diff --git a/REPORT-m3-2026-10-11.md b/REPORT-m3-2026-10-11.md
index b5790c87..3907a597 100644
--- a/REPORT-m3-2026-10-11.md
+++ b/REPORT-m3-2026-10-11.md
@@ -75,7 +75,8 @@ deliberately not built** (your brief) — the shell should settle first.
    paragraph. Found by a word-level diff.
 4. **The sidebar rendered as a horizontal row** across the content column, because „later wins" is
    per property and the controller's `.nav-links` never declares `display`. Diffing the two files'
-   declaration sets found six such leaks; only this one was visible.
+   declaration sets found six such leaks; only this one was visible. **That diff was wrong, and the
+   section below says how** — the real number was forty.
 5. **The contact pages' header slid under the sidebar** (content column at x=642 instead of 240,
    form at two-thirds width) because those pages keep their header outside `<main>`.
 
@@ -87,6 +88,32 @@ deliberately not built** (your brief) — the shell should settle first.
   *controller's* `.btn-primary`, which the shell imports; it measured 4.01:1 again. Your existing
   remedy applied. `.nav-group-label` was 3.14:1 and moved to `--text-2`.
 
+## Two defects a real phone found, after the switch
+
+Reported by you, from your own phone, hours after it went live. Both were mine. Fixed, live as
+`?v=23`, and read back on all 20 pages at 390 / 860 / 1440 px.
+
+1. **The phone menu was empty.** The drawer slid open correctly and had nothing in it. `site.css`'s
+   old mobile nav still reached the shell's sidebar: under `@media (max-width: 900px)` it sets
+   `.nav-links { position: fixed; right: -100% }`, which parked the menu off-canvas *inside* the open
+   drawer. Every old-chrome selector (29 occurrences) is now scoped `body:not(.shell)` — scoped, not
+   deleted, so a page without the shell still works.
+2. **The footer rendered as a second column** at the top right of 16 pages. `body` is a flex row on a
+   shell page, and my insert moved the footer inside `<main>` only on pages that *ended* with
+   `</main>`; a page with a `<script>` after it kept the footer as a direct child of `body`.
+
+**The first one is a failure of my own check, not bad luck.** The leak diff I ran at STOP 1 — the one
+whose „exactly six" this report quoted above — **excluded `@media` blocks**. Media-aware, the number
+is **forty**, and the worst of them only apply below 900 px, which is exactly why the desktop screens
+I took looked right. A check that silently skips part of its subject is not a check: it returned a
+confident, wrong number, and I believed it rather than the phone.
+
+**Both are now gated, each red-proofed by putting the live defect back:** gate 3b requires every shell
+page's footer to sit inside `<main class="content">`, and the new **gate 22** compares the two
+stylesheets **media-aware** and fails on any declaration of `site.css`'s that still reaches a shell
+page. Today it reads 689 site rules against 119 shell rules, 9 shared selectors, **0 leaks**. The two
+decoys (`site/shell-css-leak-in-media`, `site/footer-outside-main`) are the exact shapes that shipped.
+
 ## Still open
 
 - **R-931** — felhom.eu still serves **nothing compressed**. `shell.css` adds 15 KB per page, all of
@@ -98,5 +125,7 @@ deliberately not built** (your brief) — the shell should settle first.
 
