12; fragment attempt is +12 overall. Under Root's acknowledged tab-capture boundary, exclude only the MCP/notifications capture routes for aggregate comparison: both unchanged head and fragment attempt are [0, +12, 0, 0, 0]. All 372 other cases remain included; no classic or other real increase is excluded. Therefore nonincrease fails in both attempts.

All 12 classic case vectors follow; red and fragment attempt are identical. Vector order is [slack, nesting, tabs, eyebrows, grids].

Case	Original base	Red = fragment attempt	Delta
/einstellungen/klassisch ∣ 0 ∣ 1440x900 ∣ light	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 0 ∣ 1440x900 ∣ dark	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 0 ∣ 1920x1080 ∣ light	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 0 ∣ 1920x1080 ∣ dark	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 2 ∣ 1440x900 ∣ light	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 2 ∣ 1440x900 ∣ dark	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 2 ∣ 1920x1080 ∣ light	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 2 ∣ 1920x1080 ∣ dark	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 12 ∣ 1440x900 ∣ light	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 12 ∣ 1440x900 ∣ dark	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 12 ∣ 1920x1080 ∣ light	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch ∣ 12 ∣ 1920x1080 ∣ dark	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 1, 0, 0, 0]

Every changed acknowledged tab-capture case is listed below. These are only the previously acknowledged MCP/notifications capture-noise routes; the worker makes no native stability claim. Every case not listed in the classic or tab tables is identical to original base (381 identical for red, 379 identical for fragment).

Run	Changed capture case	Base	Actual	Delta
Red	/einstellungen@es-tab-mcp ∣ 2 ∣ 1440x900 ∣ dark	[0, 11, 0, 0, 0]	[3, 16, 0, 0, 1]	[3, 5, 0, 0, 1]
Red	/einstellungen@es-tab-benachrichtigungen ∣ 12 ∣ 1440x900 ∣ light	[0, 11, 0, 0, 0]	[0, 0, 0, 0, 0]	[0, -11, 0, 0, 0]
Red	/einstellungen@es-tab-benachrichtigungen ∣ 12 ∣ 1920x1080 ∣ dark	[0, 11, 0, 0, 0]	[0, 0, 0, 0, 0]	[0, -11, 0, 0, 0]
Fragment attempt	/einstellungen@es-tab-mcp ∣ 0 ∣ 1440x900 ∣ light	[0, 6, 0, 0, 0]	[3, 16, 0, 0, 1]	[3, 10, 0, 0, 1]
Fragment attempt	/einstellungen@es-tab-mcp ∣ 2 ∣ 1920x1080 ∣ light	[3, 16, 0, 0, 1]	[0, 11, 0, 0, 0]	[-3, -5, 0, 0, -1]
Fragment attempt	/einstellungen@es-tab-mcp ∣ 2 ∣ 1920x1080 ∣ dark	[3, 16, 0, 0, 1]	[0, 11, 0, 0, 0]	[-3, -5, 0, 0, -1]
Fragment attempt	/einstellungen@es-tab-mcp ∣ 12 ∣ 1440x900 ∣ light	[0, 11, 0, 0, 0]	[3, 16, 0, 0, 1]	[3, 5, 0, 0, 1]
Fragment attempt	/einstellungen@es-tab-mcp ∣ 12 ∣ 1920x1080 ∣ light	[3, 16, 0, 0, 1]	[0, 11, 0, 0, 0]	[-3, -5, 0, 0, -1]

