Control the Linux desktop GUI using xdotool, wmctrl, and dogtail. Use when you need to interact with non-browser applications, simulate mouse/keyboard input, manage windows, or inspect the UI hierarchy of applications on X11/GNOME. Supports: (1) Clicking/typing in apps, (2) Resizing/moving windows, (3) Extracting text-based UI trees from apps (A11y), (4) Taking screenshots for visual analysis.
Security Analysis
high confidenceThe skill's files and instructions are coherent with its stated purpose (desktop automation and UI inspection) and do not request unrelated credentials or network access, but it can control and restart user applications and capture screenshots so treat it as powerful and potentially privacy-sensitive.
The name/description match the included scripts and README: scripts call xdotool/wmctrl/scrot and the Python file uses dogtail to dump the AT-SPI accessibility tree. Nothing in the files requests unrelated credentials, config paths, or external services; the capabilities (click/type/activate/list/screenshot/inspect) are exactly what the skill claims.
Instructions are focused on window listing, activating, simulating input, taking screenshots, and reading accessibility trees. However, the README explicitly recommends killing and relaunching Electron apps with a --force-renderer-accessibility flag using pkill and nohup, which will stop and restart user processes (affecting user state). The scripts do not perform network calls or read arbitrary local files, but taking screenshots and dumping UI text can expose sensitive on-screen data — the skill does not include safeguards or explicit consent steps for that.
No install spec is provided (instruction-only with two helper scripts). This minimizes supply-chain risk. The skill assumes external binaries (xdotool, wmctrl, dogtail, scrot) are present on the system but does not attempt to install them.
No environment variables, credentials, or config paths are requested. The requested capabilities (controlling GUI and using AT-SPI) are proportional to the stated purpose. Note: AT-SPI/dogtail access implicitly requires that accessibility services are enabled and that the agent has permission to interact with the desktop session.
always:false and no special persistence or modification of other skills or system-wide agent settings. The skill can be invoked by the agent (default model invocation allowed), which is normal; because it controls the GUI, autonomous invocation could have impact, but there is no special permanent privilege requested.
Guidance
This skill appears to do exactly what it says: automate and inspect the Linux GUI using xdotool/wmctrl/dogtail/scrot. Before installing, consider: (1) It can simulate keyboard/mouse and take screenshots — that can expose sensitive information on-screen. (2) The README suggests using pkill/nohup to restart apps with flags; that will terminate and restart your applications and may change their behavior. (3) The skill assumes the listed tools are installed but does not install them. (4) If you will allow the agent to invoke this skill autonomously, limit its permissions or run it in a controlled environment (VM or disposable session) to reduce risk of unintended input or data capture. If you need network-safe behavior, ensure your agent or environment prevents automatic upload of screenshots or dump output to external endpoints.
Latest Release
v1.0.0
linux-gui-control 1.0.0 initial release - Automate Linux desktop interactions using xdotool, wmctrl, and dogtail. - Supports simulating mouse/keyboard input, window management, and text-based UI inspection on X11/GNOME. - Offers scripts for automating clicks, typing, window focus, and window movement. - Enables extraction of application UI hierarchies for accessible apps and provides screenshot support for visual tasks.
Popular Skills
Published by @dreamtraveler13 on ClawHub