 ## Gates and register
 
-`site_gates.py` green throughout; full `repo_gates.py --fast` green on DooPlex. Register unchanged at
-128 — nothing filed, because everything found was either fixed in the session or already has a row.
+`site_gates.py` green throughout, now with gate 22; full `repo_gates.py --fast` green on DooPlex,
+decoy suite included. Register unchanged at 128 — nothing filed, because everything found was either
+fixed in the session or already has a row, and the two live defects were found, fixed and pinned the
+same day.
diff --git a/STATUS.md b/STATUS.md
index d8de90f1..6fdec422 100644
--- a/STATUS.md
+++ b/STATUS.md
@@ -23,6 +23,12 @@
   contact form sat under the menu.
 - **Two phone bugs that were already live are fixed**: the download page and the privacy page used to
   scroll sideways on a phone.
+- **Two things you found on your own phone after it went live, both mine, both now fixed** (reload
+  once if the old version is still cached): the slide-out menu opened empty, and on 16 pages the
+  footer sat as a narrow second column at the top right instead of at the bottom. The menu one is the
+  worse of the two — the check I ran before switching was not looking at the phone-only part of the
+  stylesheet, so it told me six things were wrong when forty were, and the six it did find were all
+  visible on a big screen. Both now have a check that was watched failing before it was trusted.
 - **Waiting on you — nothing breaks if you do nothing:** the website still sends **everything
   uncompressed** (R-931). That is one setting on the server, it helps every page, and it is now the
   cheapest improvement left by a wide margin. Also: Search Console — the addresses did not change, so
diff --git a/scripts/CHANGELOG.md b/scripts/CHANGELOG.md
index cad651ae..29c0e165 100644
--- a/scripts/CHANGELOG.md
+++ b/scripts/CHANGELOG.md
@@ -1,3 +1,23 @@
+## 2026-10-11 (evening) — site gate 22: no CSS leak from site.css onto a shell page, MEDIA-AWARE
+
+- `site_gates.py` **gate 22** parses both stylesheets into (media context, selector, declarations) and
+  fails when `site.css` declares a property on a selector `shell.css` also styles that nothing
+  overrides — neither `shell.css` itself nor a `.shell …`-scoped reset. `site.css` loads first, so
+  „later wins" is per **property**, not per rule. Today: 689 site rules vs 119 shell rules, 9 shared
+  selectors, 0 leaking declarations. `.hamburger-box` / `.hamburger-line` are an explicit allow-list
+  with the reason beside them (the shell's burger is the website's own button markup).
+- **Why it exists.** The morning's STOP-1 leak diff excluded `@media` blocks and reported six leaks
+  where the media-aware number was forty. The worst of them, `@media (max-width: 900px) .nav-links {
+  position: fixed; right: -100% }`, parked the drawer's menu off-canvas and shipped: the operator
+  found an empty phone menu on the live site. The gate parses the media context precisely because the
+  check it replaces did not.
+- **Two decoys** in `test_gate_decoys.py`, both the exact shape that went live, both seen to convict:
+  `site/shell-css-leak-in-media` (the `body:not(.shell)` scope taken off that one rule → the gate
+  names the media query, the selector and `right: -280px`) and `site/footer-outside-main` (`<main>`
+  closed early on `gyik.html` → gate 3b's own line). They plant and restore **bytes**, with the anchor
+  translated to the file's own line ending: these files are CRLF in a Windows worktree and LF in a
+  Linux one, and `replace_in`'s text-mode round trip would have rewritten a whole file on one of them.
+
 ## 2026-10-10 (afternoon) — site gate 21: every catalog app has its logo under its slug (R-912)
 
 - `site_gates.py` gate 21 reads `app-catalog-felhom.eu/templates/*/.felhom.yml` (the `slug:`, else the directory) and
diff --git a/scripts/site_gates.py b/scripts/site_gates.py
index 38375908..8ad5f51a 100644
--- a/scripts/site_gates.py
+++ b/scripts/site_gates.py
@@ -31,6 +31,12 @@ Gates (all must pass; non-zero exit on any failure):
  19. dash-lang  — a page shows dashboard pictures only in its own language (`dashboard-*-hu.webp` / `-en.webp`)
  20. viewer     — data-gallery openers are <a href> to existing assets; gallery.js (with ?v=) exactly where openers are;
                   no inline viewer copy
+ 22. shell leak — MEDIA-AWARE: no declaration in site.css reaches a shell page on a selector that
+                  shell.css also styles. site.css loads FIRST, so a property shell.css never sets
+                  survives — „later wins” is per PROPERTY, not per rule. Earned live 2026-10-11:
+                  @media (max-width: 900px) .nav-links { position: fixed; right: -100% } parked the
+                  drawer's items off-canvas and the phone menu was empty. The check that missed it
+                  EXCLUDED @media blocks and reported six leaks where the media-aware number was 40
  21. app logos  — every catalog app (`app-catalog-felhom.eu/templates/*/.felhom.yml`, by its `slug:`) has
                   website/assets/<slug>-logo.svg or .png — the dashboard asks for exactly that name and shows a grey
                   placeholder otherwise (R-912: five apps had theirs under other names). The sibling catalog clone
@@ -631,8 +637,105 @@ else:
     else:
         print("  app logos: %d catalog apps, each with its <slug>-logo file" % len(_slugs))
 
