diff --git a/.lane/reports/fix-agent-desk-canonical-apex-lane1.md b/.lane/reports/fix-agent-desk-canonical-apex-lane1.md
new file mode 100644
index 00000000..8f64cee4
--- /dev/null
+++ b/.lane/reports/fix-agent-desk-canonical-apex-lane1.md
@@ -0,0 +1,67 @@
+# Lane 1 — retired "TinyStudio Agent Desk" title/snippet (f41c8af0f8) — 2026-08-15
+
+## Verdict
+
+The one remaining repo-side mechanism still feeding the retired name to
+Google is closed: the legacy `/agent-desk` surface no longer declares the
+apex root as its canonical/og:url, and a regression guard now fails the
+build if that claim returns. The canonical fix has been blocked in open PR
+#91 since 2026-08-10; this branch lands the corrected version.
+
+## Why the item was still live
+
+- The 2026-08-09 de-indexing pass (PR #46) added `noindex, nofollow` and
+  retired framing to `public/agent-desk.html` — that part has held.
+- The 2026-08-14 pass (the www-host 301) collapsed the duplicate
+  `www.tinystudio.io` site entity that carried the stale "TinyStudio Agent
+  Desk" site name — that part has held too.
+- But `public/agent-desk.html` still declared
+  `<link rel="canonical" href="https://tinystudio.io/" />` and
+  `og:url` = `https://tinystudio.io/` — the exact consolidation mechanism by
+  which the retired title reached the homepage URL (q5/google capture
+  2026-08-06). Verified live on 2026-08-15: `/agent-desk` serves that
+  apex-root claim, and the page is noindex, so its canonical tells Google the
+  legacy surface is a duplicate of the homepage.
+- The canonical fix has been open in PR #91 since 2026-08-10 and is still
+  unmerged on main; main and live therefore still served the apex-root claim.
+
+## The change
+
+- `public/agent-desk.html`: canonical and og:url now name the legacy page's
+  own address `https://tinystudio.io/agent-desk` — the clean form that serves
+  200 — never the apex root. (PR #91 pointed them at `/agent-desk.html`, but
+  that form 307-redirects to `/agent-desk`; the repo's own canonical
+  convention, which the audit page already follows, requires naming the
+  200-serving address.)
+- `scripts/check-site.mjs`: the retired-desk index guard now also requires
+  exactly one canonical and one og:url, both naming
+  `https://tinystudio.io/agent-desk`. Regression proof: re-pointing them at
+  the apex root makes `npm run check` fail with both messages.
+
+## Verification
+
+- `node scripts/check-site.mjs` → "TinyStudio.io checks passed."
+- `npm test` → 121 tests, 0 failures (6 headings, 7 sitemap, 80 agent-worker,
+  16 agent-UI, 8 product-contract, 4 first-viewport-audience), plus the
+  narrow-viewport suite PASS.
+- `git diff --check` clean.
+- Guard negative test: apex-root canonical → `npm run check` exit 1 with
+  both canonical and og:url failures; restored fix → green.
+- Live baseline (2026-08-15, before deploy): `/agent-desk` 200 carrying
+  `canonical https://tinystudio.io/` + `og:url https://tinystudio.io/`;
+  `/agent-desk.html` 307 → `/agent-desk`.
+
+## Files touched
+
+- `public/agent-desk.html` — canonical/og:url point at the legacy page's own
+  200-serving address.
+- `scripts/check-site.mjs` — guard requires canonical/og:url =
+  `https://tinystudio.io/agent-desk`.
+- `.lane/reports/fix-agent-desk-canonical-apex-lane1.md` — this report.
+
+## PR relationship
+
+PR #91 (`fix/agent-desk-canonical-lane1`) carries the same fix intent but
+points the canonical at the 307-redirecting `/agent-desk.html` form and its
+guard covers only the canonical, not og:url. This branch is the corrected
+delivery; #91 is closed as superseded so exactly one delivery path remains.
diff --git a/public/agent-desk.html b/public/agent-desk.html
index 4f68447e..18de745f 100644
--- a/public/agent-desk.html
+++ b/public/agent-desk.html
@@ -15,8 +15,8 @@
       content="The self-serve Agent Desk is retired and is not the current offer. TinyStudio's current offer is the free leak audit of high-ticket service homepages."