Correction-attempt commands and raw exits

Command / observation	Exit
bash scripts/sync.sh start; git fetch origin; live main/PR observations	0 each
npm run dev:start-check -- --max-cargo 2 --min-free-gb 3.5	0
Fresh shared heavy-phase guard before each heavy command	0 on admitted phases; 1 once at load 9.76, no phase started; fresh check passed after 30-second wait
npm run design:gate -- --out <private-base-output>	1
npm run design:gate -- --out <private-red-output>	1
python3 <private-evidence>/compare-classic.py <red-results.json> before editing	1 (12 real expected regression failures)
npm run typecheck on fragment attempt	0
npm run test:unit -- src/components/settings/digest/DigestFields.public.test.tsx src/components/SettingsView.test.tsx src/shell/mount/settings-routing.test.tsx on fragment attempt	0 (3 files, 26 tests)
npm run design:gate -- --out <private-fragment-output>	1
python3 <private-evidence>/compare-classic.py <fragment-results.json>	1 (all 12 still fail)
Private finite mounted-route computed-style inspection, fragment and restored head	0 each
Private exact-base computed-style inspection	1 first (port 5173 busy); 0 after allowing a free port, preserving the failed log
SHA-256 check of all tracked files after restoring the backed-up attempt; git diff --exit-code; git diff --check; 179 RAW/scope/own-row checks	0 each
git ls-remote origin refs/heads/codex/kr09b2-digest; gh pr view 1964	0; exact faa7892 SHA, open draft

Own native header reported gpt-6.1-sol/high in this assigned worktree; own usage showed 31% weekly used, below the operative 95% and unchanged native 98% rule. Attached admission-stage usage/rules were read; own native bounded goal was created and remains unmet. No prior/parent goal was completed. The local lage-jetzt mirror was unavailable; live dispatch, PLAN, native header and GitHub evidence were used.

NICHT ABGEDECKT — current correction attempt

• No successful correction is delivered; M3 fails. No full final-candidate prepush, red-first gate, secret scan/commit or push was run for this failed and restored attempt. Prior source/metadata gate evidence below remains historical and is not claimed as verification of a new fix.
• No inspected screenshot or visual acceptance of a new candidate is claimed. Complete design captures are heuristic/browser fixture evidence; native keyboard/AT, native runtime, installed/effective behavior, persistence, UAT and parent KR-09b/09a remain open.
• Linux does not compile/run #[cfg(windows)] tests or Windows clippy arms; the Windows lane belongs to the merge queue.
• Typecheck and targeted tests apply to the temporary fragment attempt only. Existing public tests were preserved byte-for-byte; no skip/retry/timeout/assertion change was made. No independent review was launched and no finding is claimed closed.

Next owner

Root: resolve the measured scope/acceptance conflict. Preserve Glass Switch and real nonincrease requirements; supply an authorized method reaching the existing material ancestors if a source correction remains required. This worker leaves the exact original clean draft head and private evidence, with no new source delta to review. No Ready, labels, queue, merge, infrastructure or parent completion action occurred.

---

Retained earlier delivery and source-review record (historical)

The following record predates this correction attempt; its previous proposed wrapper remedy and evidence counts are superseded by the observations above. Its original commits and independent review history are preserved.

Plan-ID: KR-09B-2-DIGEST-LEAF
Ticket: #330; parent #329 / KR-09b. Owner: srv-kr09b2-digest.
State: DRAFT — own row delivered; M3 nonincrease BLOCKED.
Review: ausstehend (Stufe C, metadata and final candidate). Source Tier B was accepted by Root in the retained source-review record below; the newly measured design finding has no Root disposition yet.