+# gate 22 (2026-10-11): no CSS leak from site.css onto a shell page — MEDIA-AWARE.
+#
+# site.css loads BEFORE shell.css, so for a selector both files style, shell.css wins only on the
+# properties it actually declares; everything else survives from site.css. That is how the phone menu
+# shipped empty: site.css's old mobile nav still said, under @media (max-width: 900px),
+# `.nav-links { position: fixed; right: -100% }`, which parked the drawer's items off-canvas. It was
+# invisible at desktop width AND invisible to the check that was supposed to find it — that diff
+# skipped @media blocks and reported six leaking declarations where the media-aware number was forty.
+#
+# A leak is: site.css declares property P on selector S, shell.css also styles S, and nothing
+# overrides P for S — neither shell.css itself nor a `.shell …`-scoped reset in site.css. Scoping a
+# rule `body:not(.shell)` (the fix) takes its selector out of the comparison, which is the point.
+def css_rules(css):
+    """-> [(media_context, selector, [(prop, value), ...])].
+
+    @media / @supports / @layer nest and their context is CARRIED, so a declaration that applies only
+    on a phone is still seen. @keyframes and @font-face hold no page selectors and are skipped.
+    """
+    css = re.sub(r"/\*.*?\*/", "", css, flags=re.S)
+    out = []
+
+    def walk(text, media):
+        i, n = 0, len(text)
+        while i < n:
+            j = text.find("{", i)
+            if j < 0:
+                break
+            prelude = text[i:j].strip()
+            depth, k = 1, j + 1
+            while k < n and depth:
+                if text[k] == "{":
+                    depth += 1
+                elif text[k] == "}":
+                    depth -= 1
+                k += 1
+            body = text[j + 1:k - 1]
+            if prelude.startswith("@"):
+                if prelude.split()[0].lower() in ("@media", "@supports", "@layer"):
+                    walk(body, (media + " AND " + prelude) if media else prelude)
+            else:
+                decls = []
+                for d in body.split(";"):
+                    if ":" in d and "{" not in d:
+                        prop, val = d.split(":", 1)
+                        if prop.strip():
+                            decls.append((prop.strip().lower(), val.strip()))
+                for sel in prelude.split(","):
+                    sel = re.sub(r"\s+", " ", sel).strip()
+                    if sel:
+                        out.append((media, sel, decls))
+            i = k
+
+    walk(css, "")
+    return out
+
+
+# Shared ON PURPOSE: the shell's hamburger is the website's own button markup, so site.css styles the
+# same two class names the controller does. Another name here needs the same kind of sentence.
+_LEAK_ALLOW = {".hamburger-box", ".hamburger-line"}
+_site_css_p = os.path.join(W, "assets", "site.css")
+_shell_css_p = os.path.join(W, "assets", "shell.css")
+if not (os.path.isfile(_site_css_p) and os.path.isfile(_shell_css_p)):
+    fail("shell leak: site.css or shell.css is missing — the gate checked nothing")
+else:
+    _site_r = css_rules(io.open(_site_css_p, encoding="utf-8-sig").read())
+    _shell_r = css_rules(io.open(_shell_css_p, encoding="utf-8-sig").read())
+    if len(_site_r) < 200 or len(_shell_r) < 50:
+        fail("shell leak: parsed %d site rules and %d shell rules — too few to be the real files, so "
+             "nothing was compared" % (len(_site_r), len(_shell_r)))
+    _shell_sel = set(sel for _, sel, _ in _shell_r)
+    _overridden = set((sel, prop) for _, sel, decls in _shell_r for prop, _ in decls)
+    for _, _sel, _decls in _site_r:          # a `.shell X { … }` reset in site.css overrides X too
+        if ".shell" in _sel:
+            for _prop, _ in _decls:
+                _overridden.add((_sel.split(" ")[-1], _prop))
+    _leaks = []
+    for _media, _sel, _decls in _site_r:
