{row.kind === "comment" ? "Comment content · Links and status claims are supplied by the commenter, not verified deployment evidence." : `Submitted review · revision ${row.commit?.slice(0, 8) ?? "unknown"}`}
>}
{row.kind === "deployment" &&
Provider record · revision {row.commit.slice(0, 8)}{currentHead && row.commit !== currentHead ? " · Old revision" : ""} · Open deployment
}
;
diff --git a/apps/desktop/src/renderer/code-review/PrPane.tsx b/apps/desktop/src/renderer/code-review/PrPane.tsx
index 66ea0cb9..47a09a79 100644
--- a/apps/desktop/src/renderer/code-review/PrPane.tsx
+++ b/apps/desktop/src/renderer/code-review/PrPane.tsx
@@ -34,7 +34,7 @@ interface Props {
contextBusy: ReadonlySet;
onTreeResize: (width: number) => void;
onTreeReset: () => void;
- writes?: Writes;
+ writes: Writes;
agent?: Agent;
}
diff --git a/apps/desktop/src/shared/code-review.ts b/apps/desktop/src/shared/code-review.ts
index aa9631a2..5d1437b0 100644
--- a/apps/desktop/src/shared/code-review.ts
+++ b/apps/desktop/src/shared/code-review.ts
@@ -115,6 +115,8 @@ export type PrActivity =
| { kind: "commit"; id: string; at: string; sha: string; headline: string; author: PrActor | null }
| { kind: "opened"; id: string; at: string; actor: PrActor }
| { kind: "comment"; id: string; at: string; actor: PrActor; /** Untrusted Markdown, like `PrOverview.body`. */ body: string }
+ /** A comment on a line of the diff (GitHub's review comment). `line` is null once the line is no longer in the diff. `replyTo` is the comment it answers. Whether its thread is resolved is not read. */
+ | { kind: "line-comment"; id: string; at: string; actor: PrActor; body: string; path: string; subject: "line" | "file"; line: number | null; side: "LEFT" | "RIGHT" | null; replyTo: string | null }
| { kind: "review"; id: string; at: string; actor: PrActor; verdict: "approved" | "changes-requested" | "commented" | "dismissed"; body: string; commit: string | null }
| { kind: "event"; id: string; at: string; actor: PrActor | null; event: string }
| { kind: "deployment"; id: string; at: string; environment: string; status: "success" | "pending" | "failure" | "inactive" | "unknown"; url: string | null; commit: string };
diff --git a/apps/desktop/test/code-review-read-service.test.ts b/apps/desktop/test/code-review-read-service.test.ts
index 4b33c842..8efc3749 100644
--- a/apps/desktop/test/code-review-read-service.test.ts
+++ b/apps/desktop/test/code-review-read-service.test.ts
@@ -53,6 +53,7 @@ class FakeGitHub {
["7301#74", pull(LENDING, 74, { base: "topic-73", baseSha: sha("c"), head: sha("e") })],
]);
timeline = new Map();
+ lineComments = new Map();
files = new Map();
checkRuns = new Map();
statuses = new Map();
@@ -146,6 +147,10 @@ class FakeGitHub {
}
const id = `${repo.id}#${parts[4]}`;
if (parts[3] === "pulls" && parts.length === 5) return this.pulls.has(id) ? this.json(call, this.pulls.get(id)) : this.missing();
+ if (parts[3] === "pulls" && parts[5] === "comments") {
+ const { slice, link } = this.page(this.lineComments.get(id) ?? [], query, path!);
+ return this.json(call, slice, link);
+ }
if (parts[3] === "pulls" && parts[5] === "files") {
const { slice, link } = this.page(this.files.get(id) ?? [], query, path!);
return this.json(call, slice, link);
@@ -1162,3 +1167,142 @@ test("real use: the merge base and the file page are read together between the t
};
await together({ op: "files", key, revision, cursor: null }, /\/compare\//, /\/pulls\/73\/files\?/, "the file page is asked for while the merge base is still being read");
});
+
+// --- Real-use benchmark, 2026-11-12: comments on lines of the diff were not shown (F3) -----------
+
+const lineCommentRow = (id: number, extra: Record = {}) => ({ id, user: { id: 5, login: "bea-fixture", type: "User" }, created_at: "2026-10-19T09:00:00Z",
+ body: "Should this be `>=`?", path: "policy/schema.ts", line: 41, side: "RIGHT", ...extra });
+
+test("real use: comments on lines of the diff arrive with the first activity page, with their file, line and reply", async () => {
+ const { github, read, prime } = setup();
+ await prime();
+ github.lineComments.set("7301#73", [lineCommentRow(801), lineCommentRow(802, { in_reply_to_id: 801, body: "Yes.", created_at: "2026-10-19T09:05:00Z" }), lineCommentRow(803, { line: null, side: "LEFT" })]);
+ const reply = await read({ op: "activity", key: keyOf(LENDING, 73), cursor: null });
+ assert.equal(reply.result.state, "ready");
+ const rows = value(reply).filter(r => r.kind === "line-comment");
+ assert.deepEqual(rows.map(r => r.kind === "line-comment" && [r.id, r.path, r.line, r.side, r.replyTo, r.body, r.actor.login]), [
+ ["801", "policy/schema.ts", 41, "RIGHT", null, "Should this be `>=`?", "bea-fixture"],
+ ["802", "policy/schema.ts", 41, "RIGHT", "801", "Yes.", "bea-fixture"],
+ ["803", "policy/schema.ts", null, "LEFT", null, "Should this be `>=`?", "bea-fixture"],
+ ]);
+ const asked = github.calls.filter(c => c.endpoint.includes("/pulls/73/comments"));
+ assert.deepEqual(asked.map(c => c.endpoint.split("/pulls/")[1]), ["73/comments?per_page=100&page=1"]);
+});
+
+test("real use: a refused or malformed line-comment read leaves the timeline usable and says the page is partial", async () => {
+ const { github, read, prime } = setup();
+ await prime();
+ github.overrides.push({ pattern: /\/pulls\/73\/comments\?/, answer: { kind: "http", status: 403, headers: {}, body: Buffer.from("{}") } });
+ const refused = await read({ op: "activity", key: keyOf(LENDING, 73), cursor: null });
+ assert(refused.result.state === "partial");
+ assert.equal(refused.result.reason, "denied");
+ assert.equal(refused.result.coverage.complete, false);
+ assert(value(refused).some(r => r.kind === "opened"));
+
+ github.overrides.length = 0;
+ github.lineComments.set("7301#73", [lineCommentRow(801), { id: 802, body: "no author, file or date" }]);
+ const malformed = await read({ op: "activity", key: keyOf(LENDING, 73), cursor: null });
+ assert(malformed.result.state === "partial");
+ assert.equal(malformed.result.reason, "failed");
