What problem does it solve? Operating GUI applications like LibreOffice Writer, Calc, Impress, the Thunar file manager, and terminal emulators requires visual interaction that structured file tools cannot perform. This Skill gives the agent an observe-act-verify control loop over the persistent XFCE desktop so it can launch apps, focus windows, send keystrokes, paste text, edit spreadsheet cells live, and verify results from screenshots. ## Core Features & Use Cases - Desktop Control Script: The desktopctl.sh helper wraps xdotool to launch apps, focus windows, click, type, paste, scroll, drag, and capture screenshots against the persistent agent-zero-desktop X display. - Live Calc Editing: The calc-set-cell command opens an XLSX in the visible LibreOffice Calc session via UNO, edits a cell, saves, and verifies the value on disk so users see updates in real time. - Structured State Observation: observe --json --screenshot returns window titles, active window, and a fresh screenshot for verification after every GUI action. - Use Case: A user asks to update a budget spreadsheet and confirm the change visually. The agent launches Calc, sets the cell through the UNO bridge, saves the workbook, and captures a final screenshot as proof. ## Quick Start Ask the agent to open LibreOffice Calc on the Linux desktop, set cell B2 in your workbook to a new value, save it, and show a screenshot confirming the change.