+        if _sel in _LEAK_ALLOW or ".shell" in _sel or "body:not(.shell)" in _sel:
+            continue
+        if _sel not in _shell_sel:
+            continue
+        for _prop, _val in _decls:
+            if (_sel, _prop) not in _overridden:
+                _leaks.append((_media, _sel, _prop, _val))
+    if _leaks:
+        for _media, _sel, _prop, _val in _leaks[:12]:
+            fail("shell leak: site.css %s`%s { %s: %s }` reaches a shell page — shell.css styles %s but "
+                 "never sets %s, so this one survives. Scope the rule body:not(.shell), or reset the "
+                 "property under .shell" % ((_media + " ") if _media else "", _sel, _prop, _val[:40],
+                                            _sel, _prop))
+        if len(_leaks) > 12:
+            fail("shell leak: and %d more leaking declaration(s), not listed" % (len(_leaks) - 12))
+    else:
+        print("  shell leak: %d site rules vs %d shell rules, %d shared selector(s), 0 leaking "
+              "declarations (media-aware)" % (len(_site_r), len(_shell_r),
+                                              len(set(s for _, s, _ in _site_r) & _shell_sel)))
+
 if fails:
     print("\nSITE GATES FAILED: %d problem(s)" % len(fails))
     sys.exit(1)
 print("site gates OK — BOM, emoji=0, nav/footer per language, analytics, no CDN, no legacy tokens, no <style>, "
-      "cache-busted assets, twins + hreflang + globe, lang, same apps, no Hungarian in English, FAQ JSON-LD, clean tail, dashboard pictures per language, one picture viewer")
+      "cache-busted assets, twins + hreflang + globe, lang, same apps, no Hungarian in English, FAQ JSON-LD, clean tail, dashboard pictures per language, one picture viewer, no shell CSS leak")
diff --git a/scripts/test_gate_decoys.py b/scripts/test_gate_decoys.py
index e0437f7c..e732048a 100644
--- a/scripts/test_gate_decoys.py
+++ b/scripts/test_gate_decoys.py
@@ -91,7 +91,7 @@ COVERS = {
                         "switch at the wrong twin, lang=\"hu\" on an English page, an app missing from one apps page, "
                         "ASCII-only Hungarian in English text and in a script message, a 'please', an English "
                         "retrieval promise the Hungarian does not make, a FAQ JSON-LD drifting from the visible "
-                        "answer, text after </html>, a Hungarian dashboard picture on an English page, and four picture-viewer faults (an opener without href, an opener to a missing file, openers with no gallery.js, the inline viewer put back), and (R-912) a catalog app's logo under another name than its slug — each must convict WITH its own failure line"),
+                        "answer, text after </html>, a Hungarian dashboard picture on an English page, and four picture-viewer faults (an opener without href, an opener to a missing file, openers with no gallery.js, the inline viewer put back), and (R-912) a catalog app's logo under another name than its slug — each must convict WITH its own failure line; and (2026-10-11, both of them reported from a real phone AFTER they went live) site.css's old mobile nav reaching the shell's sidebar from inside an @media block — the leak the STOP-1 check missed because it skipped @media entirely — and a <footer> left outside <main class=\"content\">, which on a shell page renders as a second column"),
     "script-tests":    ("R-885: FIVE cases in scripts/test_script_tests_gate.py, run from here: a suite that PRINTS "
                         "OK and exits 1 (label without fact), a failing suite three levels deep (scope is a walk), "
                         "an empty scripts/ tree (checked nothing), the genuine article (must pass), the nested mark"),
@@ -614,6 +614,55 @@ decoy("site/text-after-html", "site_gates.py",
       replace_in(_EN("apps.html"), "</body>\n</html>", "</body>\n</html>\n=====\nSummary: 13 files included\n"),
       must="en/apps.html: text after </html>")
 