+ assert.deepEqual(value(malformed).filter(r => r.kind === "line-comment").map(r => r.id), ["801"]);
+});
+
+test("real use: more than one page of line comments is reported as capped, and later activity pages do not ask again", async () => {
+ const { github, read, prime } = setup();
+ await prime();
+ github.lineComments.set("7301#73", Array.from({ length: 101 }, (_, i) => lineCommentRow(1000 + i)));
+ const first = await read({ op: "activity", key: keyOf(LENDING, 73), cursor: null });
+ assert(first.result.state === "partial");
+ assert.equal(first.result.reason, "capped");
+ assert.equal(value(first).filter(r => r.kind === "line-comment").length, 100);
+ github.timeline.set("7301#73", Array.from({ length: 150 }, (_, i) => ({ event: "labeled", id: 5000 + i, created_at: "2026-10-19T10:00:00Z", actor: { id: 5, login: "bea-fixture", type: "User" } })));
+ const before = github.calls.filter(c => c.endpoint.includes("/pulls/73/comments")).length;
+ await read({ op: "activity", key: keyOf(LENDING, 73), cursor: "p2" });
+ assert.equal(github.calls.filter(c => c.endpoint.includes("/pulls/73/comments")).length, before);
+});
+
+
+test("inline comment integration: a held private comment cannot survive whole-PR revocation or become an offline answer", async () => {
+ const { github, read, prime } = setup();
+ await prime();
+ const key = keyOf(LENDING, 73);
+ github.lineComments.set("7301#73", [lineCommentRow(801, { body: "INLINE_PRIVATE_REVOKED_SENTINEL" })]);
+ const release = github.hold(/\/pulls\/73\/comments\?/);
+ const pending = read({ op: "activity", key, cursor: null });
+ try {
+ await until(() => github.calls.some(call => call.endpoint.includes("/pulls/73/comments?")));
+ github.overrides.push({ pattern: /^repos\/example-org\/lending\/pulls\/73$/, answer: { kind: "http", status: 403, headers: {}, body: Buffer.from("{}") } });
+ assert.equal(reason(await read({ op: "overview", key })), "denied");
+ } finally { release(); }
+ const late = await pending;
+ assert.equal(late.result.state, "unavailable");
+ assert.ok(!JSON.stringify(late).includes("INLINE_PRIVATE_REVOKED_SENTINEL"));
+ github.cli = "offline";
+ const offline = await read({ op: "activity", key, cursor: null });
+ assert.equal(offline.result.state, "unavailable");
+ assert.ok(!JSON.stringify(offline).includes("INLINE_PRIVATE_REVOKED_SENTINEL"));
+});
+
+test("inline comment integration: a held old-account comment is refused and never kept for the replacement account", async () => {
+ const { github, read, prime } = setup();
+ await prime();
+ const key = keyOf(LENDING, 73);
+ github.lineComments.set("7301#73", [lineCommentRow(801, { body: "INLINE_PRIVATE_ACCOUNT_SENTINEL" })]);
+ const release = github.hold(/\/pulls\/73\/comments\?/);
+ const pending = read({ op: "activity", key, cursor: null });
+ try {
+ await until(() => github.calls.some(call => call.endpoint.includes("/pulls/73/comments?")));
+ github.user = KIT;
+ } finally { release(); }
+ const late = await pending;
+ assert.equal(late.result.state, "unavailable");
+ assert.equal(reason(late), "signed-out");
+ assert.ok(!JSON.stringify(late).includes("INLINE_PRIVATE_ACCOUNT_SENTINEL"));
+ const replacement = await read({ op: "activity", key, cursor: null });
+ assert.equal(reason(replacement), "signed-out");
+ assert.ok(!JSON.stringify(replacement).includes("INLINE_PRIVATE_ACCOUNT_SENTINEL"));
+});
+
+
+test("inline comment metadata: file subjects and unavailable line positions retain their distinct provenance", async () => {
+ const { github, read, prime } = setup();
+ await prime();
+ github.lineComments.set("7301#73", [lineCommentRow(901, { subject_type: "file", line: null, side: null }), lineCommentRow(902, { subject_type: "line", side: "LEFT" }), lineCommentRow(903, { line: null, side: null })]);
+ const reply = await read({ op: "activity", key: keyOf(LENDING, 73), cursor: null });
+ assert.equal(reply.result.state, "ready");
+ const rows = value(reply).filter(r => r.kind === "line-comment");
+ assert.deepEqual(rows.map(row => ({ ...row })), [
+ { kind: "line-comment", id: "901", at: "2026-10-19T09:00:00Z", actor: { id: "5", login: "bea-fixture", bot: false }, body: "Should this be `>=`?", path: "policy/schema.ts", subject: "file", line: null, side: null, replyTo: null },
+ { kind: "line-comment", id: "902", at: "2026-10-19T09:00:00Z", actor: { id: "5", login: "bea-fixture", bot: false }, body: "Should this be `>=`?", path: "policy/schema.ts", subject: "line", line: 41, side: "LEFT", replyTo: null },
+ { kind: "line-comment", id: "903", at: "2026-10-19T09:00:00Z", actor: { id: "5", login: "bea-fixture", bot: false }, body: "Should this be `>=`?", path: "policy/schema.ts", subject: "line", line: null, side: null, replyTo: null },
+ ]);
+});
+
+test("inline comment metadata: malformed locations make coverage partial and never become an inferred current side", async () => {
+ const bad = [{ line: "41" }, { line: -1 }, { line: 0 }, { line: 1.5 }, { line: undefined }, { side: "MIDDLE" }, { side: undefined }, { subject_type: "unknown" }, { subject_type: "file", line: "1", side: "RIGHT" }, { subject_type: "file", line: 1, side: "MIDDLE" }];
+ for (const patch of bad) {
+ const { github, read, prime } = setup();
+ await prime();
+ github.lineComments.set("7301#73", [lineCommentRow(901), lineCommentRow(902, patch)]);
+ const reply = await read({ op: "activity", key: keyOf(LENDING, 73), cursor: null });
+ assert.equal(reply.result.state, "partial", JSON.stringify(patch));
+ assert(reply.result.state === "partial");
+ assert.equal(reply.result.reason, "failed");
+ assert.equal(reply.result.coverage.complete, false);
+ assert.deepEqual(value(reply).filter(r => r.kind === "line-comment").map(r => r.id), ["901"]);
+ }
+});
+
+
+test("inline comment metadata: a file target may carry valid line and side metadata without becoming a line comment", async () => {
+ const { github, read, prime } = setup();
+ await prime();
+ // GitHub's public API returns file subjects with line:1/side:RIGHT as well as omitted locations.
