Date: 2026-03-02
Status: Closure evidence artifact
Scope: G-A-7, G-A-8, G-A-9, G-A-11 from ../research/2026-03-02_ux_integration_research.md.

Related:

• ../design/accessibility_baseline_checklist.md
• ../implementation_strategy/subsystem_accessibility/SUBSYSTEM_ACCESSIBILITY.md
• ../implementation_strategy/subsystem_focus/focus_and_region_navigation_spec.md

Preference-plumbed reduced-motion behavior is staged (not yet fully preference-driven).

1. Motion-critical behavior remains under explicit command authority (GraphTogglePhysics) and does not silently switch command semantics.
2. Focus and return-path transitions remain deterministic and non-animated by contract (focus/workbench specs), preserving non-pointer accessibility behavior.
3. Reduced-motion remains tracked as a first-class closure lane in canonical accessibility artifacts until preference plumbing is completed.

• Verify deterministic physics control path remains explicit and stable through command surfaces (ActionRegistry path for GraphTogglePhysics) and does not introduce hidden motion-only routing.

Run this PowerShell snippet from repository root to compute WCAG contrast ratios for explicit-color radial command-surface pairs:

function Convert-ToLinear([double]$c){ $v=$c/255.0; if($v -le 0.04045){ return $v/12.92 } else { return [math]::Pow((($v+0.055)/1.055),2.4) } }
function Get-Luminance($rgb){ return 0.2126*(Convert-ToLinear $rgb[0]) + 0.7152*(Convert-ToLinear $rgb[1]) + 0.0722*(Convert-ToLinear $rgb[2]) }
function Get-Contrast($fg,$bg){ $l1=Get-Luminance $fg; $l2=Get-Luminance $bg; if($l2 -gt $l1){ $tmp=$l1; $l1=$l2; $l2=$tmp }; return ($l1+0.05)/($l2+0.05) }

• Disabled radial command text color was remediated in render/radial_menu.rs and is now gated by an automated contrast regression test (radial_disabled_text_contrast_meets_wcag_minimum_for_text).
• Post-fix measured ratio for disabled text on disabled button is approximately 6.05:1 (>= 4.5:1) and therefore now passes WCAG 1.4.3 normal text threshold.

• Historical note: disabled-state text contrast in radial surface was previously 3.21:1; this was remediated on 2026-03-04 (see §3.4).

Verify explicit control-size constants in render/radial_menu.rs:

• COMMAND_BUTTON_RADIUS = 22.0 → diameter 44px
• WCAG 2.5.8 minimum target size: 24px

• Toolbar/menu controls without explicit size constants remain pending measurement in a follow-on target-size sweep.

Run:

• cargo test cycle_focus_region_intent_cycles_graph_node_tool_regions -- --nocapture
• cargo test close_settings_tool_pane_restores_previous_graph_focus_via_orchestration -- --nocapture
• cargo test cycle_focus_region_success_does_not_emit_ux_navigation_violation_channel -- --nocapture

All listed tests pass, validating deterministic non-pointer escape/return paths and no-trap navigation behavior in host UI focus routing.

Additional regression coverage now verifies global undo shortcut modal isolation across multiple floating command surfaces:

• global_shortcut_undo_is_consumed_when_modal_is_active (radial)
• global_shortcut_undo_is_consumed_when_command_palette_modal_is_active
• global_shortcut_undo_is_consumed_when_help_panel_modal_is_active
• global_shortcut_undo_is_consumed_when_clear_confirm_modal_is_active (clear-data confirm dialog path)

This path now routes through a unified modal-surface dispatch gate shared with UI overlay state, so clear-confirm dialog activation participates in the same no-trap shortcut isolation contract as command overlays.

All pass and provide evidence that active modal overlays consume non-modal global shortcut handling instead of trapping focus/dispatch in ambiguous paths.

Input-layer shortcut capture behavior is now explicitly covered with tests in input/mod.rs:

• collect_actions_suppresses_f6_when_keyboard_is_captured
• collect_actions_allows_f6_when_keyboard_is_not_captured
• collect_actions_keeps_f9_global_when_keyboard_is_captured
• collect_actions_suppresses_f1_when_keyboard_is_captured
• collect_actions_suppresses_f2_when_keyboard_is_captured
• collect_actions_suppresses_f3_when_keyboard_is_captured

This validates that focus-cycling shortcuts are suppressed while text-edit surfaces own keyboard focus, while designated global safety/escape controls remain available.

Keyboard operability coverage now explicitly includes workbar command-surface toggles through input-layer key-detection tests:

• collect_actions_maps_f1_to_help_panel_toggle_when_not_captured
• collect_actions_maps_f2_to_command_palette_toggle_when_not_captured
• collect_actions_maps_f3_to_radial_menu_toggle_when_not_captured

Together with existing intent-application tests (test_toggle_help_panel_action, test_toggle_command_palette_action), this provides direct evidence that core workbar command surfaces are keyboard-invokable.