Candidate, scope and identity

• Original source merge-base: 872b3d3c1a37d4faad2d9a82b4b27aa6c2e98746.
• Frozen reviewed source: 59505cedbf5551e4b607a197a8d003c6e8ee0cd8.
• Final candidate, verified remote and observed PR head: faa789217e03664c58ceefe1197f882fb64effee.
• Main at admission: 86401bd81ec62e5e684903f4d29db18730dfbd86; live ci, Push on main and dev-build were completed/success. No main merge, rebase or history rewrite.
• The only finish-phase tracked change is one nine-cell row in docs/PLAN.md, immediately after KR-09b-1-WEBPORT. It names the existing owner, Root acceptance, 01a/feat(design): add standalone UI baseline and port feedback inputs (KR-01a) #1547 predecessor, 09b/fix(fe): resync global reads on resume #329, Server, source Tier B/metadata Tier C, feat(settings): use Glass switch for the global daily digest #1964 and merge queue: checking #315 + #313 + #318 + #321 together on main (8e439b5) #330. Delivery is conditional on merge; final acceptance stays separate. No foreign row changed.
• git diff --exit-code 59505cedbf5551e4b607a197a8d003c6e8ee0cd8..HEAD -- . ':(exclude)docs/PLAN.md' returned 0. All 1,853 other tracked files, including all five source/test files, match the before-edit SHA-256 manifest.
• Removing only the added own row restores PLAN byte-for-byte to source head; normalized SHA-256: 73efbea71a1d94eef0afb5333d46d2022d24a0671829a4d9a0d68c9ee93cfe1d. All 1,851 files in the reconstructed original-base archive were also verified against the exact archive after measurement.

Path	Added + deleted vs original merge-base
docs/PLAN.md	1
src/components/SettingsView.test.tsx	2
src/components/settings/GeneralTab.tsx	23
src/components/settings/digest/DigestFields.public.test.tsx	110
src/components/settings/digest/DigestFields.tsx	34
src/components/settings/digest/texts.ts	9
Total	179

178 RAW source + 1 RAW metadata; hard ceiling 300, no exception. The metadata commit uses No-Test: document the existing digest delivery and design evidence; both existing red-first commits and their exact test trailers remain intact.

Preserved product behavior and test boundary

GeneralTab delegates only the global digest block to DigestFields using the existing Glass Switch. SettingsView owns the controlled enabled value, callback, error and existing get/set operations. The label, full cadence/path/source/no-agent/no-token/gitignored hint, no-project availability and unique stable label/hint/error associations remain covered by public GeneralTab tests. The separately authorized SettingsView second-writer resync query uses switch, with its unchecked assertion preserved. No source, test, copy, CSS, IPC, backend, persistence, routing, dependency or architecture allowlist changed in this finish phase.

The source phase demonstrated a compiling RED at 9e1fda8d486368bdd044b969b3052d22309ba583 (two intended missing-switch failures), then 26 passing targeted tests. Only Tauri core/event were mocked. No new skip, retry, timeout, snapshot or weakened assertion was added. The final red-first run again verified both exact test names missing at the original merge-base and green at the final candidate.

Exact-base and candidate design measurements — M3 not met

This is a reconstructed original-source baseline, measured after implementation, not a claimed pre-edit run. A read-only git archive 872b3d3c1a37d4faad2d9a82b4b27aa6c2e98746 extraction outside Git reused the existing node_modules; no install or extra Git worktree. The candidate ran in the assigned worktree with source byte-identical to the reviewed head; its only metadata row was subsequently corrected to use the valid predecessor wording. That documentation-only correction changes no measured input. Each required finite gate owned and closed its Vite/browser instance. No desktop/native app or persistent service was started, no production mock was changed, and neither --update nor a nonexistent --base option was used.

Command	Raw exit	Cases	Saved-ceiling failure entries	Elapsed
From exact original-base archive: npm run design:gate -- --out <private-base-output>	1	396	48	116.9 s
From candidate: npm run design:gate -- --out <private-candidate-output>	1	396	47	97.4 s

Both runs used Chromium 153.0.8010.12, 0/2/12 items, 1440×900/1920×1080, light/dark. Raw exits remain failures; fewer saved-ceiling entries do not prove nonincrease against the actual original base. No committed ceiling was changed.

Metric	Original base	Candidate	Candidate − base
Slack	249	246	-3
Nesting depth >2	2687	2700	+13
Empty tabs	132	132	0
Eyebrows	376	376	0
1fr grids	29	28	-1

All changed per-case vectors follow; vector order is [slack, nesting, tabs, eyebrows, grids]. The other 381 cases are identical. There are 16 increasing metric cells, so M3 fails even where another metric decreases.