+ github.lineComments.set("7301#73", [lineCommentRow(911, { subject_type: "file", line: 1, side: "RIGHT" }), lineCommentRow(912, { subject_type: "file", line: undefined, side: undefined })]);
+ const reply = await read({ op: "activity", key: keyOf(LENDING, 73), cursor: null });
+ assert.equal(reply.result.state, "ready");
+ const rows = value(reply).filter(row => row.kind === "line-comment");
+ assert.deepEqual(rows.map(row => ({ subject: row.subject, line: row.line, side: row.side })), [
+ { subject: "file", line: 1, side: "RIGHT" }, { subject: "file", line: null, side: null },
+ ]);
+});
diff --git a/apps/desktop/test/fixtures/code-review-gh-world.ts b/apps/desktop/test/fixtures/code-review-gh-world.ts
index 3a6d52b0..48b787f3 100644
--- a/apps/desktop/test/fixtures/code-review-gh-world.ts
+++ b/apps/desktop/test/fixtures/code-review-gh-world.ts
@@ -144,6 +144,8 @@ if (method === "PATCH" || (method === "POST" && (route === "graphql" || /\/reque
&& value && value.base && value.head && value.base.sha === comparison[2] && value.head.sha === comparison[3]);
if (declared) found = { base_commit: { sha: comparison[2] }, merge_base_commit: { sha: comparison[2] } };
}
+ // A declared pull request with no line comments listed has none.
+ if (found === undefined && /\/pulls\/\d+\/comments$/.test(route) && world.routes[route.replace(/\/comments$/, "")]) found = [];
if (found && /\/pulls\/\d+$/.test(route)) found = changed(route, found);
if (found && /\/pulls\/\d+$/.test(route) && mergedAs(route)) {
found = { ...found, state: "closed", merged: true, merged_at: "2026-10-18T20:40:00Z", merged_by: { ...world.user, type: "User" },
diff --git a/docs/design/code-review.md b/docs/design/code-review.md
index e1515479..73527f7d 100644
--- a/docs/design/code-review.md
+++ b/docs/design/code-review.md
@@ -171,7 +171,7 @@ Every call is `gh api --method GET` with validated, encoded segment
| `inbox-page` | `group`, `query`, `cursor` | `PrSummary[]` | the same search, next page; GitHub serves 1,000 results, so `p34` is the last |
| `resolve` | `input` (a pasted link) | `{ locator, key }` | `parsePrLocator`, then `repos/{owner}/{repo}` for the stable id and `pulls/{n}` to check the PR exists |
| `overview` | `key` | `PrOverview` with its `revision` | `repos/{owner}/{repo}/pulls/{n}` |
-| `activity` | `key`, `cursor` | `PrActivity[]` | `repos/{owner}/{repo}/issues/{n}/timeline`, 100 a page; the first page adds the opening from `pulls/{n}` and up to 20 current-head provider deployments from `deployments?sha={headSha}`, each with its latest status |
+| `activity` | `key`, `cursor` | `PrActivity[]` | `repos/{owner}/{repo}/issues/{n}/timeline`, 100 a page; the first page adds the opening from `pulls/{n}` up to 20 current-head provider deployments from `deployments?sha={headSha}`, each with its latest status, and the first 100 comments on lines of the diff from `pulls/{n}/comments` (more is reported as capped). Whether a comment's thread is resolved is not read: GitHub gives that only over GraphQL |
| `checks` | `key`, `revision` | `PrCheck[]` for that head | `commits/{headSha}/check-runs` and `commits/{headSha}/status`, 100 a page, up to 10 pages each; `reporter` is a run's `app.slug` or a status's `creator.login` |
| `stack` | `key` | `PrStack`, labelled `inferred` | CR-4 reads up to 1,000 open PRs with `repos/{owner}/{repo}/pulls?state=open`; equal base/head branch refs within the same stable repository form the inferred relationship. Branching, cycles, a capped read, forks or an unresolved root yield unavailable ordering. |
| `files` | `key`, `revision`, `cursor` | `PrFiles` | `repos/{owner}/{repo}/pulls/{n}/files`, 100 a page (GitHub stops at 3,000 files, so `p30` is the last) |
@@ -625,7 +625,7 @@ comparison and human keyboard/VoiceOver acceptance remain open gates.
Descriptions and comment bodies use the existing inert Markdown renderer. No avatar or remote image is loaded. Provider deployment cards come from GitHub's [deployment list](https://docs.github.com/en/rest/deployments/deployments#list-deployments) filtered by head SHA, with the latest [deployment status](https://docs.github.com/en/rest/deployments/statuses#list-deployment-statuses); only `environment_url` becomes the preview destination. Main constructs fixed GET endpoints from its repository and validated numeric deployment IDs. Comment tables and their URLs retain an explicit commenter-provenance label. Preview links show the actual host, reject credentials/control characters/unsafe schemes, and use the existing external-navigation path on user action.
-The title uses 22px/29px and prose uses the existing 14px token with a 24px pitch. At reference width the main column has a 156px left inset and 40px gap to the 305px sidebar, with a 166px trailing margin. The sidebar scrolls independently. At narrow widths, details follow activity in one readable scroll region. The Changes summary opens the existing Changes tab. Check rows expand natively, show elapsed time at the read timestamp and expose all checks on request. Comment menus offer Copy link only. Edit, reviewer and comment callbacks are explicit `OverviewActions` slots and remain unavailable through `capabilities()` until CR-7 connects them. The PR-question composer remains CR-6's; content reserves bottom clearance for that later host.
