From a9133903224330565beb376be063bc59eeccb9b6 Mon Sep 17 00:00:00 2001
From: Edward Langley <el-github@elangley.org>
Date: Mon, 21 Sep 2026 23:24:38 -0700
Subject: [PATCH] feat: the page bar becomes a control, and `[`/`]` reach the
 modes about the document
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Three tickets on one subject: paging.

improvise-rw6l — clicking a page chip opens a menu of that axis's items, with
a down-arrow saying so. Almost nothing had to be derived: `PageAxis::items` was
already projected and `ModelCommand::SetPageSelection` was already on the wire.
The one gap was that `RenderPageAxis` dropped the names, keeping the position
and count improvise-p1ym needed for `[`/`]`. `items` is relayed INSIDE `cycle`
rather than beside it, so there is one predicate and not two: a Filter axis has
no cycle, therefore no arrow, no menu, no items, and composes nothing if a click
reaches it — all four off `PageAxis::cycle`, the one home of gotcha 27's split.

improvise-771e — the browser tile bar said "Quarter Pag". `short_label` is three
characters so the TUI aligns in a fixed-width terminal (gotcha 26); a browser has
no column to fit. `render.rs` asks `plain_label` instead; `short_label` is
untouched. Note what pinned the truncation was `harness/tests/drill.spec.js`,
not `tests/grid_frame_agreement.rs`, which does not mention the tile bar.

improvise-wiew — `[`/`]` move from Normal onto a new `document_keymap` between
the global and the modes whose subject IS the document (Normal, the three
panels, TileSelect; RecordsNormal through Normal). Help and the import wizard
take the global directly. Per the user's ruling: an overlay has a subject of its
own, and these are the first shared keys that MUTATE the document — under the
help overlay that is an invisible change to the grid it is covering.

Boundary crossings to flag, both outside the brief's stated write scope and
neither a wire change:
  - `resolve.rs` gains `Click::PageItem` and `lib.rs` a `click_page_item`
    binding. Composing the command anywhere else would put a derivation in the
    page, which design-principles §5 forbids; `Command`/`ModelCommand` are
    untouched.
  - `improvise-command/src/command/keymap.rs` gains three tests. The two
    existing guards were the mechanism that forced the wiew argument, so they
    had to be answered where they live.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FC2jVn8Au7E9JPkw6bwXXH
---
 context/code/clients.md                       |  78 +++++++++-
 context/code/protocol.md                      |  63 ++++++++
 context/test-evidence/clients.md              |  96 ++++++++++++
 context/test-evidence/protocol.md             |  87 +++++++++++
 .../improvise-command/src/command/keymap.rs   | 108 +++++++++++++
 crates/improvise-protocol/src/keymap.rs       |  28 +++-
 crates/improvise-protocol/src/resolve.rs      |  66 ++++++++
 crates/improvise-wasm-client/src/lib.rs       |   7 +
 crates/improvise-wasm-client/src/render.rs    |  81 ++++++++--
 harness/tests/contract.spec.js                |   6 +-
 harness/tests/drill.spec.js                   |   4 +-
 harness/tests/pages.spec.js                   | 125 +++++++++++++++
 web/app.js                                    | 147 ++++++++++++++++--
 web/index.html                                |   7 +
 web/shell.js                                  |  55 ++++++-
 web/style.css                                 |  70 ++++++++-
 web/tests/dispatch.test.js                    |  90 ++++++++++-
 17 files changed, 1076 insertions(+), 42 deletions(-)
 create mode 100644 harness/tests/pages.spec.js

diff --git a/context/code/clients.md b/context/code/clients.md
index 514b4281..9b5b21d2 100644
--- a/context/code/clients.md
+++ b/context/code/clients.md
@@ -1630,10 +1630,80 @@ now spelled twice — `src/ui/grid.rs` and `web/shell.js::gridPlan` — which fa
 the `indicator()` precedent (the collapse glyph is spelled in both ends while
 the *fact* is shared) and because the alternative is a protocol-level string
 table for one word; `invariant(a_ui_string_is_keyed_by_a_command_never_by_a_key)`
-is where it belongs if a second such string ever appears. And the page bar is
-**display-only**: it shows the cycle position so `[`/`]` have an affordance, but
-clicking an entry does nothing, because composing a page change from a click
-needs a `click_page` on the wasm surface that does not exist yet.
+is where it belongs if a second such string ever appears.
+
+### The page bar is a control (improvise-rw6l)
+
+It was display-only until improvise-rw6l, and this file said so, naming the
+missing `click_page` on the wasm surface as the reason. That is now built, and
+it cost less than the note implied: `PageAxis::items` was already projected and
+`ModelCommand::SetPageSelection` was already on the wire. The only gap was that
+`RenderPageAxis` **dropped the names** — it carried a position and a count,
+which is everything improvise-p1ym needed for `[`/`]` and nothing a menu needs.
+
+**The widening puts `items` inside `cycle`, not beside it.** That is the whole
+design: one predicate, not two. A Filter axis has no `cycle`, so it has no
+items, no down-arrow, is painted as a `<span>` rather than a `<button>`, and
+composes nothing if a click somehow reaches it — four answers off the one home
+gotcha 27 already named. The tempting alternative, a separate `items` field on
+`RenderPageAxis`, gives the page a second thing to test and a way to disagree
+with itself.
+
+**The arrow follows `navigable`, never "is this on the page bar".** A Filter
+axis *is* on the bar and must get no arrow: offering a menu that cannot do
+anything is worse than offering none, because the user spends a click finding
+out. A fixture with only a Page axis cannot see that distinction, which is why
+every test of it — Rust, `node --test`, and the two in the harness — carries a
+Filter axis beside the Page one wherever the shape allows.
+
+**The menu lives outside `#pages`**, in its own `#page-menu` element. `#pages`
+has its children replaced on every repaint, and a menu rebuilt under the pointer
+is a menu that closes itself mid-click. Being outside also means a repaint is
+free to *close* it explicitly, which is the honest answer: the chip it was
+anchored to has just been rebuilt, and re-anchoring would be a guess.
+
+**Two surfaces, one meaning.** improvise-b8ot's ruling is one delegated handler
+per surface on a stable root, operands off data attributes, no structural
+inference — so the chip carries `data-axis` and a menu row carries `data-axis`
+*and* `data-item`, and `shell.js::pageItemAction` is one function over both.
+That single attribute is the entire disambiguation. The case that has to be got
+right is an **empty** `data-item`: `Number("")` is `0`, so a naive read turns
+"no item" into "item zero", and clicking the chip of an axis already on its
+first item would silently re-pick it instead of opening anything.
+
+**The menu owns the keyboard while it is open, which is the third state of the
+kind `editing` already was.** `app.js`'s window `keydown` handler returns early
+for `editing`; it now returns early for `pageMenu` too, into `menuKey`. That is
+the same mechanism rather than a new one, and the reason is specific: the menu
+hangs over the grid, so letting `hjkl` through would scroll the thing it is
+pointing at out from under it. `menuKey` therefore answers **every** key, with
+`ignore` as the default — an unlisted key is swallowed, not passed on — and
+`count: 0` closes rather than wrapping modulo zero, because a repaint can empty
+an axis under an open menu.
+
+`active` and `selected` are two different facts and both are drawn: `selected`
+is what the *document* is fixed to (a check mark), `active` is where the
+keyboard is. They start together and part on the first arrow key.
+
+### The tile bar says "Page", not "Pag" (improvise-771e)
+
+`Axis::short_label` returns three characters so the TUI's tile bar aligns in a
+fixed-width terminal (gotcha 26); `"·"` for `None` is an answer there and not a
+truncation. A browser has no column to fit, so the abbreviation buys nothing and
+reads to a user as a rendering defect — it was spotted in PR #120's screenshots
+as "Quarter Pag". `render.rs` now asks `plain_label`, which is the other
+vocabulary already on `Axis`: the decision stays in the shared layer and only
+the *choice* of vocabulary is local, exactly the `AxisColor → ratatui::Color`
+shape §5 blesses. `short_label` is untouched, because the TUI's width is
+load-bearing.
+
+**Where the truncation was actually pinned is the useful part.** The ticket says
+to check `tests/grid_frame_agreement.rs` first, in case it fixes the tile bar's
+vocabulary. It does not mention the tile bar at all. What held "Quarter Pag" in
+place was `harness/tests/drill.spec.js`, which asserts the tile's text exactly —
+a Playwright spec in a directory the ticket never names, asserting the bug as
+its expected value. A rendered string is pinned wherever anything reads the
+screen, so the search is for the *string*, not for the function that makes it.
 
 ### The keyboard
 
diff --git a/context/code/protocol.md b/context/code/protocol.md
index f1af8157..a2b49452 100644
--- a/context/code/protocol.md
+++ b/context/code/protocol.md
@@ -1487,6 +1487,69 @@ agreement, and it is what `no_mode_binds_a_panel_toggle_of_its_own` now keeps.
 is the thing being removed, and two equal-but-separate globals would satisfy
 any value comparison.
 
