fix(agent): give wrapped canvas rows headroom for frame labels

The canvas paints each top-level frame's name ABOVE the frame at a fixed
screen-space offset that does not scale with zoom (canvas_frame_labels.rs
puts the baseline at sy-18 and the label box top at sy-32). A row gap
equal to the column gap therefore cannot work: at the zoom where a
wrapped canvas is actually viewed, the gap is narrower than the label
itself, so every second-row label sat pressed against the bottom edge of
the row above.

For the 3x2 deck the numbers are exact: 6000 doc px of content framed by
zoom_to_fit into a 944-1424 px canvas region lands at zoom 0.14-0.22,
where a 120 doc px row gap is only 16-26 screen px against a label that
always occupies 32. It overlapped at every realistic window size.

Add 240 doc px of headroom to the VERTICAL step only — 33-52 screen px at
that zoom, clearing the label with room to spare. Columns are untouched:
labels are left-aligned to their own frame and never reach sideways. The
slide-deck template takes the same allowance on top of its own gap, so
both the generated canvas and the shipped template space rows the same
way.

Claude-Session: https://claude.ai/code/session_01FqKQqNj8exYwopGDpYUU7x

crates/op-editor-core/assets/scene_templates/slide-deck.op

@@ -1232,7 +1232,7 @@
         }
       ],
       "x": 0,
-      "y": 1200
+      "y": 1440
     },
     {
       "type": "frame",
@@ -1575,7 +1575,7 @@
         }
       ],
       "x": 2040,
-      "y": 1200
+      "y": 1440
     },
     {
       "type": "frame",
@@ -1664,7 +1664,7 @@
         }
       ],
       "x": 4080,
-      "y": 1200
+      "y": 1440
     }
   ]
 }

crates/op-orchestrator/src/scaffold.rs

@@ -462,6 +462,24 @@ pub(crate) const SCREEN_GROUP_GAP: f64 = 80.0;
 /// - 390 mobile       → 17 per row, i.e. unchanged for any realistic fan-out
 const MAX_ROW_WIDTH: f64 = 8200.0;
 
+/// Extra vertical clearance between rows, on top of `SCREEN_GROUP_GAP`.
+///
+/// The canvas paints each top-level frame's name ABOVE the frame at a FIXED
+/// screen-space offset that does not scale with zoom
+/// (`canvas_frame_labels.rs`: baseline at `sy - 18`, label box top at
+/// `sy - 32`). So a row gap equal to the column gap makes every second-row
+/// label collide with the bottom edge of the row above — 80 doc px is only
+/// ~11-17 screen px at the zoom where a wrapped canvas is actually viewed,
+/// well inside the label's fixed 32 px footprint (user report 2026-08-02).
+///
+/// 240 doc px is the allowance: fitting a wrapped multi-row canvas to the
+/// screen lands around zoom 0.14-0.22 (a 3-wide 1920 deck is 6000 doc px
+/// across, framed by `zoom_to_fit` into a 944-1424 px canvas region), where
+/// 240 doc px reads as 33-52 screen px — the label's 32 px plus breathing
+/// room. Horizontal spacing is untouched: labels are left-aligned to their
+/// own frame and never reach sideways.
+const ROW_LABEL_HEADROOM: f64 = 240.0;
+
 /// How many screen roots fit in one row at `board_width`.
 ///
 /// Always ≥ 1: a board wider than the whole row budget still has to go
@@ -527,14 +545,16 @@ pub(crate) fn build_screen_group_scaffold(
 