+The title uses 22px/29px and prose uses the existing 14px token with a 24px pitch. At reference width the main column has a 156px left inset and 40px gap to the 305px sidebar, with a 166px trailing margin. The sidebar scrolls independently. At narrow widths, details follow activity in one readable scroll region. The Changes summary opens the existing Changes tab. Check rows expand natively, show elapsed time at the read timestamp and expose all checks on request. Comment menus offer Copy link only. Edit, reviewer and comment controls use the connected `Writes` boundary, supplied by the Code Review host; current capabilities and main-owned preview/send checks determine availability. The PR-question composer remains CR-6's; content reserves bottom clearance for that later host.
These are implementation decisions within the accepted contract, not original-reference or human accessibility acceptance. The original R1/R2 images were absent from the manifest paths on 2026-10-23; synthetic normal, narrow, stale and error captures are separate evidence.
diff --git a/docs/reviews/2026-11-13-cr-inline-integration.md b/docs/reviews/2026-11-13-cr-inline-integration.md
new file mode 100644
index 00000000..05336ca6
--- /dev/null
+++ b/docs/reviews/2026-11-13-cr-inline-integration.md
@@ -0,0 +1,45 @@
+# Code Review line-comment integration
+
+This owned integration starts from protected main `927b443414befe5f4300bb3f3d39af4666032609` and merges ready PR #513 source `c599dd7a9fbaeb4d6946666d8859e892e6847aca`, preserving its history and authors. The original owner branch and worktree are unchanged. Git independently reports only a `docs/status.md` conflict; this branch retains the full current-main file and applies narrow, evidence-backed status additions instead of reviving closed #378 or discarding delivered LM/CF/browser facts.
+
+The feature reads the first 100 REST diff-line comments into Activity, preserves line/file target provenance, labels file/line/replies and unavailable current-line positions, and reports partial, malformed or capped reads. Comment text remains untrusted Markdown. Resolved-thread state, placement beside changed lines and later comment pages are not implemented by this slice and remain part of their accepted parent requirements. No default activation or new live-write authority is implied.
+
+## First hosted iPhone job cancellation
+
+Source `249154d57f7f4f849b79f87ea0fd259768ce6a86` started [CI run 43360620861](https://github.com/TypeSafeAI/modex/actions/runs/43360620861). GitHub's annotation confirms that its iPhone job exceeded the configured 30-minute maximum; no unexpected UI assertion failure is recorded before cancellation. Setup, build and native units consumed roughly 14.5 minutes before the paired UI suite started. The native-unit result bundle independently reports 65 passes, zero skips and zero failures. The UI log records 39 passed cases, including one intentional expected-failure control, before cancellation during another case. Its paired result bundle was not finalized and has no Info.plist; the complete UI suite is unverified, not green. The first raw log, cancellation annotations and 13,539,345-byte artifact are retained with independent API-digest and ZIP-integrity verification.
+
+Eight preceding completed PR runs were inspected: seven iPhone jobs passed in approximately 9.9–20.3 minutes and one earlier job was cancelled. This run's cutoff is a confirmed whole-job budget failure; slower setup and UI execution are observed, but no application defect or single resource cause is established. The iPhone job budget is raised to a bounded 45 minutes, leaving about 30 minutes for UI execution after similarly slow setup. Test cases, individual waits, assertions and retry behavior are unchanged. The original required macOS job finished successfully before publication of this correction, with its first result retained separately below. A fresh hosted run at the corrected head must complete before source qualification; the cancelled run remains retained and is not converted into a pass.
+
+The original required macOS job `131672985266` passed 2,587 units with 36 skips and zero failures/cancellations, 533 desktop cases with fourteen opt-in skips in 36.7 minutes, eight site cases in 20.0 seconds and four web cases in 4.4 seconds, without retry annotations. Its actual checkout was merge preview `c34a26f9b0d11155d6206db666ef2986de30b4bf`, parents `927b443414befe5f4300bb3f3d39af4666032609` and `249154d57f7f4f849b79f87ea0fd259768ce6a86`, tree `552d9b575c5006862872e7301044292a16084e76`. Root independently reproduced that tree and retained all four source-run artifacts with API SHA-256 and ZIP-integrity checks. The fourteen desktop skips remain opt-in gates, not acceptance passes; the cancelled iPhone job prevents a whole-source-green claim.
+
+Both existing workflow setup regression tests pass. Local actionlint 1.7.12 reports the same pre-existing unsupported `code-quality` permission scope on the baseline and corrected workflow; with only that exact diagnostic excluded, the remaining workflow lint passes. The permission is unchanged. All 852 previously qualified runtime/test/dependency producer hashes remain byte-identical, so the complete local and separate native receipts below retain their original scope; this workflow-only repair still requires fresh corrected-head hosted qualification.
+
+The separate scoped review of the CI correction at signed `7bce382a11b95ca6ae04b76c264249a8d28e1395` found zero Critical, Important or Minor findings, with none deferred. It confirmed that only the iPhone whole-job watchdog changes, and leaves fresh corrected-head hosted qualification and actual protected delivery required. The final receipt update changes documentation only; the reviewed workflow and all 852 runtime/test/dependency inputs stay identical.
+
+Fresh integrated build and typecheck pass. All 55 read-service cases pass, including two new held-comment controls: a whole-PR denial cannot expose or retain a late private comment as offline data, and an account switch refuses the late old-account comment and subsequent old-key read. All fifteen actual Electron regression cases pass in 41.7 seconds with zero retries, including the inline-comment/filter/inert-content case and the existing revocation/path tests. This found no production defect requiring a repair; legitimate comment-subresource denial still returns an explicitly partial timeline. One fresh whole-branch review found no Critical or Important defects and three Minor defects; all three are corrected below. Full corrected-source repository units/E2E, signed recovery, source CI and protected delivery remain pending. Existing source PR checks do not qualify this new integration. Canonical resource invalidation and final account confirmation were inspected and directly exercised by the two held-comment controls; a comment-subresource denial legitimately leaves a partial timeline, while whole-PR denial or account replacement refuses late private comment data. The independent CR-8 delivery audit in this branch records already-delivered evidence and does not depend on this feature's pending delivery.
+
+## Independent review correction
+
+One fresh read-only review of signed integration `38bab0968a36b9fdc75dc35ea0819984dcd32bfe` found no Critical or Important defect and three Minor defects. M1 identified that GitHub file-level comments were labeled as obsolete line comments. M2 identified malformed/missing location values being coerced into apparently valid null/RIGHT metadata. M3 identified current design prose naming the removed `OverviewActions` callback interface. All were accepted in one coalesced review correction; the pre-push schema audit below completes the file-target coverage. No Minor finding is deferred, and no second review was requested. The dated historical producer description remains historical.