+### The layer below it: `document_keymap` (improvise-wiew)
+
+The global above is **all seventeen modes**, and that answer is unchanged. What
+changed is that a second shared layer now sits between it and some of them.
+
+The user asked for `[`/`]` "probably in the global keymap", and then answered
+the question that left open himself: *"the help display (and import wizard,
+etc.) shouldn't respond to these keys, but the 'normal' modes like
+categories/tiles/formulas/etc. should."*
+
+The distinction is **is the document the subject right now**, and it is genuinely
+a new one rather than the global's argument applied again. Every key the global
+binds — `?`, `:`, the three panel toggles — is a **UI** action: it opens a room,
+it does not change the document. `page-next`/`page-prev` resolve to
+`ModelCommand::SetPageSelection`, which is `EffectKind::Model(LayoutGeometry)`.
+They are the first shared keys that *mutate*, and under the Help overlay that is
+a mutation the user cannot see, because the overlay is covering the grid it
+just moved.
+
+So the chain is three deep:
+
+```
+global                  ?  :  and the panel toggles          — every mode
+  └── document          [  ]                                 — Normal, the three
+        ├── Normal          panels, TileSelect; RecordsNormal
+        │     └── RecordsNormal                                through Normal
+        ├── FormulaPanel / CategoryPanel / ViewPanel
+        └── TileSelect
+  ├── Help               global only
+  ├── ImportWizard       global only
+  └── the text-entry modes and SearchMode, unchanged
+```
+
+**The name is the guard.** Nothing in the code stops the next person parenting
+Help to this layer; what stops them is that it is called `document_keymap` and
+not `common` or `shared`, so its name answers "should my new mode inherit it?"
+A name describing *who* inherits a layer invites everything to join, and the
+distinction dies within two waves.
+
+**RecordsNormal inherits it once, through Normal.** It already parents Normal,
+so it reaches the layer transitively; parenting it directly as well would be a
+second edge with nothing keeping the two in step. That is asserted on the chain
+identity rather than on behaviour, because a duplicated edge answers `page-next`
+correctly on the day it is added and only diverges later.
+
+**The layer owes the same Char-only constraint as the global**, and for the
+identical reason: it is reached by *fallthrough*, so a non-Char binding on it
+would fire while a descendant's `AnyChar` was typing. No descendant binds
+`AnyChar` today — which is exactly why the constraint is written down as a test
+rather than left to the arrangement. An unconstrained shared layer is where the
+next "just put it in the shared one" lands.
+
+**Two things that cost nothing and were checked anyway.** Text entry is
+untouched: the fallback order is `exact → Char(NONE) → AnyChar → Any → parent`,
+so a text-entry mode's `AnyChar` answers `[` before any parent is consulted, and
+typing a bracket into a category name still types it. And `prefix_keymaps_are_flat`
+is about the *sub-keymaps a prefix opens*, not about inheritance depth, so a
+third level of parents violates nothing it asserts.
+
+The browser follows for free: `resolve.rs` dispatches on the command **name**,
+so every mode that now reaches `page-next` composes it in a browser too, with no
+second table to update. That is the whole point of one shared keymap.
+
 ### `TextEntry` — the spine, welded
 
 Every text-entry mode has the same spine: characters append to one named
diff --git a/context/test-evidence/clients.md b/context/test-evidence/clients.md
index 100121f7..60f4ae6f 100644
--- a/context/test-evidence/clients.md
+++ b/context/test-evidence/clients.md
@@ -3362,3 +3362,99 @@ pin the palette, which is not the property.
 Inject the fault by moving the `td:not(.editable)` block back below
 `td.error`, or by restoring either `color: var(--fg)`; each makes exactly the
 corresponding assertion fail.
+
+## The page bar as a control (improvise-rw6l, improvise-771e)
+
+The user, pointing at the deployed page: *"The page items here (e.g. Quarter)
+should open a menu when clicked so you can pick a different option. Also have a
+down-arrow indicating there's a menu."*
+
+Almost nothing had to be derived for it. `PageAxis::items` was already
+projected, `ModelCommand::SetPageSelection` was already on the wire, and the one
+gap was that `RenderPageAxis` **threw the names away**: it kept a position and a
+count, which is all improvise-p1ym needed for `[`/`]`. Widening
+`RenderPageCycle` with `items` is a pass-through, and the items sit *inside*
+`cycle` rather than beside it so there is exactly one predicate — a Filter axis
+has no `cycle`, therefore no arrow, no menu and no items, all off the one home
+`PageAxis::cycle` already was (gotcha 27).
+
+### render::tests::a_navigable_page_axis_relays_the_items_it_may_be_fixed_to
+
+Fault injected: `items: Vec::new()` in place of `items.to_vec()`. The menu then
+paints empty on a real document and nothing in the browser can see it, because
+the down-arrow and the chip's `<button>`-ness both follow `cycle`, not the item
+count. The Filter row in the same test is what stops the fix being "just always
+send `axis.items`".
+
+### render::tests::a_page_tile_reads_as_the_whole_word
+
+improvise-771e: the browser tile bar drew "Quarter Pag". `Axis::short_label`
+returns three characters so the TUI's tile bar aligns in a fixed-width terminal
+(gotcha 26); a browser has no column to fit, so the abbreviation buys nothing
+and reads as a rendering defect. `plain_label` is the other vocabulary already
+on `Axis`, so the decision stays in the shared layer and only the *choice* of
+vocabulary is local — the `AxisColor → ratatui::Color` shape §5 blesses.
+
+**The fixture is the whole test.** The two vocabularies differ on exactly two
+of five axes — `Page` ("Pag") and `None` ("·") — so `client_with_panels`, whose
+tiles are `Row` and `None`, could see the `None` half and **not** the `Page`
+half the ticket is about. This test projects a `Page` tile and a `Column` one
+beside it: the second is what says this is a change of vocabulary rather than a
+rewrite, since `Col` is spelled the same either way.
+
+**What actually pinned the truncation was not where the ticket said to look.**
+improvise-771e says to check `tests/grid_frame_agreement.rs` before changing
+anything. It does not mention the tile bar at all. The assertion that held
+"Quarter Pag" in place was `harness/tests/drill.spec.js`, which asserts the tile
+text exactly, twice — a Playwright spec, in a directory the ticket never names,
+asserting the bug as the expected value. The lesson is the search, not the file:
+a *rendered string* is pinned wherever anything reads the screen, so the grep is
+for the string, not for the function that produces it.
+
+### a page axis offers its items, and a filter axis offers none
+
+`web/tests/dispatch.test.js`. `pageEntries` relays `cycle.items` and gives a
+Filter axis `[]`. Fault injected: `items: axis.cycle ? ... : axis.items`, i.e.
+falling back to the raw list — the Filter row then offers items nothing can act
+on, which is the "a menu that cannot do anything" case the ticket names as
+worse than no menu at all.
+
+### a chip opens the menu and a menu row picks, one attribute apart
+
+improvise-b8ot's shape: one delegated handler per surface on a stable root,
+operands off data attributes, and **no structural inference**. A chip carries
+`data-axis`; a menu row carries `data-axis` *and* `data-item`. That single
+difference is the whole disambiguation, so `pageItemAction` is one function over
+both surfaces and `app.js` registers the same listener on each.
+
+The row worth having is `{ axis: "0", item: "" }` → `open`. `Number("")` is `0`,
+so a naive read turns an absent item into item zero — which picks Q1 whenever
+the user clicks the chip for an axis already on Q1, i.e. silently does nothing,
+which is indistinguishable from the menu failing to open.
+
+### the menu answers its own keys and swallows the rest
+
+The menu is the **third** state of the kind `editing` already was: `app.js`'s
+window `keydown` handler returns early for it. That is the same treatment and
+not a new mechanism, and the reason is the same — the menu is anchored to a
+chip above the grid, so `hjkl` reaching the grid would scroll the thing the menu
+is pointing at out from under it.
+
+`menuKey` therefore answers **every** key, with `ignore` as the default rather
+than a fallthrough: an unlisted key is swallowed, not passed on. The
+`count: 0` row is a real state, not defensive padding — a repaint can empty the
+axis under an open menu — and it closes rather than wrapping modulo zero.
+
+### the menu is driven by the keyboard and does not leak keys to the grid
+
+`harness/tests/pages.spec.js`, and the assertion that earns the slow test is
+the **negative** one: `j` moves the menu's active row *and* leaves
+`#grid td.selected` where it was. A spec that only checked the menu moved would
+pass with the key leaking, because both handlers would have run.
+
+### the help overlay pages nothing behind itself
+
+improvise-wiew's product call, executed. The overlay is open, `]` is pressed,
+and the page bar still reads `Quarter = Q1 (1/4)`. Its sibling — a panel open,
+`]` pressed, bar moves to Q2 *and the panel is still open* — is what stops the
+first one passing because the key does nothing anywhere.
diff --git a/context/test-evidence/protocol.md b/context/test-evidence/protocol.md
index 5446c13c..7bc0a91e 100644
--- a/context/test-evidence/protocol.md
+++ b/context/test-evidence/protocol.md
@@ -821,3 +821,90 @@ Faults injected against the fixed code:
   duplication improvise-636d removes, and it is invisible to every check that
   asks what a key *does*.
 