+# -- site: the two defects that reached a real phone (2026-10-11) --------------------------------
+#
+# Both shipped in the M3 switch and were reported by the operator from his own phone; neither was
+# found here. `replace_in` cannot build them: it round-trips through text mode, and these files are
+# CRLF in a Windows worktree and LF in a Linux one, so a text-mode restore would rewrite the whole
+# file on one of the two. These plant and restore BYTES, with the anchors translated to the file's
+# own line ending.
+def replace_bytes(path, pairs):
+    """Mutate one or more exact spans of a real file IN BYTES; restore it byte-for-byte.
+
+    Anchors are written with \\n and translated at plant time. A span that is no longer there makes
+    the decoy UNBUILDABLE, which is a failure of this suite — never a quietly skipped case."""
+    def _plant():
+        backup = io.open(path, "rb").read()
+        nl = b"\r\n" if b"\r\n" in backup else b"\n"
+        cur = backup
+        for old, new in pairs:
+            o = old.encode("utf-8").replace(b"\n", nl)
+            n = new.encode("utf-8").replace(b"\n", nl)
+            assert cur.count(o) == 1, ("%s: the decoy's anchor %r is gone or ambiguous — rebuild the "
+                                       "decoy" % (path, old[:60]))
+            cur = cur.replace(o, n, 1)
+        io.open(path, "wb").write(cur)
+        return lambda: io.open(path, "wb").write(backup)
+    return _plant
+
+
+# 1. THE PHONE MENU WAS EMPTY. site.css's old mobile nav still reached the shell's sidebar: under
+#    @media (max-width: 900px) it parked `.nav-links` off-canvas (position: fixed; right: -280px) —
+#    inside the open drawer. Desktop looked right, so only a phone showed it, and the STOP-1 check
+#    that should have caught it diffed the two stylesheets with @media blocks EXCLUDED: it reported
+#    six leaking declarations where the media-aware number was forty. The decoy takes the
+#    `body:not(.shell)` scope off that one rule, which is exactly the state that went live.
+decoy("site/shell-css-leak-in-media", "site_gates.py",
+      replace_bytes(os.path.join(_WEB, "assets", "site.css"),
+                    [("    body:not(.shell) .nav-links {\n        display: flex;\n        flex-direction: column;",
+                      "    .nav-links {\n        display: flex;\n        flex-direction: column;")]),
+      must="shell leak: site.css @media (max-width: 900px) `.nav-links { right: -280px }`")
+
+# 2. THE FOOTER RENDERED AS A SECOND COLUMN at the top right of 16 pages: body is a flex row on a
+#    shell page, so a <footer> left outside <main class="content"> is a flex SIBLING of the sidebar
+#    and the content, not the bottom of the page. The decoy closes <main> early and leaves the footer
+#    outside it — the shape the insert produced on every page that had a <script> after </main>.
+decoy("site/footer-outside-main", "site_gates.py",
+      replace_bytes(os.path.join(_WEB, "gyik.html"),
+                    [("    <footer>\n", "    </main>\n    <footer>\n"),
+                     ("</footer>\n    </main>\n", "</footer>\n")]),
+      must='gyik.html: <footer> is outside <main class="content">')
+
 # ── hostinstall (R-426, 2026-10-06) ──────────────────────────────────────────────────────────────
 #
 # Each decoy is a one-line edit a real session would make to felhom-host-install.sh, leaving the LABEL
diff --git a/website/CHANGELOG.md b/website/CHANGELOG.md
index 21d1f391..2dab4dd1 100644
--- a/website/CHANGELOG.md
+++ b/website/CHANGELOG.md
@@ -1,4 +1,30 @@
-## 2026-10-11 (M3) — the whole site moves into the controller's shell
+## 2026-10-11 (evening) — the two M3 defects a real phone found, and the gate that now holds them
+
+Both were mine, both shipped in the morning's switch, and both were reported by the operator from his
+own phone rather than caught here. `site.css?v=23` / `shell.css?v=23`, live and read back on all 20
+pages at 390 / 860 / 1440 px.
+
+- **The phone menu was empty.** The drawer slid open correctly and had nothing in it: `site.css`'s old
+  mobile nav still reached the shell's sidebar, because under `@media (max-width: 900px)` it sets
+  `.nav-links { position: fixed; right: -100% }` — which parked the menu off-canvas *inside* the open
+  drawer. Fixed at the source: every selector belonging to the website's OLD chrome (`nav`,
+  `.nav-links*`, `.mobile-overlay`, `.hamburger`, `.logo*`, `.lang-globe-li` — 29 occurrences) is now
+  scoped `body:not(.shell)`. Scoped, not deleted, so a page without the shell still works.
+  `.hamburger-box` and `.hamburger-line` stay unscoped on purpose: the shell's burger uses them.
+- **Why the morning's check missed it.** The STOP-1 leak diff compared the two stylesheets'
+  declaration sets **with `@media` blocks excluded**, and reported „exactly six" leaking declarations.
+  Media-aware, the number is **forty** — and the worst of them only apply below 900 px, which is why
+  the desktop looked right and a phone did not. A check that silently skips part of its subject is not
+  a check; this one returned a confident wrong number and was believed.
+- **The footer rendered as a second column** at the top right of 16 pages. `body` is a flex row on a
+  shell page, and the insert moved the footer inside `<main>` only on pages that *ended* with
+  `</main>`; a page with a `<script>` after it kept the footer as a direct child of `body` — a flex
+  sibling of the sidebar. Moved inside `<main class="content">` on all 16, each move asserted to have
+  landed inside.
+- **Both are now gated, each red-proofed by putting the live defect back** (`site_gates.py` gate 3b
+  and the new gate 22; decoys `site/footer-outside-main` and `site/shell-css-leak-in-media`).
+
+## 2026-10-11 (M3) — the whole site moves into the controller's shell
 
 The operator's own direction, after three mockups were rejected: the **controller's frame** (left
 sidebar, phone top bar, off-canvas drawer) around **the website's own content**. Switched in one