+
+The mapper now preserves file/line subject and explicit nullable location metadata, rejects malformed location shapes so coverage becomes partial, and avoids inventing a current side. Activity labels file comments as file comments and unknown current line/side as unavailable, without asserting a comment is outdated. Current design prose names the connected `Writes` boundary. GitHub's [official review comment API](https://docs.github.com/en/rest/pulls/comments) documents file and line targets; later pages, complete thread/resolution, inline placement, multi-line ranges and review-comment Add to chat/copy-link remain separate product requirements.
+
+Before production changes, two new service tests failed on lost file/unknown provenance and incorrectly complete malformed coverage. The actual Electron case failed on the obsolete-line label. Those first failures and four Electron artifacts are retained. The initial post-correction service run had 56 passes and one test-helper failure: an unavailable-only helper was incorrectly used on the correctly partial response. The assertion now checks the narrowed partial result directly; no production or coverage expectation was weakened. Fresh corrected build and typecheck pass; all 57 read-service cases and all fifteen Electron regression cases pass without framework retries. Electron coverage includes valid file comments, replies, LEFT-side labeling, absent current positions, filtering and inert hostile Markdown. Focused artifacts are independently copied and byte-verified before full E2E may overwrite their source directory.
+
+The initial full run on pre-correction `38bab09` was intentionally interrupted during units using one SIGINT to its independently verified driver/descendant process group. All recorded processes were then absent, and E2E had not started. That run is incomplete and superseded, not a full-suite pass or a product failure. Final corrected-source full qualification and protected delivery remain pending; no parent issue, live/human gate or activation is accepted from these focused results.
+
+## Pre-push file-target schema correction
+
+A read-only public GitHub API check found a valid file subject carrying `line: 1` and `side: RIGHT`: [public metadata example](https://github.com/github/docs/pull/46247#discussion_r4236229612). Only identifier/target/location metadata was retained, never the comment body. The earlier review correction incorrectly required file targets to have null location fields and would drop that valid comment. This was an executor-introduced validation assumption, not evidence that GitHub returned malformed data. The synthetic regression preserves the shape without copying the real comment content.
+
+One new service regression failed because a valid file target made the aggregate partial, and the real Electron case failed because that file comment was absent. Both first failures and four Electron artifacts are retained. The mapper now uses the explicit file/line subject to determine target provenance, retains valid supplied numeric line/side metadata, permits omitted locations for file subjects, and still rejects malformed line/side shapes for both kinds. The existing file-target renderer needs no further change: it labels the comment as a file even when optional line metadata is present.
+
+Fresh build and typecheck pass; all 58 read-service cases and all fifteen Electron regression cases pass in 42.2 seconds without retries. Sixteen focused output files were independently byte-compared before another full run. The first corrected-source full run at `defa81e` had completed all six unit groups (2,586 passes, 36 skips, zero failures/cancellations), then was intentionally stopped during E2E to repair this newly proved defect: 482 desktop passes, seventeen skips, 47 not run and one interrupted case in 20.3 minutes. No site/web suite ran. All 169 partial E2E artifact files were copied and byte-verified. That incomplete run is superseded, not a full E2E pass or an ordinary product-test failure; its units remain evidence only for that earlier source. The next full qualification must bind to the corrected signed head before push or delivery.
+
+## Final local qualification
+
+Frozen signed runtime/test source `57d2773e755259a2e618e40f62e179bd8f5d6d8c`, tree `943510ba6f20cb19559a27dacd98745eb5b853d1`, passed fresh build/typecheck and all six full unit groups: 2,587 passes, 36 explicit skips, zero failures/cancellations. Full E2E passed 530 desktop cases with seventeen explicit opt-in skips in 27.9 minutes, eight site cases in 8.7 seconds and four web cases in 3.5 seconds, without framework retries or flaky annotations. All fifteen Code Review regressions passed, including the valid file-target shape, account/revocation cases and unsafe paths. Every one of the 852 tracked producer-file hashes remained unchanged. All 187 full-suite output files were copied and independently rehashed before running the native suite.
+
+The seventeen desktop skips are four live-provider cases, four local-model native cases, three Open Knowledge cases and six installed-CLI Code Review cases. They are not acceptance passes. The six Code Review cases were then separately qualified on the same unchanged inputs: four ordinary passes and two intentional expected-failure controls in 22.1 seconds, using installed Claude Code 2.1.323, Electron 44.4.5, Chromium 152.0.7977.130 and embedded Node 24.21.0 on macOS 27.2.1 arm64. The actual CLI consumed hostile PR content, allowed selected evidence and refused outside-secret/symlink reads, shell, write, fetch, MCP and the placeholder; forbidden effect markers remained absent. Three native starts were recorded; root independently observed their three PIDs absent and their three provider ports refusing connections. Fifty native output files and all three owned fixture trees were independently byte-compared. Incoming GH_TOKEN was already absent, so this is not a supplied-token native scrub challenge. The provider and GitHub payloads were scripted, with no paid inference or live product write; cleanup observations cover sampled recorded children, not every possible descendant.
+
+One fresh whole-branch review and every ruling are retained, with no Critical/Important finding and no deferred Minor. The first coalesced review correction and the subsequent public-schema correction remain separately identified above. No second feature review was requested; the later CI-only correction received its own scoped review. The complete signed recovery restores into an empty bare repository, passes full fsck and reproduces all thirteen authored files byte-for-byte; original #513 source history and authorship remain ancestors. The final qualification-record commit changes documentation only; every runtime/test/dependency input stays byte-exact to the qualified source. Source-hosted CI, actual protected delivery and merged-tree equivalence remain pending. This is a reviewable implementation candidate, not whole-parent, human/live-provider/live-write, release or default-activation acceptance.
diff --git a/docs/reviews/2026-11-13-cr8-adversarial-delivery.md b/docs/reviews/2026-11-13-cr8-adversarial-delivery.md
new file mode 100644
index 00000000..e0196d0f
--- /dev/null
+++ b/docs/reviews/2026-11-13-cr8-adversarial-delivery.md
@@ -0,0 +1,21 @@
+# CR-8 adversarial/race delivery audit
+
+The literal adversarial/race test item in #327 is now accepted at signed actual protected main `927b443414befe5f4300bb3f3d39af4666032609`, tree `32853cd5fd78046cc3e8e12aac38b7e910916b85`, through #518. Its sole parent is `b6e3026234cf39c51f5ad0dbb6183f3c5c5b0b83`; the complete tree exactly matches qualified source `81e54f816b49850f6acdeb1c3f668c7411637309`. GitHub records the merge at 2026-11-13 01:47:13 UTC. This supersedes pending-delivery sentences in the [source receipt](2026-11-12-cr8-access-revocation.md), preserving that receipt's first failures and correction history.