+
+## `src/keymap.rs` — the layer between the global and the modes (improvise-wiew)
+
+The user asked for `[`/`]` "probably in the global keymap", and then ruled on
+the question that ask left open: *"maybe we need an intermediate keymap: the
+help display (and import wizard, etc.) shouldn't respond to these keys, but the
+'normal' modes like categories/tiles/formulas/etc. should."*
+
+So the distinction is **is the document the subject right now**, and it is a new
+distinction rather than the global's. Every key the global already binds — `?`,
+`:`, the three panel toggles — is a **UI** action. `page-next`/`page-prev`
+resolve to `ModelCommand::SetPageSelection`, which is
+`EffectKind::Model(LayoutGeometry)`: the first key on a shared layer that
+*changes the document*. Under the Help overlay that is a real mutation the user
+cannot see, because the overlay is covering the grid it just moved.
+
+`document_keymap` is therefore a third level: `global → document → {Normal, the
+three panels, TileSelect}`, with Help, ImportWizard and the text-entry modes
+taking `global` directly. RecordsNormal reaches it **through Normal**, which it
+already parents — parenting it directly as well would be a second edge to keep
+in step, and `records_normal_inherits_the_document_layer_once_through_normal`
+asserts the chain is exactly four long.
+
+Two things about the name. It is `document_keymap` and not `common`/`shared`
+deliberately: the name is the only thing that tells the next person whether a
+new mode should parent it, and a name that describes *who inherits it* invites
+everything to join. And the layer is **derived** in every test off
+`Normal.parent()` rather than exported, so a test cannot restate a membership
+the table disagrees with.
+
+### the_page_keys_reach_the_document_modes_and_no_overlay
+
+The six document modes answer `]`/`[` with `page-next`/`page-prev`; Help
+resolves **neither at all** (`lookup` is `None`, not "shadowed into something
+harmless"), ImportWizard answers `handle-wizard-key` through its `Any`, and
+CategoryAdd answers `append-char` — so typing a bracket into a name still types
+it, which is the fallback order (`AnyChar` before the parent) doing its job.
+
+Fault injected: parenting Help to `document_keymap` instead of to `global`.
+`]` in Help then resolves to `page-next` and the test fails. Note the assertion
+that discriminates is the Help one — the six positive rows pass either way.
+
+### records_normal_inherits_the_document_layer_once_through_normal
+
+Fault injected: `Keymap::with_parent(document.clone())` for RecordsNormal in
+place of `with_parent(normal)`. The test above **still passes** — RecordsNormal
+answers `page-next` either way — and this one fails on the parent identity.
+That pair is the reason it is a separate test: a duplicated inheritance edge is
+invisible to every behavioural assertion on the day it is added, and only shows
+up the first time Normal and the layer disagree.
+
+### the_document_layer_binds_only_unmodified_char_keys
+
+The sibling of `the_global_binds_only_unmodified_char_keys`, and it is owed for
+exactly the same reason: the layer is reached by **fallthrough**, so a non-Char
+binding on it would fire while a descendant mode's `AnyChar` was typing. No
+descendant binds `AnyChar` today, which is precisely why the constraint has to
+be written down rather than relied on — an unconstrained new layer is where the
+next "just put it in the shared one" lands.
+
+### resolve::tests::picking_a_page_item_moves_that_axis_and_only_that_axis
+
+improvise-rw6l. `Click::PageItem { axis, item }` is the browser's "the user
+picked straight off the bar", and the property that separates it from `]` is
+that it names **one** axis: `page_cycle` is an odometer that restates every
+navigable axis and carries between them, so on a two-axis fixture `]` from
+`Year=2024, Quarter=Q4` emits both `Year=2025` and `Quarter=Q1`. The click emits
+`Quarter=Q1` alone. A fixture with one page axis cannot tell the two apart.
+
+The basis it claims is `state.basis.claim([])` — the same claim `page_cycle`
+makes, because `SetPageSelection` names no record (gotcha 43). It is written as
+that call rather than as the `published()` fixture so the test says *which*
+claim it means; the fixture carries a record tier this command must not state.
+
+### resolve::tests::a_filter_axis_offers_no_item_to_pick
+
+Gotcha 27, from the click side. A Filter axis is on the page bar and carries no
+items, so `PageAxis::cycle` is `None` and the click composes nothing — the same
+one home the arrow and the menu follow. The two out-of-range rows beside it
+(an axis past the end of the bar, an item past the end of the list) are there
+because the page names both by index, so a stale bar under a repaint is the
+realistic way either arrives.
+
+Fault injected: resolving `items` off `axis.items` directly instead of through
+`cycle()`. The Filter arm then indexes an empty `Some(vec![])` for a Page axis
+with no items — and, more to the point, the `items.is_some()` spelling is the
+one gotcha 27 says is *not* the split.
diff --git a/crates/improvise-command/src/command/keymap.rs b/crates/improvise-command/src/command/keymap.rs
index 2528002e..5c8b57c6 100644
--- a/crates/improvise-command/src/command/keymap.rs
+++ b/crates/improvise-command/src/command/keymap.rs
@@ -831,6 +831,114 @@ mod tests {
         }
     }
 
+    /// The layer between the global and the modes whose subject is the
+    /// document. Derived off Normal's parent, never restated (improvise-wiew).
+    fn document_layer(ks: &KeymapSet) -> Arc<Keymap> {
+        ks.maps()
+            .get(&ModeKey::Normal)
+            .expect("Normal has a keymap")
+            .parent()
+            .expect("Normal inherits the document layer")
+            .clone()
+    }
+
+    fn chain_of(km: &Arc<Keymap>) -> Vec<Arc<Keymap>> {
+        let mut chain = vec![km.clone()];
+        while let Some(parent) = chain.last().expect("non-empty").parent() {
+            chain.push(parent.clone());
+        }
+        chain
+    }
+
+    /// improvise-wiew: `[`/`]` page the DOCUMENT, so they reach every mode
+    /// whose subject IS the document — and no overlay with one of its own.
+    #[test]
+    fn the_page_keys_reach_the_document_modes_and_no_overlay() {
+        let ks = KeymapSet::default_keymaps();
+        for mode in [
+            ModeKey::Normal,
+            ModeKey::RecordsNormal,
+            ModeKey::CategoryPanel,
+            ModeKey::ViewPanel,
+            ModeKey::FormulaPanel,
+            ModeKey::TileSelect,
+        ] {
+            assert_eq!(
+                bound(&ks, &mode, KeyCode::Char(']')).0,
+                "page-next",
+                "{mode:?} is about the document, so it pages"
+            );
+            assert_eq!(bound(&ks, &mode, KeyCode::Char('[')).0, "page-prev");
+        }
+
+        // Help draws over the grid it would be mutating, so the keys must not
+        // reach it at all — not merely be shadowed into something harmless.
+        let help = ks.maps().get(&ModeKey::Help).expect("Help has a keymap");
+        for key in [KeyCode::Char('['), KeyCode::Char(']')] {
+            assert!(
+                help.lookup(key, KeyModifiers::NONE).is_none(),
+                "{key:?} resolves in Help; it would page the document under the overlay"
+            );
+        }
+        assert_eq!(
+            bound(&ks, &ModeKey::ImportWizard, KeyCode::Char(']')).0,
+            "handle-wizard-key",
+            "the wizard swallows everything, which is its own subject answering"
+        );
+        assert_eq!(
+            bound(&ks, &ModeKey::CategoryAdd, KeyCode::Char('[')).0,
+            "append-char",
+            "and typing a bracket into a name still types it"
+        );
+    }
+
+    /// improvise-wiew: RecordsNormal reaches the layer THROUGH Normal. Parented
+    /// directly as well it would inherit twice and drift on the next change.
+    #[test]
+    fn records_normal_inherits_the_document_layer_once_through_normal() {
+        let ks = KeymapSet::default_keymaps();
+        let document = document_layer(&ks);
+        let normal = ks.maps().get(&ModeKey::Normal).unwrap().clone();
+        let rn = ks.maps().get(&ModeKey::RecordsNormal).unwrap();
+        assert!(
+            Arc::ptr_eq(rn.parent().expect("RecordsNormal has a parent"), &normal),
+            "RecordsNormal must parent Normal, not the layer directly"
+        );
+        let chain = chain_of(rn);
+        assert_eq!(
+            chain.iter().filter(|km| Arc::ptr_eq(km, &document)).count(),
+            1,
+            "the layer appears exactly once in RecordsNormal's chain"
+        );
+        assert_eq!(chain.len(), 4, "rn -> normal -> document -> global");
+    }
+
+    /// The layer is reached by fallthrough exactly as the global is, so it owes
+    /// the same shape — improvise-636d's argument, one level down.
+    #[test]
+    fn the_document_layer_binds_only_unmodified_char_keys() {
+        let ks = KeymapSet::default_keymaps();
+        let document = document_layer(&ks);
+        assert!(
+            !Arc::ptr_eq(&document, &root_of(&document)),
+            "the layer is not the global: it must have the global as its parent"
+        );
+        for (pattern, binding) in document.bindings() {
+            let KeyPattern::Key(code, mods) = pattern else {
+                panic!("the layer binds {pattern:?}; only exact Char keys are shadowable");
+            };
+            assert!(
+                matches!(code, KeyCode::Char(_)),
+                "the layer binds {code:?}, which AnyChar cannot shadow while typing"
+            );
+            assert_eq!(*mods, KeyModifiers::NONE, "{code:?} is held with {mods:?}");
+            assert!(
+                matches!(binding, Binding::Cmd { .. }),
+                "{code:?}: a prefix here would break prefix_keymaps_are_flat"
+            );
+        }
+    }
+
     /// The keys the global binds, read off the global rather than restated —
     /// a second list here would be the duplication improvise-636d removed.
     fn global_keys(global: &Keymap) -> Vec<KeyCode> {
diff --git a/crates/improvise-protocol/src/keymap.rs b/crates/improvise-protocol/src/keymap.rs
index 7f345c7b..5d6e34a8 100644
--- a/crates/improvise-protocol/src/keymap.rs
+++ b/crates/improvise-protocol/src/keymap.rs
@@ -403,6 +403,16 @@ fn global_keymap() -> Keymap {
     km
 }
 
