This is the tree as it stands now: a leader-driven live walk grew it from 22 to 26 screens. Below is the current After map (a node is a screen, an edge is an action), then the findings organized into problems / improvements and the custom tools the walk proved we need to build.
A node is a screen (click it for the device screenshot); an edge is an action (click it for the action + the screen it lands on). Green = grown live this survey; dashed amber = a fragmentation duplicate (same screen split by dynamic content — problem IB-1).
Select a node (screen) or edge (action).
Structural fixes to the framework. The two MAJOR items gate most further growth.
Each is a real widget the walk reached but the current tool-set could not actuate or read. The screenshot shows the widget; the gap is concrete; the tool is the deliverable.
set_weightfound #004set_clock_facefound #006read_chart_valuefound #012The tree is now 26 screens and the walk works. To go deeper, build the 2 MAJOR fixes (general de-fragmentation + open the feature-vocab) and the 3 custom tools (set_weight, clock-face, chart-read). Then the same walk reaches the screens it was blocked on today.
Source: committed menu-tree fold (post live survey) + stage snapshots; runlog frontier-RUNLOG.md. Screenshots are real device captures (chart screen has no fixture — drawn canvas).