+
+[Actual protected run 43343769795](https://github.com/TypeSafeAI/modex/actions/runs/43343769795/job/131618073158) passed 2,579 unit cases with 36 skips, 532 desktop cases with fourteen skips in 24.9 minutes, eight site cases in 19.3 seconds and four web cases in 3.0 seconds, with zero failures/cancellations or framework retries. Both protected artifacts match their API SHA-256 digests and pass ZIP integrity. The queue request was sent once; no bypass was used. The protected iPhone job skipped and provides no device acceptance.
+
+| Literal requirement | Source and observed assertions |
+| --- | --- |
+| Wrong-account delayed replies and two repositories sharing a PR number | `e2e/code-review-regressions.spec.ts`: account-switch and held-answer cases reject old-account/private and cross-repository content. |
+| Revoked access and head changes | The same file verifies accessible-to-403/404/410 transitions, six sibling reads after offline failure, held private context after denial, unrelated repository fallback and head movement before/during file reads. |
+| Incomplete pagination/diffs and unsafe Markdown/URLs/paths | Partial inbox/files/patch states remain explicit; hostile content stays inert; unsafe links and eight context-path IPC inputs are refused after a valid positive control. |
+| Injected instructions and secret leakage | `e2e/code-review-native-boundary.spec.ts` consumes an injected PR snapshot through the actual installed CLI, allows selected evidence, refuses outside-secret/symlink reads, shell/write/fetch/MCP/placeholder proposals, and checks absent effect markers and outside-secret content. |
+| Stop/unknown completion and duplicate starts | `e2e/code-review-agent.spec.ts` verifies Stop, refused concurrent start, crash/relaunch unknown state, exact intentional starts and no automatic provider replay. |
+| Lost write replies and replay prevention | `e2e/code-review-actions.spec.ts` records one synthetic POST before crash, restart in doubt and explicit read-back with no replay; the separate 502 path retains one POST after further read-back and tracking decisions. |
+| Local Discard routing | The regression file checks absent remote revert controls, zero intercepted local-revert calls and unchanged local HEAD/status/diff/file bytes. |
+
+All non-opt-in case names above are present as passing cases in the actual protected log. The separate installed-CLI qualification used Claude Code 2.1.323, Electron 44.4.5, Chromium 152.0.7977.130 and embedded Node 24.21.0 on macOS 27.2.1 arm64. It passed four ordinary checks and two intentional expected-failure controls in 18.9 seconds, with three native starts, three sampled recorded PIDs absent and three provider ports closed. All 853 tracked producer-file hashes independently match actual merged main byte-for-byte; fifty native output files and three owned fixture trees were copied and byte-verified. These unchanged-source results remain reusable without pretending the skipped hosted native rows ran.
+
+Incoming GH_TOKEN was already absent during the native run; this does not prove a supplied-token native scrub challenge. PID/port observations are sampled, not exhaustive descendant proof. Provider and GitHub payloads are scripted; there is no paid inference or live product write. This accepts only the literal integrated adversarial/race test item, not whole #327/#319, visual/reference agreement, human accessibility, performance budgets, live provider/write, release or default activation. [Criterion-level publication and read-back](https://github.com/TypeSafeAI/modex/issues/327#issuecomment-6841738907) records the single checkbox change to seven of 17.
+
+The session-owned merged worktree was explicitly released, then removed only through the guarded helper after fresh merge proof and recovery verification. An independent empty-bare restore passed full fsck, verified the three own SSH signatures and reproduced all nine authored files byte-for-byte. All thirty scratch files retain both their original tree outside the worktree and a separately byte-verified copy. Other owners' worktrees and remote branches were not removed.
diff --git a/docs/status.md b/docs/status.md
index 4eca5403..37bc4e66 100644
--- a/docs/status.md
+++ b/docs/status.md
@@ -136,7 +136,7 @@ before relying on a tick.
| [#324](https://github.com/TypeSafeAI/modex/issues/324) CR-5 | 12/14 | Original R4 comparison (AC1), retired on 2026-11-04 but still named by the box; the Viewed toggle is decided (local, revision-bound) and not built | Build the Viewed toggle; run the benchmark |
| [#325](https://github.com/TypeSafeAI/modex/issues/325) CR-6 | 13/14 | Authorized live native smoke (AC6) | Run the smoke once on a small open TypeSafeAI/modex PR (authorized 2026-11-04) |
| [#326](https://github.com/TypeSafeAI/modex/issues/326) CR-7 | 13/14 | Live canary on a disposable repository (AC6) | Val names a disposable repository |
-| [#327](https://github.com/TypeSafeAI/modex/issues/327) CR-8 | 6/17 | The full fixture journey is accepted at its [named source/CI scope](https://github.com/TypeSafeAI/modex/issues/327#issuecomment-6830284937); reference comparison (retired 2026-11-04, boxes still name it), VoiceOver and the live gates. Budgets and C76 landed in #490; re-check item 18 and AC5 once Val confirms the budget numbers | Val's [one session](reviews/2026-11-02-v0.5.0-val-session.md) |
+| [#327](https://github.com/TypeSafeAI/modex/issues/327) CR-8 | 7/17 | The full fixture journey and adversarial/race item are accepted at their [named source/CI/native scopes](reviews/2026-11-13-cr8-adversarial-delivery.md); reference comparison (retired 2026-11-04, boxes still name it), VoiceOver and the live gates. Budgets and C76 landed in #490; re-check item 18 and AC5 once Val confirms the budget numbers | Val's [one session](reviews/2026-11-02-v0.5.0-val-session.md) |
| [#376](https://github.com/TypeSafeAI/modex/issues/376) LM tracker | 2/14 | LM-1/LM-2 accepted; LM-3 to LM-7 core acceptance; LM-8 is optional and does not block the core release; `LocalExecution` is not constructed, so the product refuses local turns | Val answers the five questions in `docs/reviews/2026-11-03-lm5-val-decision.md` on branch `lm-ready` |
| [#379](https://github.com/TypeSafeAI/modex/issues/379) LM-3 | 12/13 | AC1's real native read/edit/check and AC4's regression/configuration receipts now bind to actual main. AC6 still has native redirect and managed-policy gaps; simultaneous isolation claims retain their tested fixture scope | Resolve AC6 without weakening destination binding; product admission stays separate. [Receipt](https://github.com/TypeSafeAI/modex/issues/379#issuecomment-6811858808) |
| [#380](https://github.com/TypeSafeAI/modex/issues/380) LM-5 | 6/13 | S5 shared capacity/cancellable waiting/explicit status and AC3/AC5 are reconciled on protected main. Redirect privacy, action-safe compaction, genuine OOM and older-client presentation remain open | Complete remaining engineering/native gates; preserve #475's owner disposition. [Delivered S5 evidence](https://github.com/TypeSafeAI/modex/issues/380#issuecomment-6815410455) |
@@ -147,8 +147,10 @@ before relying on a tick.