+/// The keys every mode answers **whose subject is the document itself**, and
+/// no overlay with one of its own: `context/code/protocol.md` §6.
+fn document_keymap(global: Arc<Keymap>) -> Keymap {
+    let none = KeyModifiers::NONE;
+    let mut km = Keymap::with_parent(global);
+    km.bind(KeyCode::Char(']'), none, "page-next");
+    km.bind(KeyCode::Char('['), none, "page-prev");
+    km
+}
+
 /// Bind every panel's toggle key. One call site -- the global -- so a new panel
 /// reaches every mode from one `Panel::ALL` entry.
 fn bind_panel_toggles(km: &mut Keymap) {
@@ -499,7 +509,10 @@ impl Keymaps {
         // ── Normal mode ──────────────────────────────────────────────────
         // The global every mode falls through to (improvise-636d).
         let global = Arc::new(global_keymap());
-        let mut normal = Keymap::with_parent(global.clone());
+        // The layer between: the modes whose subject is the document. Help
+        // and the wizard take `global` directly — §6 (improvise-wiew).
+        let document = Arc::new(document_keymap(global.clone()));
+        let mut normal = Keymap::with_parent(document.clone());
 
         // Navigation
         for (key, dr, dc) in [
@@ -611,9 +624,8 @@ impl Keymaps {
         );
         normal.bind(KeyCode::Char('N'), none, "search-or-category-add");
 
-        // Page navigation
-        normal.bind(KeyCode::Char(']'), none, "page-next");
-        normal.bind(KeyCode::Char('['), none, "page-prev");
+        // Page navigation is the global's, exactly as `?` and `:` are: a
+        // mode that re-bound it would be the asymmetry §6 removed.
 
         // Group / hide
         normal.bind(KeyCode::Char('z'), none, "toggle-group-under-cursor");
@@ -709,7 +721,7 @@ impl Keymaps {
         set.insert(
             ModeKey::FormulaPanel,
             Arc::new(
-                PanelKeymap::new(Panel::Formula, global.clone())
+                PanelKeymap::new(Panel::Formula, document.clone())
                     .bind_args(
                         KeyCode::Char('a'),
                         "enter-mode",
@@ -733,7 +745,7 @@ impl Keymaps {
         set.insert(
             ModeKey::CategoryPanel,
             Arc::new(
-                PanelKeymap::new(Panel::Category, global.clone())
+                PanelKeymap::new(Panel::Category, document.clone())
                     .bind(KeyCode::Enter, "filter-to-item")
                     .bind(KeyCode::Char(' '), "cycle-axis-at-cursor")
                     .bind_args(
@@ -751,7 +763,7 @@ impl Keymaps {
         set.insert(
             ModeKey::ViewPanel,
             Arc::new(
-                PanelKeymap::new(Panel::View, global.clone())
+                PanelKeymap::new(Panel::View, document.clone())
                     .bind(KeyCode::Enter, "switch-view-at-cursor")
                     .bind(KeyCode::Char('n'), "create-and-switch-view")
                     .bind(KeyCode::Char('o'), "create-and-switch-view")
@@ -761,7 +773,7 @@ impl Keymaps {
         );
 
         // ── Tile select ──────────────────────────────────────────────────
-        let mut ts = Keymap::with_parent(global.clone());
+        let mut ts = Keymap::with_parent(document.clone());
         ts.bind_args(KeyCode::Esc, none, "enter-mode", vec!["normal".into()]);
         ts.bind_args(KeyCode::Tab, none, "enter-mode", vec!["normal".into()]);
         ts.bind_args(KeyCode::Left, none, "move-tile-cursor", vec!["-1".into()]);
diff --git a/crates/improvise-protocol/src/resolve.rs b/crates/improvise-protocol/src/resolve.rs
index 03f60922..6a3f8c09 100644
--- a/crates/improvise-protocol/src/resolve.rs
+++ b/crates/improvise-protocol/src/resolve.rs
@@ -84,6 +84,9 @@ pub enum Click {
     Tile(usize),
     /// A grid cell, in **viewport** coordinates as the page painted it.
     Cell { row: usize, col: usize },
+    /// An item of a page axis, both named by index: `axis` into
+    /// `GridCache::page_coords`, `item` into that axis's own item list.
+    PageItem { axis: usize, item: usize },
 }
 
 /// One rule for all the widgets: **go there, then do what `Enter` does
@@ -97,6 +100,7 @@ pub fn click(state: &ClientState, target: Click) -> Vec<Command> {
         Click::FormulaRow(index) => click_formula_row(state, index),
         Click::Tile(index) => click_tile(state, index),
         Click::Cell { row, col } => click_cell(state, row, col),
+        Click::PageItem { axis, item } => click_page_item(state, axis, item),
     };
     resolved.unwrap_or_default()
 }
@@ -720,6 +724,22 @@ fn page_cycle(state: &ClientState, way: Odometer) -> Option<Vec<Command>> {
         .collect()
 }
 
+/// The user picked an item off a page axis directly. `PageAxis::cycle` is
+/// asked, so a Filter axis composes nothing (gotcha 27).
+fn click_page_item(state: &ClientState, axis: usize, item: usize) -> Option<Vec<Command>> {
+    let page = state.grid.page_coords.get(axis)?;
+    let (items, _) = page.cycle()?;
+    let picked = items.get(item)?;
+    Some(vec![Command::Model {
+        basis: state.basis.claim([])?,
+        cmd: ModelCommand::SetPageSelection {
+            view: state.view.active_view.clone(),
+            category: page.cat.clone(),
+            item: picked.clone(),
+        },
+    }])
+}
+
 /// Which way `CommitAndAdvance` moves after writing.
 #[derive(Debug, Clone, Copy)]
 enum Advance {
@@ -2715,6 +2735,52 @@ mod tests {
         assert_eq!(moved.len(), 2, "the two Page axes still move: {moved:?}");
     }
 
+    /// improvise-rw6l: picking straight off the bar, which is what a menu
+    /// does. It names ONE axis, so the odometer's carry must not appear.
+    #[test]
+    fn picking_a_page_item_moves_that_axis_and_only_that_axis() {
+        let state = paged();
+        let commands = click(&state, Click::PageItem { axis: 1, item: 0 });
+        assert_eq!(
+            commands,
+            vec![Command::Model {
+                // The claim `]` makes for the same document: this command
+                // names no record, so the record tier is omitted (gotcha 43).
+                basis: state.basis.claim([]).expect("the basis is published"),
+                cmd: ModelCommand::SetPageSelection {
+                    view: state.view.active_view.clone(),
+                    category: "Quarter".into(),
+                    item: "Q1".into(),
+                },
+            }],
+            "Q4 to Q1 directly: Year is untouched, where `]` would have carried into it"
+        );
+    }
+
+    /// A Filter axis is on the bar and offers nothing to pick (gotcha 27).
+    /// Offering a menu that cannot act is worse than offering none.
+    #[test]
+    fn a_filter_axis_offers_no_item_to_pick() {
+        let mut state = paged();
+        state.grid.page_coords.push(crate::PageAxis {
+            cat: "Region".into(),
+            selected: "East".into(),
+            items: None,
+        });
+        assert!(
+            click(&state, Click::PageItem { axis: 2, item: 0 }).is_empty(),
+            "a Filter axis has no item list, so there is nothing a click could mean"
+        );
+        assert!(
+            click(&state, Click::PageItem { axis: 9, item: 0 }).is_empty(),
+            "and an axis that is not on the bar is not one either"
+        );
+        assert!(
+            click(&state, Click::PageItem { axis: 0, item: 9 }).is_empty(),
+            "nor an item past the end of the list the bar was painted from"
+        );
+    }
+
     /// Enter in `Editing` is `[commit-cell-edit editing, clear-buffer edit]`
     /// and every part shows up here — including the odd-looking trailing clear.
     #[test]
diff --git a/crates/improvise-wasm-client/src/lib.rs b/crates/improvise-wasm-client/src/lib.rs
index d736ce26..a6b89e9a 100644
--- a/crates/improvise-wasm-client/src/lib.rs
+++ b/crates/improvise-wasm-client/src/lib.rs
@@ -139,6 +139,13 @@ impl Client {
         self.clicked(Click::Tile(index))
     }
 