Case: route ∣ items ∣ viewport ∣ theme	Original base	Candidate	Delta
/einstellungen/klassisch∣0∣1440x900∣light	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[0, 1, 0, 0, 0]
/einstellungen@es-tab-mcp∣0∣1440x900∣dark	[3, 16, 0, 0, 1]	[0, 6, 0, 0, 0]	[-3, -10, 0, 0, -1]
/einstellungen@es-tab-benachrichtigungen∣0∣1440x900∣dark	[0, 0, 0, 0, 0]	[0, 6, 0, 0, 0]	[0, 6, 0, 0, 0]
/einstellungen/klassisch∣0∣1440x900∣dark	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[0, 1, 0, 0, 0]
/einstellungen@es-tab-mcp∣0∣1920x1080∣light	[0, 6, 0, 0, 0]	[3, 16, 0, 0, 1]	[3, 10, 0, 0, 1]
/einstellungen/klassisch∣0∣1920x1080∣light	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch∣0∣1920x1080∣dark	[1, 42, 0, 15, 0]	[1, 43, 0, 15, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch∣2∣1440x900∣light	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch∣2∣1440x900∣dark	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch∣2∣1920x1080∣light	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch∣2∣1920x1080∣dark	[1, 49, 0, 17, 0]	[1, 50, 0, 17, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch∣12∣1440x900∣light	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch∣12∣1440x900∣dark	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 1, 0, 0, 0]
/einstellungen@es-tab-mcp∣12∣1920x1080∣light	[3, 16, 0, 0, 1]	[0, 11, 0, 0, 0]	[-3, -5, 0, 0, -1]
/einstellungen/klassisch∣12∣1920x1080∣light	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 1, 0, 0, 0]
/einstellungen/klassisch∣12∣1920x1080∣dark	[0, 59, 0, 20, 0]	[0, 60, 0, 20, 0]	[0, 1, 0, 0, 0]

D1964-1 — needs Root disposition: /einstellungen/klassisch adds exactly one nesting finding in each of its 12 cases (aggregate 600→612). For 0 items/light/1440×900 it is 42→43; details add the digest control #:rp: at depth 3. The measured element is the existing Glass Switch used at src/components/settings/digest/DigestFields.tsx:18, rendered by src/design/controls/Switch.tsx:12. This is a measured heuristic increase; no visual defect or native failure is inferred. Other changed vectors occur in captured settings sub-tabs and remain unexplained; no stability or causal claim is made for them. The worker preserved frozen source and did not change the gate, styles, tests or acceptance to obtain a pass. Root/Chief must disposition the finding and authorize any further source work; M3 is not satisfied.

Finish-phase evidence — foreground commands and raw exits

Actual own native turn_context at 2026-10-27T21:33:41.155Z reported gpt-6.1-sol, high, in the assigned worktree before editing. Own get_goal initially returned null and create_goal created this bounded finish goal in session 01a17f48-7f63-7db3-a39e-b3569356cb07. It remains active because M3 is unmet; no parent or prior goal was completed. The local lage-jetzt mirror was unavailable; current dispatch, canonical PLAN, native header and live GitHub evidence were used.

Every heavy phase had a fresh guard.py srv-kr09b2-digest 7 && <phase> in the same awaited foreground shell, slot 7 and one Cargo build job, without profile overrides. Busy/load guards refused admission with exit 1; required waits and fresh checks preceded later passing admission. Nothing was detached. Private evidence and original reports remain outside the repository.