## In flight
+- **Code Review line comments integration** (#513, #520). A new owned branch integrates ready source `c599dd7` from protected main `927b443`, retaining source authorship/history and every unrelated current-main status fact. Activity loads the first 100 REST review comments with distinct file/line targets, path/line/reply labels and explicit partial/capped states. Resolved-thread state, inline placement and further comment pages remain separate requirements. Signed runtime `57d2773` passes build/typecheck, 2,587 units with 36 skips, 530 desktop cases with seventeen opt-in skips, eight site cases and four web cases. Separate installed Claude Code 2.1.323 qualification passes four ordinary checks and two intentional fault controls with a scripted provider. The public-schema follow-up preserves valid location metadata on file targets. The first hosted iPhone job hit its 30-minute job maximum after 65 native-unit passes and 39 UI-case passes, including one expected-failure control; its full UI suite remains unverified. The original required macOS job passed 2,587 units, 533 desktop cases with fourteen opt-in skips, eight site cases and four web cases; the cancelled iPhone suite prevents a whole-source-green claim. The separately reviewed correction allows 45 minutes for that job while preserving all test deadlines and assertions. Corrected-head source-hosted qualification and protected delivery remain pending. [Integration receipt](reviews/2026-11-13-cr-inline-integration.md).
+
- **Local-model refresh preservation** (#381). Protected #503 delivered signed `558b54f8`, exact reviewed tree `3ba66a4`. Its queue passed 2,478 units with 36 skips, 522 desktop E2E cases with eight skips and seven site cases. Canonical cases preserve two thread drafts, staged image bytes, search/edit focus and selected text through delayed metadata, and keep an old revision-bound target after endpoint replacement. The full S2/AC3 UI/race audit and conservative S5 option/pinning guards are reconciled at five of 13 boxes; qualification/provenance/admission/human requirements remain open. [Delivery and limits](reviews/2026-11-12-local-model-refresh-preservation.md#actual-protected-delivery).
-- **CR-8 integrated native boundary fixture** (#327). An opt-in real Electron/main review uses installed Claude 2.1.322 with a scripted provider: hostile PR text is read, outside reads and shell/write/fetch/MCP proposals are denied, one explicit start completes and its native PID exits. The default model’s extra reserved placeholder is explicitly attempted and denied. The fixture also verifies retained provider observations and cleanup under intentional callback/close failures; unknown process-inspection errors fail qualification. This is named local evidence; the full adversarial/race box, live-provider smoke and human gates stay open. [Evidence and limits](reviews/2026-11-12-cr8-native-boundary.md).
+- **CR-8 integrated native boundary fixture** (#327). Protected #517 delivered the boundary harness; fresh integrated Claude Code 2.1.323 qualification on the exact #518 tree passed four ordinary checks and two intentional expected-failure controls. The actual Electron/main/installed-CLI run consumes hostile PR text, permits selected evidence and refuses outside-secret/symlink reads, shell, write, fetch, MCP and the reserved placeholder. All 853 tracked inputs match protected main `927b443` byte-for-byte. Incoming GH_TOKEN was already absent, so this is not a supplied-token native scrub challenge. Full adversarial/race integration is now accepted at this named fixture/native scope; live-provider and human gates remain open. [Delivery audit and limits](reviews/2026-11-13-cr8-adversarial-delivery.md).
- **CF-8 repeated source observations** (#281). Three fresh-app history and log replicates per source add 189 measured opens and 63 measured streams on each source. Public per-replicate numbers retain named host/runtime, timestamps and original receipt hashes. The comparison is descriptive, with shared current dependencies and sequential fixed-shape workloads; it does not qualify a historical shipped binary. The full S5 clause audit also checks delivered payload/log caps, cleanup and child regressions; S5 is supported. Protected #514 also delivered the portable reproduction and named capture/difference evidence at signed `30a56557`; the combined reproducibility criterion is accepted, at ten of 27 issue boxes. Native, human, physical-device and fresh integrated-candidate acceptance remain open. [Measurements and limits](reviews/2026-11-12-cf8-retention.md).
- **Local-service queue progress** (#380). A follow-up from merged #499 adds one waiting notice and a neutral continuation notice through the existing host-event path, plus an abort check at capacity handoff. Three canonical regressions failed before repair; 39 focused cases now pass, including actual runner/store persistence before dispatch and cancellation with no continuation or extra child. Full local build/typechecks, 2,478 units with 36 skips, 515 desktop E2E cases with 11 skips and seven site cases passed. Protected #500 delivered exact tree `b354cf7` at signed main `4f1d80c`; its queue passed 2,478 units with 36 skips, 518 desktop E2E cases with eight skips and seven site cases. S5 is reconciled at six of 13 issue boxes; product admission and remaining acceptance stay open. [Evidence and limits](reviews/2026-11-12-local-queue-status.md).
@@ -163,7 +165,7 @@ before relying on a tick.
- **Local draft destination fence** (#380, #381, #498). Final review found that an asynchronous local selection could accept input/send through its old non-local configuration before the target was pinned. Two controlled regressions failed; the repair reserves and fences only the source/new target, preserves the exact raw draft, and leaves unrelated Stop and editing available. All three focused Electron cases and full frozen build/typechecks/units pass (2,465 unit passes, 36 skips), followed by 511 desktop E2E passes with 11 skips and seven site passes. Protected [#498](https://github.com/TypeSafeAI/modex/pull/498) has since delivered this repair; its exact merge-group macOS check passed. The counts above retain their original candidate scope, and product local execution remains unadmitted. [Failure, repair and limits](reviews/2026-11-11-local-selection-fence.md).