+    /// The user picked item `item` off page axis `axis`, both indexing what
+    /// the bar was painted from. A Filter axis composes nothing (gotcha 27).
+    #[wasm_bindgen]
+    pub fn click_page_item(&mut self, axis: usize, item: usize) -> Result<Option<String>, JsError> {
+        self.clicked(Click::PageItem { axis, item })
+    }
+
     /// The user clicked a grid cell at a **viewport** position: move the
     /// selection there. Opening an editor is the page's own affair.
     #[wasm_bindgen]
diff --git a/crates/improvise-wasm-client/src/render.rs b/crates/improvise-wasm-client/src/render.rs
index e7f52e6b..33d9e11d 100644
--- a/crates/improvise-wasm-client/src/render.rs
+++ b/crates/improvise-wasm-client/src/render.rs
@@ -162,12 +162,15 @@ pub struct RenderPageAxis {
     pub cycle: Option<RenderPageCycle>,
 }
 
-/// A position in a navigable page axis, as a named pair: two bare counts read
-/// as "how many page axes are there".
-#[derive(Debug, Clone, Copy, Default, PartialEq, Serialize)]
+/// A position in a navigable page axis, and what it may be moved to. `items`
+/// is RELAYED, never derived here — `context/code/clients.md`.
+#[derive(Debug, Clone, Default, PartialEq, Serialize)]
 pub struct RenderPageCycle {
     pub at: usize,
     pub count: usize,
+    /// Every item this axis can be fixed to, in the document's own order, so
+    /// a page may offer them. `count` is `items.len()` and stays for `[`/`]`.
+    pub items: Vec<String>,
 }
 
 /// One row of the category panel. `item_count` and `expanded` are a category
@@ -206,8 +209,8 @@ pub struct RenderFormulaRow {
     pub selected: bool,
 }
 
-/// One tile: a category the view may move between axes, with `Axis`'s own
-/// short label (`"·"` for None is an answer, not a truncation — gotcha 26).
+/// One tile: a category the view may move between axes, with `Axis`'s PLAIN
+/// label — the three-column one is the TUI's (improvise-771e, clients.md).
 #[derive(Debug, Clone, Default, PartialEq, Serialize)]
 pub struct RenderTile {
     pub category: String,
@@ -330,6 +333,7 @@ fn page_axes(state: &ClientState) -> Vec<RenderPageAxis> {
             cycle: axis.cycle().map(|(items, at)| RenderPageCycle {
                 at,
                 count: items.len(),
+                items: items.to_vec(),
             }),
         })
         .collect()
@@ -383,7 +387,7 @@ fn tile_row(state: &ClientState) -> Vec<RenderTile> {
         .enumerate()
         .map(|(i, tile)| RenderTile {
             category: tile.category.clone(),
-            axis: tile.axis.short_label(),
+            axis: tile.axis.plain_label(),
             selected: selecting && i == state.view.tile_cat_idx,
         })
         .collect()
@@ -829,8 +833,8 @@ mod tests {
         let normal = client.model();
         assert_eq!(normal.tiles.len(), 2);
         assert_eq!(
-            normal.tiles[1].axis, "·",
-            "None is an answer, not a truncation"
+            normal.tiles[1].axis, "None",
+            "the browser has no column to fit, so the word is the answer"
         );
         assert!(
             !normal.tiles.iter().any(|t| t.selected),
@@ -992,7 +996,11 @@ mod tests {
         let bar = client.model().page_coords;
         assert_eq!(
             bar[0].cycle,
-            Some(RenderPageCycle { at: 2, count: 3 }),
+            Some(RenderPageCycle {
+                at: 2,
+                count: 3,
+                items: vec!["Q1".into(), "Q2".into(), "Q3".into()],
+            }),
             "a Page axis says where `[`/`]` currently sit"
         );
         assert_eq!(
@@ -1005,6 +1013,61 @@ mod tests {
         );
     }
 
+    /// improvise-rw6l: the names were projected and the browser dropped them,
+    /// so the bar could say WHERE the cycle sits and not WHAT it may move to.
+    #[test]
+    fn a_navigable_page_axis_relays_the_items_it_may_be_fixed_to() {
+        let mut client = init();
+        client
+            .state_mut()
+            .fold(&Command::Project(Projection::Grid(GridUpdate::PageCoords(
+                vec![
+                    page_axis("Quarter", "Q3", Some(vec!["Q1", "Q2", "Q3"])),
+                    page_axis("Region", "North", None),
+                ],
+            ))));
+        let bar = client.model().page_coords;
+        let cycle = bar[0].cycle.as_ref().expect("a Page axis cycles");
+        assert_eq!(
+            cycle.items,
+            vec!["Q1", "Q2", "Q3"],
+            "the document's own order, relayed — a menu offers these and the page derives none of them"
+        );
+        assert_eq!(cycle.at, 2, "and which of them it is fixed to now");
+        assert!(
+            bar[1].cycle.is_none(),
+            "a Filter axis offers NO items: there is nothing a menu could pick (gotcha 27)"
+        );
+    }
+
+    /// improvise-771e: `short_label` is three columns so the TUI's tile bar
+    /// aligns; a browser has no column to fit and read "Pag" as a bug.
+    #[test]
+    fn a_page_tile_reads_as_the_whole_word() {
+        let mut client = init();
+        client
+            .state_mut()
+            .fold(&Command::Project(Projection::TileBar(vec![
+                TileEntry {
+                    category: "Quarter".into(),
+                    axis: Axis::Page,
+                },
+                TileEntry {
+                    category: "Region".into(),
+                    axis: Axis::Column,
+                },
+            ])));
+        let tiles = client.model().tiles;
+        assert_eq!(
+            tiles[0].axis, "Page",
+            "the ONE label the two vocabularies disagree on among the visible axes"
+        );
+        assert_eq!(
+            tiles[1].axis, "Col",
+            "and the ones they agree on are unchanged, so this is a choice of vocabulary and not a rewrite"
+        );
+    }
+
     /// The totals strip is relayed, never summed here: the total runs over
     /// rows this viewport scrolled past (gotcha 44). improvise-p7lh.
     #[test]
diff --git a/harness/tests/contract.spec.js b/harness/tests/contract.spec.js
index b679cfac..34e764b2 100644
--- a/harness/tests/contract.spec.js
+++ b/harness/tests/contract.spec.js
@@ -47,7 +47,7 @@ test.describe("every field the render model projects reaches the screen", () =>
     await expect(page.locator("#status")).toContainText("in-page worker");
 
     // `tiles`: drawn on every frame, as the TUI draws them. The axis label is
-    // `Axis::short_label` — "Pag", and "·" for None, an answer not a stub.
+    // `Axis::plain_label` here; the three-column one is the TUI's (771e).
     await expect(page.locator("#tiles .tile")).toHaveCount(5);
     await expect(page.locator("#tiles .tile").filter({ hasText: "Product" })).toHaveText(
       "Product Row",
@@ -151,7 +151,7 @@ test.describe("every field the render model projects reaches the screen", () =>
 
     // `web/demo.improv` is `Quarter: page, Q1` — the fact this bead is about.
     await expect(page.locator("#pages")).toBeVisible();
-    await expect(page.locator("#pages .page-axis")).toHaveText("Quarter = Q1 (1/4)");
+    await expect(page.locator("#pages .page-axis")).toHaveText(/Quarter = Q1 \(1\/4\)/);
     // Page is navigable where Filter is not (gotcha 27), and that is what
     // says `[`/`]` will move it.
     await expect(page.locator("#pages .page-axis")).toHaveClass(/navigable/);
@@ -167,7 +167,7 @@ test.describe("every field the render model projects reaches the screen", () =>
 