Command / check	Exit
bash scripts/sync.sh start; git fetch origin; live main/PR observations	0 each
npm run dev:start-check -- --max-cargo 2 --min-free-gb 3.5 — RAM 12.5 GB, 0 Cargo builds; native model/limit observed separately	0
git diff --check; own nine-cell/scope/179 RAW/foreign-byte/normalized-PLAN/archive identity checks	0 each
bash scripts/ci/secret-scan.sh before each commit attempt	0 each
First git commit -F <message-file> — precommit rejected the row’s nonexistent KR-01a PLAN dependency; no commit created	1
Corrected own row only; git commit -F <message-file> — unchanged full precommit including doc-links and offline plan-check	0
bash scripts/ci/red-first.sh at faa789217e03664c58ceefe1197f882fb64effee	0
git push -u origin codex/kr09b2-digest — unchanged active hook ran bash scripts/ci/gates.sh lane prepush at faa789217e03664c58ceefe1197f882fb64effee in the assigned worktree; 292.3 seconds including push	0
git ls-remote origin refs/heads/codex/kr09b2-digest — full final SHA matched	0
git status --short — clean; observed gh pr view 1964 final head matched and remained draft	0

Final full prepush: 247 frontend files / 1,814 tests passed; Rust nextest 2,331 passed, 7 existing skips. Every prepush gate passed, including doc-links, plan-check, typecheck, lint, clippy and Rust suite. It was run once on this final candidate through the normal push hook; no redundant full lane. Architecture drift: no new occurrences, no allowlist change. The separate design check above failed and remains explicitly outside any claim that this candidate meets M3.

NICHT ABGEDECKT

• Linux does not compile/run the #[cfg(windows)] tests (run_in_pty, npm_shim, ConPTY, DPAPI) or Windows clippy arms. The Windows lane runs in the merge queue.
• Precommit and prepush exclude Browser-Smoke, frontend build and additional workflow gates of the full linux lane. Offline plan-check excludes live PR/plan-age reconciliation and running-job checks. Start-check had no usage/observe-log argument; separate native/coordinator observations supplied bounded admission evidence.
• First precommit stopped at the dependency diagnostic; six later gates did not run in that failed attempt. The corrected full precommit and final full prepush both passed. Red-first filters sibling tests; the unfiltered final frontend suite ran all 1,814 tests. Existing platform/capability/ignored skips remain unchanged.
• No external-service health (Updater endpoint, OmniRoute, Mirror), native keyboard/AT, native runtime, persistence/UAT, installed/effective behavior, performance acceptance or full Glass acceptance is claimed. The source handover retained inspected static light/dark fixtures; new design captures are DOM-heuristic/mock evidence, not native or visual acceptance.
• M3 nonincrease is NOT covered/met: both design commands exited 1; nesting increases by 13 overall and by 12 on the classic settings route. Other sub-tab differences are not explained or treated as stable measurements.
• Final Tier C metadata/candidate acceptance, Ready, merge, full KR-09b/09a′, native/AT/installation/UAT and parent goals remain open.

Handover

Next owner: Root, for Tier C row/candidate review and disposition of D1964-1/M3, preserving the distinction from the accepted source Tier B reviews below. Further source edits require Root/Chief authorization; none were attempted. No Ready change, review launch, label, queue action, new PR or parent completion was performed. The checked metadata commit is delivered and pushed, while the full finish acceptance is blocked by M3. Historical source-stop diagnostics and the source handover report are preserved.

Root review record (2026-10-27, source head 59505ce)

Retained source-stage record: its then-pending own row is delivered above; the later measured M3 finding remains unresolved. The recorded source verdicts and dispositions are unchanged.

Candidate source head 59505cedbf5551e4b607a197a8d003c6e8ee0cd8 (178 RAW, 5 files, no seam). Tier B; author family OpenAI (Codex gpt-6.1-sol). Scope: Root authorized exactly one extra file at 21:03Z, SettingsView.test.tsx:356, changing the query from checkbox to switch with .not.toBeChecked() kept.