- **iPhone image-binding diagnostics** (#109, #498). The prior manifest collapsed container-query failures, realpath failures and image mismatches into app-image refusals. A diagnostic-only increment retains at most 16 callback/time-bound outcomes with sanitized error metadata and unchanged identity/command/stop guards. Six baseline controls fail; all 25 observer cases pass after the change. Full frozen build/typechecks/units pass with 2,471 passes, 36 skips and zero failures/cancellations. Fresh hosted observation remains required; no successful stack binding or older TLS cause is claimed. [Evidence and limits](reviews/2026-11-11-ios-image-binding-diagnostics.md).
- **iPhone trust observer collection** (#498). Published `188af74` passed 44 iPhone tests with zero actual failures, but its observer stopped after six poll errors and unconfirmed command closure, with no callbacks/stacks captured. A predicate-only experiment narrows app/Security and trustd-peer logs while keeping every bound and identity guard. Native Foundation predicate replay goes from output-limit failure to 21 selected records with preserved app/trustd correlation; all 19 focused tests pass. Full build/typechecks and units passed with 2,465 passes, 36 skips and zero failures/cancellations. Fresh `aa095b2` app qualification passes 44 tests, but nine image-binding refusal entries yielded zero bound observers/stacks; five distinct callbacks in offline diagnostics completed normally. Live stack binding remains unqualified and the older trust-stall cause remains unknown. [Evidence and limits](reviews/2026-11-11-ios-trust-predicate.md).
-- **Code Review revocation integration evidence** (#327). An owned candidate adds actual Electron/IPC accessible-to-revoked cases for 403/404/410, a held context response and an unsafe context-path boundary. Four focused revocation cases pass; a deliberate temporary 403-as-offline mutation fails with the private body still visible, and the exact restored production source passes all thirteen then-current regressions. The path case also passes, and fresh build/typechecks plus 2,570 units with 36 explicit skips pass. Full E2E passes 525 desktop cases with eleven explicit opt-in skips, eight site cases and four web cases, including all fourteen regressions. One fresh review is complete; an immediate denial-diagnostic assertion gap is corrected and all fourteen regressions pass again. Post-correction full E2E also passes 525 desktop cases with eleven skips, eight site cases and four web cases; the first source-hosted macOS run fails an existing human-question privacy scan on an atomic temporary-file rename, while all fourteen regressions pass. The follow-up closes the private app before the unchanged scan, follows the keyboard target’s associated details card and repairs a reproduced title-bar bug where an old Rename command opened editing on later chats. Independent follow-up review found no blocking defects, and both minor test improvements are implemented. The current-main build, typecheck, 2,579 units with 36 skips, all 22 affected cases and full E2E pass: 529 desktop cases with 17 skips, eight site cases and four web cases. Separate Claude Code 2.1.323 native qualification passes four ordinary checks and two intentional expected-failure controls, with three native starts and no unexpected failure. Fresh source-hosted checks, protected delivery and the final full-clause audit remain pending. No full adversarial, live-provider, real-write or human acceptance is claimed. [Evidence and limits](reviews/2026-11-12-cr8-access-revocation.md).
+- **Code Review revocation integration evidence** (#327). Protected #518 delivered signed main `927b443414befe5f4300bb3f3d39af4666032609`, exact source tree `32853cd5fd78046cc3e8e12aac38b7e910916b85`. Its queue passed 2,579 units with 36 skips, 532 desktop cases with fourteen skips, eight site cases and four web cases, without failures or retries. A literal clause audit reconciles account/repository/head races, 403/404/410 revocation, incomplete data, hostile content/paths, native injection resistance, Stop/unknown completion, write replay prevention and local Discard isolation. The full adversarial/race item is now accepted as seven of 17 #327 boxes; broader producer/live/human requirements remain open. The slice also repairs replayed menu-rename intent; first failures and review corrections remain in the [source receipt](reviews/2026-11-12-cr8-access-revocation.md), with final delivery in the [audit](reviews/2026-11-13-cr8-adversarial-delivery.md).
- **Code Review native-modal investigation** (#498). Parent `fc593e0` timed out clicking Settings before Jev contrast collection; teardown then timed out and the job reached its 45-minute limit. The retained Electron stack is inside `NSAlert runModal`, with dialog message and cause unknown. A test-only candidate records bounded owned-child stderr and observes uncaught exceptions without handling them; five real-child controls and all eight accessibility cases pass locally. Full build/typechecks and units passed with 2,460 passes, 36 skips and zero failures/cancellations. Exact-head hosted qualification remains required; no causal repair or acceptance is claimed. [Observation and limits](reviews/2026-11-11-electron-modal-observation.md).
- **Linux Debian source continuity** (#498). Candidate `188af74` acquired all four signed Noble pockets and installed host dependencies, then failed baseline Debian installation because `libnotify4` was not installable. The next candidate keeps explicit signed sources through both baseline and upgrade dependency resolution, preserving artifact verification, installed checks and the three-minute direct command bound. All 13 synthetic helper/wrapper cases and final corrected-source build/typechecks/units pass, with 2,465 passes, 36 skips and zero failures/cancellations. Fresh `aa095b2` hosted installed checks now pass eight Linux launches plus controlled KWallet5 and three Windows NSIS/DPAPI launches; original installer acceptance remains pending, including historical-schema, full graphical-login and real-account limits. [Failure, candidate and limits](reviews/2026-11-11-linux-debian-sources.md).
- **Linux qualification host preparation** (#498). Candidate `5edbb84` passed required hosted macOS, iPhone and Windows checks, but both Linux attempts stopped at the 30-minute job limit during Azure Ubuntu mirror DNS failures before installed-app acceptance. The CI-only repair selects official signed Noble archive/security sources per APT invocation, bounds acquisition and the host step, and preserves the nine packages and downstream installed acceptance. Build passed; final full units passed with 2,427 passes, 35 skips and zero failures. Seven focused helper cases and independent review cover argument boundaries, failure propagation and owned temporary cleanup. Application/native inputs were unchanged by that CI-only repair, which has since landed through protected [#498](https://github.com/TypeSafeAI/modex/pull/498). Its later delivery is distinct from the historical host failures above; native account/graphical-session acceptance and product admission remain open. [Failure evidence, repair and limits](reviews/2026-11-11-linux-host-preparation.md).