     // BOTH move, and the numbers are Q2's: a bar that showed a stale quarter
     // over fresh numbers is the exact lie this bead names.
-    await expect(page.locator("#pages .page-axis")).toHaveText("Quarter = Q2 (2/4)", {
+    await expect(page.locator("#pages .page-axis")).toHaveText(/Quarter = Q2 \(2\/4\)/, {
       timeout: 10_000,
     });
     await expect(page.locator("#grid tfoot td").first()).toHaveText("66,400");
diff --git a/harness/tests/drill.spec.js b/harness/tests/drill.spec.js
index 193e41ed..1a03d702 100644
--- a/harness/tests/drill.spec.js
+++ b/harness/tests/drill.spec.js
@@ -17,14 +17,14 @@ async function anAggregateGrid(page) {
   await page.keyboard.press("Shift+T");
   const tiles = page.locator("#tiles .tile");
   const quarter = tiles.filter({ hasText: "Quarter" });
-  await expect(quarter).toHaveText("Quarter Pag");
+  await expect(quarter).toHaveText("Quarter Page");
   const names = await tiles.allTextContents();
   for (let i = 0; i < names.findIndex((t) => t.startsWith("Quarter")); i++) {
     await page.keyboard.press("ArrowRight");
   }
   await expect(quarter).toHaveClass(/selected/);
   await page.keyboard.press("Enter");
-  await expect(quarter).toHaveText("Quarter ·", { timeout: 10_000 });
+  await expect(quarter).toHaveText("Quarter None", { timeout: 10_000 });
   await page.keyboard.press("Escape");
 
   // The precondition, asserted rather than assumed: nothing is editable now,
diff --git a/harness/tests/pages.spec.js b/harness/tests/pages.spec.js
new file mode 100644
index 00000000..3a542747
--- /dev/null
+++ b/harness/tests/pages.spec.js
@@ -0,0 +1,125 @@
+// The page bar as a control: the arrow, the menu, and the keys that reach it.
+// The RULES behind each decision are `node --test web/tests` — clients.md.
+
+import { test, expect } from "@playwright/test";
+
+/// The shipped demo, focused. `#grid` must have focus or the page's keydown
+/// listener never sees a key and every key assertion would pass vacuously.
+async function theDemo(page) {
+  await page.goto("/app/index.html");
+  await expect(page.locator("#grid table")).toBeVisible({ timeout: 20_000 });
+  await page.locator("#grid").focus();
+}
+
+const chip = (page) => page.locator("#pages .page-axis");
+const menu = (page) => page.locator("#page-menu");
+
+test.describe("a page axis can be picked from a menu", () => {
+  // improvise-rw6l. The demo has Quarter on Page, so the chip is navigable and
+  // the arrow is the affordance that says a menu is behind it.
+  test("the chip carries a down-arrow and opens a menu of every item", async ({
+    page,
+  }) => {
+    await theDemo(page);
+    await expect(chip(page)).toHaveText(/Quarter = Q1 \(1\/4\)/);
+    // A <button> and not a <span>: the affordance is the element as much as
+    // the glyph, and a Filter axis gets neither.
+    await expect(chip(page)).toHaveAttribute("aria-haspopup", "listbox");
+    await expect(chip(page).locator(".page-arrow")).toHaveText("▾");
+
+    await expect(menu(page)).toBeHidden();
+    await chip(page).click();
+    await expect(menu(page)).toBeVisible();
+    await expect(menu(page).locator(".page-item")).toHaveText([
+      "✓ Q1",
+      "  Q2",
+      "  Q3",
+      "  Q4",
+    ]);
+    // Which item the document is fixed to is MARKED, so the menu says where
+    // you are as well as where you may go.
+    await expect(menu(page).locator(".page-item.selected")).toHaveText("✓ Q1");
+  });
+
+  // The whole point: the picked item reaches the host and the grid it filters
+  // changes. Nothing on the page moves the bar itself.
+  test("picking an item moves the page axis and the numbers under it", async ({
+    page,
+  }) => {
+    await theDemo(page);
+    const firstCell = page.locator("#grid td").first();
+    const q1 = await firstCell.textContent();
+
+    await chip(page).click();
+    await menu(page).getByText("Q4", { exact: false }).click();
+
+    await expect(chip(page)).toHaveText(/Quarter = Q4 \(4\/4\)/);
+    await expect(menu(page)).toBeHidden();
+    // Espresso Revenue is 18,400 in Q1 and 20,300 in Q4, so a menu that moved
+    // only the label would leave this unchanged.
+    await expect(firstCell).not.toHaveText(q1 ?? "");
+    await expect(firstCell).toHaveText("20,300");
+  });
+
+  // The menu is operable without a mouse, and while it is open it owns the
+  // keyboard: a grid key must not reach the grid behind it.
+  test("the menu is driven by the keyboard and does not leak keys to the grid", async ({
+    page,
+  }) => {
+    await theDemo(page);
+    await chip(page).click();
+    await expect(menu(page)).toBeVisible();
+
+    // `j` is the grid's own "move down" and here it moves the MENU. If it
+    // leaked, the selection under the menu would move as well.
+    const selected = page.locator("#grid td.selected");
+    const before = await selected.textContent();
+    await page.keyboard.press("j");
+    await expect(menu(page).locator(".page-item.active")).toHaveText("  Q2");
+    await expect(selected).toHaveText(before ?? "");
+
+    await page.keyboard.press("Enter");
+    await expect(menu(page)).toBeHidden();
+    await expect(chip(page)).toHaveText(/Quarter = Q2 \(2\/4\)/);
+  });
+
+  test("Escape closes the menu and changes nothing", async ({ page }) => {
+    await theDemo(page);
+    await chip(page).click();
+    await expect(menu(page)).toBeVisible();
+    await page.keyboard.press("Escape");
+    await expect(menu(page)).toBeHidden();
+    await expect(chip(page)).toHaveText(/Quarter = Q1 \(1\/4\)/);
+    // And the keyboard is the grid's again: `]` is the page key, so the bar
+    // moving here proves the menu gave it back.
+    await page.keyboard.press("]");
+    await expect(chip(page)).toHaveText(/Quarter = Q2 \(2\/4\)/);
+  });
+});
+
+test.describe("the page keys follow the document, not the grid", () => {
+  // improvise-wiew, end to end: `[`/`]` are on the layer between the global
+  // and the modes whose subject IS the document, so a panel pages too.
+  test("a panel is still somewhere the page axis can be moved", async ({ page }) => {
+    await theDemo(page);
+    await expect(chip(page)).toHaveText(/Quarter = Q1/);
+
+    await page.keyboard.press("C");
+    await expect(page.locator("#mode")).toHaveText("CATEGORIES");
+    await page.keyboard.press("]");
+    await expect(chip(page)).toHaveText(/Quarter = Q2 \(2\/4\)/);
+    // The key did its own job and no other: the panel is still open.
+    await expect(page.locator("#mode")).toHaveText("CATEGORIES");
+  });
+
+  // …and the help overlay is not, because it covers the grid it would be
+  // changing. That is the one case the layer exists to exclude.
+  test("the help overlay pages nothing behind itself", async ({ page }) => {
+    await theDemo(page);
+    await page.keyboard.press("?");
+    await expect(page.locator("#help")).toBeVisible();
+    await page.keyboard.press("]");
+    await expect(page.locator("#help")).toBeVisible();
+    await expect(chip(page)).toHaveText(/Quarter = Q1 \(1\/4\)/);
+  });
+});
diff --git a/web/app.js b/web/app.js
index a412096f..3374d13c 100644
--- a/web/app.js
+++ b/web/app.js
@@ -17,7 +17,10 @@ import {
   indexAction,
   indicator,
   isFatal,
+  MENU_ARROW,
+  menuKey,
   NOTICE_KEY,
+  pageItemAction,
   noticeOpen,
   noticeState,
   noticeText,
@@ -41,6 +44,7 @@ const el = {
   views: document.getElementById("views"),
   tiles: document.getElementById("tiles"),
   pages: document.getElementById("pages"),
+  pageMenu: document.getElementById("page-menu"),
   mode: document.getElementById("mode"),
   minibuffer: document.getElementById("minibuffer"),
   hint: document.getElementById("hint"),
@@ -233,6 +237,9 @@ function repaint() {
   }
   if (model.revision === painted) return;
   painted = model.revision;
+  // The bar is about to be rebuilt, so the chip this menu is anchored to is
+  // gone. Closing is the honest answer; re-anchoring would be a guess.
+  closePageMenu();
 