     // Roots wrap into rows instead of one endless strip (see `MAX_ROW_WIDTH`).
     // `y` is the top of the FIRST row; each later row steps down by a board
-    // height plus the same gutter the columns use. The row step reads the
+    // height, the same gutter the columns use, and the label headroom the
+    // columns do NOT need (see `ROW_LABEL_HEADROOM`). The row step reads the
     // height the scaffold actually builds with — a plan's `height: 0` means
     // "size me from content", and `build_root_frame_node` presets that to the
     // device-class artboard, so both must ask `resolved_root_height` or the
     // rows would overlap by exactly the amount that preset adds.
     let per_row = boards_per_row(rf.width);
     let column_step = rf.width + SCREEN_GROUP_GAP;
-    let row_step = resolved_root_height(rf.height, rf.width) + SCREEN_GROUP_GAP;
+    let row_step =
+        resolved_root_height(rf.height, rf.width) + SCREEN_GROUP_GAP + ROW_LABEL_HEADROOM;
 
     for (index, group) in groups.iter().enumerate() {
         // Placeholder id: `{root_frame.id}-{screen}` — mirrors the deleted

crates/op-orchestrator/src/scaffold_tests.rs

@@ -520,10 +520,13 @@ fn wide_screen_group_boards_wrap_onto_a_second_row() {
     );
 
     // The fifth board is the wrap: back to the left edge, one board height
-    // plus a gutter further down.
+    // plus a gutter plus the frame-name label's headroom further down.
     assert_eq!(
         positions[4],
-        (start_x, start_y + 1080.0 + SCREEN_GROUP_GAP),
+        (
+            start_x,
+            start_y + 1080.0 + SCREEN_GROUP_GAP + ROW_LABEL_HEADROOM
+        ),
         "board 5 starts a new row: {positions:?}"
     );
     assert_eq!(
@@ -583,7 +586,7 @@ fn content_sized_boards_still_step_down_by_the_preset_height() {
     let positions = root_positions(nodes);
     assert_eq!(
         positions[4],
-        (0.0, 900.0 + SCREEN_GROUP_GAP),
+        (0.0, 900.0 + SCREEN_GROUP_GAP + ROW_LABEL_HEADROOM),
         "the desktop preset height drives the row step: {positions:?}"
     );
 }

templates/step0/_generators/tpl_slides.py

@@ -231,10 +231,23 @@ def closing():
     return s
 
 
-# Gap between boards on the canvas, used on both axes. Without explicit
-# positions every frame defaults to the origin and the deck stacks into one
-# visible slide — the other five are underneath it (reported 2026-08-02).
-BOARD_GAP = 120
+# Horizontal gap between boards. Without explicit positions every frame
+# defaults to the origin and the deck stacks into one visible slide — the
+# other five are underneath it (reported 2026-08-02).
+BOARD_GAP_X = 120
+
+# Vertical gap. Larger than the horizontal one because the canvas paints each
+# frame's name ABOVE the frame at a fixed screen-space offset that does not
+# scale with zoom (canvas_frame_labels.rs: baseline at sy-18, label box top at
+# sy-32). At the zoom where a 3x2 deck is framed to the screen (~0.14-0.22 —
+# the deck is 6000 doc px across, fitted into a 944-1424 px canvas region), a
+# 120 doc px row gap is only ~16-26 screen px, i.e. NARROWER than the label's
+# fixed 32 px footprint, so every second-row label sat on the bottom edge of
+# the row above (reported 2026-08-02). The extra 240 doc px reads as 33-52
+# screen px there, clearing the label with room to spare. Mirrors
+# `scaffold.rs::ROW_LABEL_HEADROOM`, which fixes the generated multi-screen
+# canvas the same way.
+BOARD_GAP_Y = BOARD_GAP_X + 240
 
 # Boards per row. Six 1920px slides in one strip are ~12,000px across, which
 # can only be read by panning; wrapping them into a 3x2 grid matches the
@@ -247,8 +260,8 @@ BOARDS_PER_ROW = 3
 def build():
     boards = [cover(), agenda(), points(), metrics(), chart(), closing()]
     for index, board in enumerate(boards):
-        board["x"] = (index % BOARDS_PER_ROW) * (W + BOARD_GAP)
-        board["y"] = (index // BOARDS_PER_ROW) * (H + BOARD_GAP)
+        board["x"] = (index % BOARDS_PER_ROW) * (W + BOARD_GAP_X)
+        board["y"] = (index // BOARDS_PER_ROW) * (H + BOARD_GAP_Y)
     return boards