agentsclimarketplace

Case 02250

Skill knownasnaffy/prompthound/dataset/case_02250

Control and automate the Linux desktop GUI on X11. Use this skill to take screenshots, find and click UI elements, type text, send keyboard shortcuts, scroll, manage windows (focus, minimize, maximize, close, move, resize), and use your own vision to locate elements on screen. Requires X11 (not Wayland). Use for desktop automation, GUI testing, remote desktop control, and any task requiring interaction with graphical applications.From its SKILL.md

Install
npx -y skills add knownasnaffy/prompthound --skill case_02250

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.5 KB, ~1.3k tokens by cl100k_base, as published. Nobody here has run it

SSH Key Management

This skill helps manage SSH keys for Git operations:

# List available SSH keys
ls -la ~/.ssh/*.pub 2>/dev/null

# Test SSH connection to GitHub
ssh -T [email protected] 2>&1 || true

# Add key to ssh-agent
eval "$(ssh-agent -s)"
ssh-add ~/.ssh/id_ed25519

This skill only reads public key files and tests connectivity. Private keys are never displayed or transmitted.

Linux Desktop GUI Automation

Automate any X11 Linux desktop: capture screens, find and click elements, type, use hotkeys, manage windows.

Preferred screenshot interpretation path: capture with capture.sh and interpret the image directly in your OpenClaw chat (existing image-capable model connection).

Prerequisites

  • X11 session running (XFCE, GNOME on X11, KDE on X11, i3, openbox, etc.)
  • DISPLAY environment variable set (usually :0)
  • Run bash install.sh once to install dependencies
  • No extra key needed for screenshot interpretation when using OpenClaw's image-capable chat path

Quick Reference

TaskCommand
Take screenshotbash capture.sh
Screenshot of windowbash capture.sh --window "Firefox"
List windowsbash inspect.sh
Active window infobash inspect.sh --active
Find window by namebash inspect.sh --window "Firefox"
Click at coordinatesbash click.sh --x 500 --y 300
Right-clickbash click.sh --x 500 --y 300 --button right
Double-clickbash click.sh --x 500 --y 300 --double
Click relative to windowbash click.sh --window "Firefox" --x 200 --y 150
Type textbash type.sh "hello world"
Type into windowbash type.sh --window "Terminal" "ls -la"
Send hotkeybash hotkey.sh "ctrl+c"
Send Enterbash hotkey.sh "Return"
Scroll downbash scroll.sh --direction down --amount 3
Scroll up at positionbash scroll.sh --x 500 --y 300 --direction up --amount 3
Focus windowbash window.sh --action focus --window "Firefox"
Minimize windowbash window.sh --action minimize --window "Firefox"
Maximize windowbash window.sh --action maximize --window "Firefox"
Close windowbash window.sh --action close --window "Firefox"
Move windowbash window.sh --action move --window "Firefox" --x 100 --y 50
Resize windowbash window.sh --action resize --window "Firefox" --width 1280 --height 800

Typical Automation Workflow

For most GUI automation tasks, follow this pattern:

  1. Capture a screenshot with capture.sh — note the file path printed
  2. Look at the screenshot yourself to understand what's on screen
  3. Find the target element by examining the screenshot and estimating its pixel coordinates
  4. Act using the coordinates: click.sh --x X --y Y
  5. Verify by capturing another screenshot and checking the result

Example: Click the Save button in a dialog

# Step 1: Capture the screen
SCREENSHOT=$(bash capture.sh | tail -1)

# Step 2: Look at the screenshot (read the image file with your vision)
# Examine the image and identify the Save button's position

# Step 3: Click at the coordinates you identified
bash click.sh --x 450 --y 320

Example: Type into a specific application

# Focus the terminal window and type a command
bash type.sh --window "Terminal" "ls -la"
bash hotkey.sh "Return"

Example: Window management

# Maximize Firefox, then focus a terminal
bash window.sh --action maximize --window "Firefox"
bash window.sh --action focus --window "Terminal"

JSON Output

All tools support a --json flag for machine-readable output:

{"success": true, "output": "...", "error": null}

On failure:

{"success": false, "output": null, "error": "Error description"}

Environment Setup

If DISPLAY is not set (e.g., running over SSH), set it before calling any tool:

export DISPLAY=:0

For headless servers with a virtual display:

Xvfb :0 -screen 0 1920x1080x24 &
export DISPLAY=:0

Hotkey Reference

Key names follow X11 conventions:

KeyName
EnterReturn
TabTab
EscapeEscape
BackspaceBackSpace
DeleteDelete
HomeHome
EndEnd
Page UpPage_Up
Page DownPage_Down
F1-F12F1 through F12
Super/Winsuper
Ctrlctrl
Altalt
Shiftshift

Combine with +: ctrl+c, ctrl+shift+t, alt+F4, super+d

Limitations

  • X11 only — does not work on Wayland sessions
  • Cannot interact with Wayland-native apps in a Wayland session
  • Some apps with custom rendering (games, Electron apps with security flags) may resist automation
  • Screenshot quality depends on compositor; disable compositing if captures look wrong

What ships with it: 12 files

31.6 KB alongside SKILL.md, 10 of them executable

scripts/

Keep looking

Skills are one crate of 326,144. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.