   el.mode.textContent = model.mode;
   // The mode's own hint, off `help::hint_line`. The page spells no keys
@@ -328,22 +335,94 @@ function paintHelp(model) {
  *  (improvise-p7lh). An empty list is what says there are none. */
 function paintPages(model) {
   const entries = pageEntries(model);
+  pageAxes = entries;
   el.pages.hidden = entries.length === 0;
-  el.pages.replaceChildren(
-    ...entries.map((entry) => {
-      const span = document.createElement("span");
-      span.className = "page-axis";
-      // Only a Page axis is navigable; a Filter one constrains without being
-      // something `[`/`]` can move (gotcha 27).
-      span.classList.toggle("navigable", entry.navigable);
-      span.textContent = entry.position
-        ? `${entry.text} (${entry.position})`
-        : entry.text;
-      return span;
+  el.pages.replaceChildren(...entries.map(pageChip));
+}
+
+/** One page chip. A navigable axis is a <button> carrying the arrow, because
+ *  it opens a menu; a Filter axis is a <span> with neither (improvise-rw6l). */
+function pageChip(entry) {
+  const label = entry.position
+    ? `${entry.text} (${entry.position})`
+    : entry.text;
+  if (!entry.navigable) {
+    // Only a Page axis is navigable; a Filter one constrains without being
+    // something `[`/`]` or a menu can move (gotcha 27).
+    const span = document.createElement("span");
+    span.className = "page-axis";
+    span.textContent = label;
+    return span;
+  }
+  const button = document.createElement("button");
+  button.type = "button";
+  button.className = "page-axis navigable";
+  button.dataset.axis = String(entry.index);
+  button.setAttribute("aria-haspopup", "listbox");
+  const text = document.createElement("span");
+  text.textContent = label;
+  const arrow = document.createElement("span");
+  arrow.className = "page-arrow";
+  arrow.setAttribute("aria-hidden", "true");
+  arrow.textContent = MENU_ARROW;
+  button.append(text, arrow);
+  return button;
+}
+
+// The open page menu: which axis it belongs to and which row the keyboard is
+// on. `null` is closed, exactly as `editing` is.
+let pageAxes = [];
+let pageMenu = null;
+
+function openPageMenu(axis, anchor) {
+  const entry = pageAxes[axis];
+  // A Filter axis carries no items, so there is nothing to open — the chip is
+  // not a button either, and both answers come off the same `cycle`.
+  if (!entry || entry.items.length === 0) return;
+  const active = entry.items.findIndex((item) => item.selected);
+  pageMenu = { axis, active: active === -1 ? 0 : active };
+  const box = anchor.getBoundingClientRect();
+  el.pageMenu.style.left = `${Math.round(box.left)}px`;
+  el.pageMenu.style.top = `${Math.round(box.bottom + 2)}px`;
+  paintPageMenu();
+}
+
+function closePageMenu() {
+  pageMenu = null;
+  el.pageMenu.hidden = true;
+  el.pageMenu.replaceChildren();
+}
+
+function paintPageMenu() {
+  if (!pageMenu) return;
+  const entry = pageAxes[pageMenu.axis];
+  el.pageMenu.hidden = false;
+  el.pageMenu.replaceChildren(
+    ...entry.items.map((item) => {
+      const row = document.createElement("button");
+      row.type = "button";
+      row.className = "page-item";
+      row.setAttribute("role", "option");
+      // `selected` is which item the DOCUMENT is fixed to; `active` is where
+      // the keyboard is. They start together and part on the first arrow key.
+      row.setAttribute("aria-selected", String(item.selected));
+      row.classList.toggle("selected", item.selected);
+      row.classList.toggle("active", item.item === pageMenu.active);
+      row.dataset.axis = String(entry.index);
+      row.dataset.item = String(item.item);
+      row.textContent = item.selected ? `✓ ${item.label}` : `  ${item.label}`;
+      return row;
     }),
   );
 }
 
+/** Pick an item and close. The value that comes back is the server's: nothing
+ *  here moves the page bar, exactly as an editor does not move a cell. */
+function pickPageItem(axis, item) {
+  closePageMenu();
+  send(client.click_page_item(axis, item));
+}
+
 /** The grid, painted from a plan in which every decision is already made
  *  (improvise-anr9, and `shell.js::gridPlan`). Nothing below decides anything:
  *  which labels repeat and are therefore blank, how many header levels there
@@ -586,6 +665,31 @@ function delegateIndex(root, selector, act) {
   });
 }
 
+// The page bar and its menu are two surfaces and ONE meaning: an action with
+// an axis and maybe an item, decided in `shell.js` (improvise-rw6l, b8ot).
+for (const root of [el.pages, el.pageMenu]) {
+  root.addEventListener("click", (event) => {
+    const node = event.target.closest("[data-axis]");
+    if (!node || !root.contains(node)) return;
+    const action = pageItemAction(node.dataset);
+    if (!action) return;
+    // A live editor closes first, before the click is composed — the order
+    // the grid's own handler keeps, and for the same reason.
+    if (editing) editing.close(false);
+    if (action.kind === "pick") pickPageItem(action.axis, action.item);
+    else if (pageMenu && pageMenu.axis === action.axis) closePageMenu();
+    else openPageMenu(action.axis, node);
+  });
+}
+
+// A click anywhere else dismisses it. `pointerdown` rather than `click`, so a
+// press that lands on the grid closes this before the grid acts on it.
+window.addEventListener("pointerdown", (event) => {
+  if (!pageMenu) return;
+  if (el.pages.contains(event.target) || el.pageMenu.contains(event.target)) return;
+  closePageMenu();
+});
+
 delegateIndex(el.categories, "li[data-index]", (i) => client.click_category(i));
 delegateIndex(el.views, "li[data-index]", (i) => client.click_view(i));
 delegateIndex(el.formulas, "li[data-index]", (i) => client.click_formula(i));
@@ -617,10 +721,31 @@ el.grid.addEventListener("mousedown", (event) => {
 // The keyboard. `Client::on_key_event` folds locally where it can; what it
 // cannot answer is sent and does nothing, by design (context/code/clients.md).
 
+/** The menu's own keys, decided in `shell.js`. Every one of them is claimed:
+ *  an unhandled key must not reach the grid behind the menu. */
+function pageMenuKeydown(e) {
+  const items = pageAxes[pageMenu.axis].items;
+  const action = menuKey(e.key, { count: items.length, active: pageMenu.active });
+  if (action.kind === "close") closePageMenu();
+  else if (action.kind === "pick") pickPageItem(pageMenu.axis, action.item);
+  else if (action.kind === "move") {
+    pageMenu.active = action.item;
+    paintPageMenu();
+  }
+  // Ctrl and Alt combinations stay the browser's, exactly as below.
+  if (!e.ctrlKey && !e.altKey) e.preventDefault();
+}
+
 window.addEventListener("keydown", (e) => {
   // The inline editor owns the keyboard while it is open. It stops
   // propagation, so this is belt and braces for a focus we did not predict.
   if (editing) return;
+  // …and so does the page menu, which is the third state of that kind: it is
+  // anchored to the grid, so letting `hjkl` through would scroll it away.
+  if (pageMenu) {
+    pageMenuKeydown(e);
+    return;
+  }
   // BrowserKey drops anything held with the platform meta key precisely so
   // Cmd-R does not arrive as a bare `r`; do not claim it here either.
   if (e.metaKey) return;
diff --git a/web/index.html b/web/index.html
index 5ed8239a..86b70525 100644
--- a/web/index.html
+++ b/web/index.html
@@ -108,6 +108,13 @@
     -->
     <div id="pages" aria-label="Page axes" hidden></div>
 
+    <!--
+      The page menu (improvise-rw6l). It lives OUTSIDE #pages because that
+      element's children are replaced on every repaint, and a menu rebuilt
+      under the pointer is a menu that closes itself.
+    -->
+    <div id="page-menu" role="listbox" aria-label="Page items" hidden></div>
+
     <main>
       <aside id="categories" aria-label="Categories"></aside>
       <aside id="formulas" aria-label="Formulas" hidden></aside>
diff --git a/web/shell.js b/web/shell.js
index 482e7d6e..ce3be21a 100644
--- a/web/shell.js
+++ b/web/shell.js
@@ -88,7 +88,8 @@ export function frameCostOf(model) {
 /** The page bar: one entry per category fixing which records are visible. The
  *  Page/Filter split is `cycle`, decided in Rust — gotcha 27, clients.md. */
 export function pageEntries(model) {
-  return model.page_coords.map((axis) => ({
+  return model.page_coords.map((axis, index) => ({
+    index,
     // The TUI's own spelling (`src/ui/grid.rs`); only the spelling is local,
     // exactly as `indicator`'s is.
     text: `${axis.cat} = ${axis.selected}`,
@@ -96,9 +97,61 @@ export function pageEntries(model) {
     // navigable: there is no position to show and no key that would move it.
     position: axis.cycle ? `${axis.cycle.at + 1}/${axis.cycle.count}` : "",
     navigable: Boolean(axis.cycle),
+    // The items a menu offers, RELAYED from `cycle` — the one home of the
+    // Page/Filter split, so a Filter axis offers none (improvise-rw6l).
+    items: axis.cycle
+      ? axis.cycle.items.map((label, item) => ({
+          item,
+          label,
+          selected: item === axis.cycle.at,
+        }))
+      : [],
   }));
 }
 