     />
     <meta property="og:type" content="website" />
-    <meta property="og:url" content="https://tinystudio.io/" />
-    <link rel="canonical" href="https://tinystudio.io/" />
+    <meta property="og:url" content="https://tinystudio.io/agent-desk" />
+    <link rel="canonical" href="https://tinystudio.io/agent-desk" />
     <meta property="og:image" content="https://tinystudio.io/og-image.png" />
     <meta property="og:image:width" content="1200" />
     <meta property="og:image:height" content="630" />
diff --git a/scripts/check-site.mjs b/scripts/check-site.mjs
index dd69347e..df5bfcc2 100644
--- a/scripts/check-site.mjs
+++ b/scripts/check-site.mjs
@@ -1266,7 +1266,12 @@ for (const [pageName, pageHtml] of ownedPages) {
 // The legacy page must therefore stay out of the index: its head keeps a
 // robots noindex, nofollow meta, and its title and description frame the
 // surface as retired, so neither the search index nor a scraper can re-present
-// the retired self-serve product as the current offer.
+// the retired self-serve product as the current offer. Its canonical and
+// og:url must name the legacy page itself — the clean /agent-desk address
+// that serves 200 — never the apex root: while the page declared the root as
+// its canonical, Google consolidated the retired title onto the homepage URL
+// (the q5/google capture), and a canonical that points at the root keeps
+// handing the retired name to the homepage's SERP entry.
 const retiredDeskHead = retiredDesk.match(/<head\b[^>]*>([\s\S]*?)<\/head>/i)?.[1] ?? "";
 const robotsMeta = /<meta\b(?=[^>]*\bname="robots")(?=[^>]*\bcontent="noindex,\s*nofollow")[^>]*>/i;
 if (!robotsMeta.test(retiredDeskHead)) {
@@ -1280,6 +1285,29 @@ const retiredDeskDescription =
 if (!/\bretired\b/i.test(retiredDeskDescription)) {
   failures.push("Retired Agent Desk description must frame the surface as retired.");
 }
+// The retired page must never claim the apex root. Exactly one canonical and
+// one og:url, both naming the clean /agent-desk address that serves 200.
+const retiredDeskLive = retiredDesk.replace(/<!--[\s\S]*?-->/g, "");
+const retiredDeskCanonical =
+  retiredDeskLive.match(/<link\b[^>]*\brel\s*=\s*["']canonical["'][^>]*>/gi) ?? [];
+if (retiredDeskCanonical.length !== 1) {
+  failures.push(`Retired Agent Desk canonical must appear exactly once (found ${retiredDeskCanonical.length}).`);
+} else {
+  const href = retiredDeskCanonical[0].match(/\bhref\s*=\s*["']([^"']*)["']/i)?.[1] ?? "";
+  if (href.trim() !== "https://tinystudio.io/agent-desk") {
+    failures.push(`Retired Agent Desk canonical must point at https://tinystudio.io/agent-desk (found "${href}").`);
+  }
+}
+const retiredDeskOgUrl =
+  retiredDeskLive.match(/<meta\b[^>]*\bproperty\s*=\s*["']og:url["'][^>]*>/gi) ?? [];
+if (retiredDeskOgUrl.length !== 1) {
+  failures.push(`Retired Agent Desk og:url must appear exactly once (found ${retiredDeskOgUrl.length}).`);
+} else {
+  const content = retiredDeskOgUrl[0].match(/\bcontent\s*=\s*["']([^"']*)["']/i)?.[1] ?? "";
+  if (content.trim() !== "https://tinystudio.io/agent-desk") {
+    failures.push(`Retired Agent Desk og:url must point at https://tinystudio.io/agent-desk (found "${content}").`);
+  }
+}
 
 // ---- Meta descriptions (dogfood) -------------------------------------------
 // The leak audit this site sells flags a homepage whose served HTML carries no