2.8 KiB
Capella scenario occurrence editing observation
Observed 2026-09-13 on the installed Capella 7.1.0 Windows distribution, existing local IncidentScenarioBenchmark comparison workspace, [OAS] OperationalCapability 1. This is direct Computer Use observation of this fixture/window only.
The app initially remained at its workspace launcher. After Launch, the launcher disappeared; activation and one window-list request timed out during startup. A fresh window list and activation then reached the existing scenario. This supersedes the earlier assumption that capture remained unavailable, not its recorded historical failures.
At a 1010 by 761 captured window, the diagram has two Operational Activity lifelines and an asynchronous Report incident. Single click on the message selects the Sequence Message occurrence and exposes Name, Summary, Invoked Interaction, Exchanged Items and kind through the Properties pane. Name and Invoked Interaction both display Report incident in this fixture but are distinct fields. The bottom pane initially exposes only Name and Summary; remaining controls require scrolling.
Double-click on the message label opened a Properties dialog with Name and Summary, linked Invoked Interaction, Exchanged Items, Sequence Message Kind, and Exchange Context. The kind choices were visibly disabled with ASYNCHRONOUS_CALL selected. No text was entered; Cancel returned to the selected occurrence. This does not prove editing/commit semantics or why kinds are disabled.
The preserved image pair shows a reversible selection-induced layout shift without changing the outer window: selected message uses two toolbar rows and its horizontal line is at approximately y334; background/diagram selection removes a toolbar row and the same line moves to approximately y304. Headers and viewport move with it. The two activity names also wrap in the diagram while the pinned header labels are ellipsized. These are observed pixel/layout facts, not authored coordinate changes or a benchmark of all Capella layouts.
Product implications: preserve direct, discoverable occurrence editing and explicit separation from referenced interaction identity; keep the canvas bounds stable when selection-specific actions appear. Compare Studio using its real editable inspector and actual selection route. Avoid making additional panels a prerequisite for basic edits, while keeping independently selected model records intact during view changes. Existing Studio mounted layout/selection tests support these intentions; current native Studio validation remains required.
Evidence: message-selected.jpg (after dialog Cancel), diagram-selected.jpg (after background click), selection-metadata.json. No deletion, Name edit, Save or model mutation was attempted. Screenshots were preserved as audit artifacts, not re-read to drive inputs.