+/** The down-arrow. Present exactly where there is a menu to open, which is
+ *  `navigable` and never "is this on the page bar" (gotcha 27). */
+export const MENU_ARROW = "▾";
+
+/** What a delegated click on the page bar or its menu MEANS. `null` where the
+ *  node names neither, which is every gap between the chips. */
+export function pageItemAction(dataset) {
+  const axis = whole(dataset.axis);
+  if (axis === null) return null;
+  const item = whole(dataset.item);
+  // A chip carries an axis and no item: it OPENS the menu. A menu row carries
+  // both: it picks. One attribute apart, so nothing infers from structure.
+  return item === null ? { kind: "open", axis } : { kind: "pick", axis, item };
+}
+
+/** What a key does while the page menu is open. The menu owns the keyboard
+ *  exactly as the inline editor does — `context/code/clients.md`. */
+export function menuKey(key, { count, active }) {
+  if (count === 0) return { kind: "close" };
+  switch (key) {
+    case "Escape":
+    case "Tab":
+      return { kind: "close" };
+    case "Enter":
+    case " ":
+      return { kind: "pick", item: active };
+    case "ArrowDown":
+    case "j":
+      return { kind: "move", item: (active + 1) % count };
+    case "ArrowUp":
+    case "k":
+      return { kind: "move", item: (active - 1 + count) % count };
+    case "Home":
+      return { kind: "move", item: 0 };
+    case "End":
+      return { kind: "move", item: count - 1 };
+    // Everything else is SWALLOWED rather than passed on: a menu that let
+    // `hjkl` reach the grid would scroll the thing it is anchored to.
+    default:
+      return { kind: "ignore" };
+  }
+}
+
 /** The grid, as a description: every decision made, nothing created. The page
  *  decides nothing here and neither does this — which labels repeat, how many
  *  header levels there are and which columns share a group all arrive
diff --git a/web/style.css b/web/style.css
index e6aab1c5..b1a46a4c 100644
--- a/web/style.css
+++ b/web/style.css
@@ -269,12 +269,78 @@ header h1 {
   font-size: 0.85em;
 }
 
-/* A Page axis can be moved with `[`/`]`; a Filter one constrains without
-   being navigable (gotcha 27), which is why only one of them lights up. */
+/* A Page axis can be moved with `[`/`]` and picked from a menu; a Filter one
+   constrains without being navigable (gotcha 27), so only one of them lights
+   up and only one of them is a button at all. */
 #pages .navigable {
   border-color: var(--accent);
   color: var(--accent);
   font-weight: 600;
+  display: inline-flex;
+  gap: 0.3rem;
+  align-items: center;
+  font: inherit;
+  font-weight: 600;
+  cursor: pointer;
+}
+
+#pages .navigable:hover {
+  background: var(--accent-soft);
+}
+
+/* The arrow says a menu is there. It follows `navigable`, never "is this on
+   the page bar" — an arrow over nothing to pick is worse than none. */
+#pages .page-arrow {
+  font-size: 1.15em;
+  line-height: 1;
+  margin-left: -0.05rem;
+}
+
+/* The menu itself. Absolutely positioned from the chip it was opened on, and
+   outside #pages so a repaint cannot rebuild it under the pointer. */
+#page-menu[hidden] {
+  display: none;
+}
+
+#page-menu {
+  position: fixed;
+  z-index: 20;
+  min-width: 9rem;
+  max-height: 60vh;
+  overflow-y: auto;
+  padding: 0.2rem;
+  display: flex;
+  flex-direction: column;
+  background: var(--raised);
+  border: 1px solid var(--line-strong);
+  border-radius: var(--radius);
+  box-shadow: var(--shadow);
+  scrollbar-width: thin;
+}
+
+#page-menu .page-item {
+  text-align: left;
+  padding: 0.2rem 0.5rem;
+  border: 0;
+  border-radius: 3px;
+  background: none;
+  color: var(--fg);
+  font: inherit;
+  font-size: 0.82rem;
+  white-space: pre;
+  cursor: pointer;
+}
+
+/* `selected` is what the document is fixed to; `active` is where the keyboard
+   is. They start together and part on the first arrow key. */
+#page-menu .page-item.selected {
+  color: var(--accent);
+  font-weight: 600;
+}
+
+#page-menu .page-item.active,
+#page-menu .page-item:hover {
+  background: var(--selected);
 }
 
 /* ── The frame: panels and grid ──────────────────────────────────────── */
diff --git a/web/tests/dispatch.test.js b/web/tests/dispatch.test.js
index 3dcd1b0b..3246634e 100644
--- a/web/tests/dispatch.test.js
+++ b/web/tests/dispatch.test.js
@@ -4,7 +4,15 @@
 import { test } from "node:test";
 import assert from "node:assert/strict";
 
-import { cellAction, cellAttrs, gridPlan, indexAction, pageEntries } from "../shell.js";
+import {
+  cellAction,
+  cellAttrs,
+  gridPlan,
+  indexAction,
+  menuKey,
+  pageEntries,
+  pageItemAction,
+} from "../shell.js";
 
 const cell = (text, extra = {}) => ({
   text,
@@ -85,7 +93,13 @@ test("an index surface reads its row back, and nothing else", () => {
 // moved it in silence. The bar is what says which quarter this is.
 test("the page bar names the category and the item it is fixed to", () => {
   const [entry] = pageEntries({
-    page_coords: [{ cat: "Quarter", selected: "Q2", cycle: { at: 1, count: 4 } }],
+    page_coords: [
+      {
+        cat: "Quarter",
+        selected: "Q2",
+        cycle: { at: 1, count: 4, items: ["Q1", "Q2", "Q3", "Q4"] },
+      },
+    ],
   });
   assert.equal(entry.text, "Quarter = Q2");
   assert.equal(entry.position, "2/4", "one-based, as a user counts");
@@ -106,3 +120,75 @@ test("a filter axis is shown, and is not offered as something to cycle", () => {
 test("no fixed categories is an empty bar, not a blank one", () => {
   assert.deepEqual(pageEntries({ page_coords: [] }), []);
 });
+
+// improvise-rw6l: the arrow and the menu follow `cycle`, which is the Page /
+// Filter split decided in Rust — never "is there an entry on the bar".
+test("a page axis offers its items, and a filter axis offers none", () => {
+  const [page, filter] = pageEntries({
+    page_coords: [
+      {
+        cat: "Quarter",
+        selected: "Q3",
+        cycle: { at: 2, count: 3, items: ["Q1", "Q2", "Q3"] },
+      },
+      { cat: "Region", selected: "North", cycle: null },
+    ],
+  });
+  assert.deepEqual(page.items, [
+    { item: 0, label: "Q1", selected: false },
+    { item: 1, label: "Q2", selected: false },
+    { item: 2, label: "Q3", selected: true },
+  ]);
+  assert.equal(page.index, 0, "the axis names itself by its place on the bar");
+  assert.deepEqual(
+    filter.items,
+    [],
+    "a Filter axis constrains without being navigable: nothing to pick",
+  );
+  assert.equal(filter.index, 1);
+});
+
+// One meaning across two surfaces (improvise-b8ot): a chip carries an axis, a
+// menu row carries an axis AND an item. Nothing infers from structure.
+test("a chip opens the menu and a menu row picks, one attribute apart", () => {
+  assert.deepEqual(pageItemAction({ axis: "1" }), { kind: "open", axis: 1 });
+  assert.deepEqual(pageItemAction({ axis: "1", item: "0" }), {
+    kind: "pick",
+    axis: 1,
+    item: 0,
+  });
+  assert.equal(pageItemAction({}), null, "the gaps between the chips are not a chip");
+  assert.equal(pageItemAction({ axis: "" }), null);
+  assert.equal(pageItemAction({ axis: "-1" }), null);
+  assert.deepEqual(
+    pageItemAction({ axis: "0", item: "" }),
+    { kind: "open", axis: 0 },
+    "an empty item attribute is NO item, where Number('') would say 0",
+  );
+});
+
+// The menu owns the keyboard while it is open, so every key is answered and
+// nothing falls through to the grid it is anchored to.
+test("the menu answers its own keys and swallows the rest", () => {
+  const at = (active) => ({ count: 3, active });
+  assert.deepEqual(menuKey("ArrowDown", at(0)), { kind: "move", item: 1 });
+  assert.deepEqual(menuKey("j", at(2)), { kind: "move", item: 0 }, "and it wraps");
+  assert.deepEqual(menuKey("ArrowUp", at(0)), { kind: "move", item: 2 }, "both ways");
+  assert.deepEqual(menuKey("k", at(1)), { kind: "move", item: 0 });
+  assert.deepEqual(menuKey("Home", at(2)), { kind: "move", item: 0 });
+  assert.deepEqual(menuKey("End", at(0)), { kind: "move", item: 2 });
+  assert.deepEqual(menuKey("Enter", at(1)), { kind: "pick", item: 1 });
+  assert.deepEqual(menuKey(" ", at(1)), { kind: "pick", item: 1 });
+  assert.deepEqual(menuKey("Escape", at(1)), { kind: "close" });
+  assert.deepEqual(menuKey("Tab", at(1)), { kind: "close" });
+  assert.deepEqual(
+    menuKey("h", at(1)),
+    { kind: "ignore" },
+    "a grid key is SWALLOWED, not passed on — the grid is behind the menu",
+  );
+  assert.deepEqual(
+    menuKey("ArrowDown", { count: 0, active: 0 }),
+    { kind: "close" },
+    "a menu with nothing in it cannot be moved around",
+  );
+});