Omnibar submit-dispatch behavior now has explicit helper-level coverage in toolbar_location_panel.rs:

• submit_dispatch_triggers_for_focused_enter
• submit_dispatch_triggers_for_queued_submit
• submit_dispatch_ignores_enter_after_focus_loss
• submit_dispatch_triggers_for_queued_submit_after_focus_loss
• submit_dispatch_does_not_trigger_without_enter_or_queue

This confirms keyboard-submit semantics remain deterministic (focused Enter and queued-submit paths dispatch; passive Enter after focus loss does not), reducing no-trap/accidental-submit ambiguity for omnibar workflows.

Clear-data destructive confirmation timing now has explicit unit coverage in dialog_panels.rs:

• clear_data_confirm_is_not_armed_without_deadline
• clear_data_confirm_is_armed_until_deadline_inclusive
• clear_data_confirm_expires_after_deadline_passes
• next_clear_data_confirm_deadline_uses_expected_window

This validates deterministic two-step confirmation semantics (arm on first activation, execute only while armed, require re-arm after expiry), which contributes direct error-prevention evidence for destructive dialog actions.

Instructional copy invariants now have explicit regression coverage:

• clear_data_confirm_warning_text_includes_instruction_and_timing
• clear_data_confirm_success_text_describes_completed_action
• location_input_hint_text_provides_search_and_address_instruction

These tests preserve baseline instruction clarity for destructive dialogs and the omnibar primary text-entry affordance, providing concrete partial evidence for WCAG 3.3.2 mapping while broader multi-surface validation remains pending.

Input-layer shortcut suppression under text capture now explicitly includes character-key shortcuts:

• collect_actions_suppresses_character_shortcut_n_when_keyboard_is_captured
• collect_actions_suppresses_character_shortcut_t_when_keyboard_is_captured
• collect_actions_suppresses_character_shortcut_questionmark_when_keyboard_is_captured

This extends keyboard-capture evidence beyond function keys, showing that single-character command shortcuts are not triggered while text-entry controls own keyboard focus.

Dialog clear-confirm input processing now has explicit action-classification regressions:

• clear_data_confirm_action_arms_when_no_deadline_is_present
• clear_data_confirm_action_executes_when_deadline_is_active
• clear_data_confirm_action_rearms_when_deadline_expired

Together with existing omnibar submit-dispatch tests (submit_dispatch_*), this provides concrete evidence that input-side effects are deterministic and state-dependent rather than surprising implicit context changes.

Command-surface disabled precondition guidance now has explicit regressions:

• all_disabled_actions_expose_textual_precondition_reason_in_default_context
• disabled_action_reasons_use_actionable_text_not_color_cues

These tests provide partial evidence that disabled-state communication is conveyed via textual/actionable guidance rather than relying on color-only cues, supporting WCAG 1.3.3 and 1.4.1 mapping in command surfaces.

Action label purpose clarity now has targeted regression coverage in action_registry.rs:

• test_representative_action_labels_convey_purpose_in_context

The test verifies representative Open/Copy/Connect/Delete/Save/Restore command labels contain explicit purpose terms, providing partial evidence for WCAG 2.4.4 mapping on command surfaces.

Clipboard status/toast wording now has targeted regression coverage in gui_orchestration:

• clipboard_success_status_text_is_deterministic_per_copy_kind
• clipboard_status_messages_describe_outcomes_explicitly
• clipboard_status_success_messages_identify_copied_subject

This coverage locks deterministic success mapping (URL/title) and explicit success/warning/error/unavailable wording, providing partial evidence for WCAG 4.1.3 mapping on command-surface status feedback while screen-reader announcement verification remains pending.

Clipboard failure wording now has targeted regression coverage in gui_orchestration:

• clipboard_failure_message_prefix_is_stable_and_identifiable
• clipboard_missing_node_failure_message_is_explicit

This coverage preserves explicit error identification semantics by enforcing a stable failure prefix and concrete missing-node failure wording in clipboard command feedback, providing partial evidence for WCAG 3.3.1 mapping while broader multi-surface error-state audit remains pending.

Clipboard missing-node failure guidance now has targeted regression coverage in gui_orchestration:

• clipboard_missing_node_failure_message_includes_recovery_suggestion

The missing-node failure path now includes explicit recovery guidance (select a node and try again), providing partial evidence for WCAG 3.3.3 mapping in command-surface error feedback while broader multi-surface suggestion quality validation remains pending.

• Reduced-motion behavior implemented or explicitly staged with guardrails.
• Contrast/target-size audit artifacts committed.
• Keyboard trap mitigation validated.