ID	Source	Model (observed)	Verdict
R1964-K	Kimi CLI 2.3.2 on the PC, 21:28:10–21:29:40Z, exit 0	kimi-code/k3.1-pro (Moonshot)	APPROVE (none; 12 checks)
R1964-G	.pa/review_transport.py, Ollama, 21:23:18Z, exit 0	glm-5.2:cloud (Zhipu; self-label differs, noted)	APPROVE (2 low; 12 checks)

ID	Sev	Finding	Root disposition
R1964-a1	low	DigestFields.tsx:16: the new settings-field wrapper may change spacing.	Follow-up / visual check. It reuses the existing neighbouring settings class for consistency, and no new CSS was added.
R1964-a2	low	DigestFields.tsx:17: settings-check moved from <label> to <div>, so element-scoped selectors could miss.	Follow-up. The Switch brings its own g-sw design styling. Both reviewers confirmed no new local styles.

Source accepted by Root. Before ready, still required: its own PLAN row in this PR (the user’s rule 2 of 27.10.: no separate mini-PLAN PR) and the design:gate before/after delta in the PR text. The PLAN-row commit is checked as a Tier-C delta, then review-ok is set on the final head.

Root review record — round 2 (bounded LAST delta, 2026-10-27 23:42Z)

Delta faa78921..3200045d: src/styles.css −1 line (background: var(--bg) removed from the existing .settings-view rule). Reviewer glm-5.2:cloud via .pa/review_transport.py (Zhipu; self-labels "codex-reviewer-o4-mini" in its text, transport reports glm-5.2), one non-author family. Verdict APPROVE WITH CONDITIONS, 10 checks.

ID	Severity	Finding	Disposition
R1964-D1	low	--bg (= --color-content, styles.css:98) differs from the .g-mount-legacy surface --color-window (styles.css:38/260) → slight tone shift inside classic settings	accepted as known, no change: verified tokens differ; the classic route is being retired (GI02), the author retained light/dark screenshots, and contrast-check is part of prepush (exit 0)
R1964-D2	low	.settings-view outside .g-mount-legacy would lose its surface	rejected: verified the only mount is App.tsx:1139, inside the legacy mount (ShellMount.tsx:102)

Design (D1964-1 / M3): all 12 /einstellungen/klassisch cases back at their ceilings; every non-MCP case non-increasing (author report srv-kr09b2-digest-recut-once.md). Design accepted. PLAN row present. Decision: accepted; Ready + review-ok bound to 3200045d84af763fb741fd66b248a1b790482b60.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-27T23:38:45.997910Z	3200045	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[mergify]

✅ Merged into main as 9ba7b54 at 23:53 UTC

8 min in the queue, 8 min of it CI.

So far

• ✅ Entered the queue Tue 27 Oct, 23:45 · rule default · triggered by merge protections
• ✅ Passed · draft merge queue: checking #1964 on main (416d5a7) #1976
• ✅ Merged 23:53 · 9ba7b54 · merge

Merge conditions: 9 of 9 met

This pull request

	Condition	Required by
✅	github-review-approved	GitHub repository ruleset rule main-pr-required-checks
✅	label != do-not-merge	—

The batch, tested on draft #1976

	Condition	Required by
✅	check-success = gates (linux)	—
✅	check-success = gates (windows)	—
✅	check-success = red-first	—
✅	any of: check-success = @github-actions/gates (linux), check-neutral = @github-actions/gates (linux), check-skipped = @github-actions/gates (linux)	GitHub repository ruleset rule main-pr-required-checks
✅	any of: check-success = @github-actions/gates (windows), check-neutral = @github-actions/gates (windows), check-skipped = @github-actions/gates (windows)	GitHub repository ruleset rule main-pr-required-checks
✅	any of: check-success = @github-actions/red-first, check-neutral = @github-actions/red-first, check-skipped = @github-actions/red-first	GitHub repository ruleset rule main-pr-required-checks
✅	any of: check-neutral = @mergify/Mergify Merge Protections, check-skipped = @mergify/Mergify Merge Protections, check-success = @mergify/Mergify Merge Protections	GitHub repository ruleset rule main-pr-required-checks

_{Times in UTC · dashboard · how the merge queue works}