PyAutoGUI Automation Skill

Cross-platform mouse/keyboard automation for Windows, Linux, and macOS.

Features

• Mouse control: move, click, drag, scroll
• Keyboard control: key press, hotkeys, type text
• Screen operations: screenshot, mouse position, screen size
• Image utilities: image metadata (size/format/file size), crop images
• Screen overlay: draw temporary markers to validate coordinates
• Draw on images: draw persistent markers into an image and save
• Image locating: template matching and OCR-based text locating
• Cleanup: remove generated screenshots/marked files to free disk space

Activation

Activate when the user asks to do things like:

• "Click a position on the screen"
• "Move the mouse to (x, y)"
• "Type text / press keys"
• "Take a screenshot"
• "Run repetitive UI automation"
• "Get the current mouse position"
• "Get image size / image info"
• "Crop an image"
• "Draw a marker on the screen"
• "Draw a marker on an image"
• "Locate an element by template"
• "Locate text on the screen (OCR)"
• "Clean up screenshots / temporary files"

Usage

Install dependencies

# Mouse/keyboard automation
pip3 install pyautogui

# Image utilities
pip3 install Pillow

Screen info

# Screen size
python3 scripts/keyboard_mouse.py screen_size

# Mouse position
python3 scripts/keyboard_mouse.py mouse_position

Mouse actions

# Move mouse to (x, y)
python3 scripts/keyboard_mouse.py mouse_move 500 300
python3 scripts/keyboard_mouse.py mouse_move 500 300 --duration 1.0

# Mouse click (left/right/middle)
python3 scripts/keyboard_mouse.py mouse_click left
python3 scripts/keyboard_mouse.py mouse_click right
python3 scripts/keyboard_mouse.py mouse_click middle --clicks 2

# Click at a specific location
python3 scripts/keyboard_mouse.py mouse_click_at 500 300 left
python3 scripts/keyboard_mouse.py mouse_click_at 500 300 right --clicks 2

# Double click
python3 scripts/keyboard_mouse.py mouse_double_click 500 300

# Drag
python3 scripts/keyboard_mouse.py mouse_drag 500 300 800 600
python3 scripts/keyboard_mouse.py mouse_drag 500 300 800 600 --duration 2.0

# Scroll (positive = up, negative = down)
python3 scripts/keyboard_mouse.py mouse_scroll 5
python3 scripts/keyboard_mouse.py mouse_scroll -3

Keyboard actions

# Single key
python3 scripts/keyboard_mouse.py key_press enter
python3 scripts/keyboard_mouse.py key_press escape
python3 scripts/keyboard_mouse.py key_press tab
python3 scripts/keyboard_mouse.py key_press space

# Hotkeys
python3 scripts/keyboard_mouse.py key_hotkey ctrl c
python3 scripts/keyboard_mouse.py key_hotkey ctrl v
python3 scripts/keyboard_mouse.py key_hotkey win r
python3 scripts/keyboard_mouse.py key_hotkey alt tab
python3 scripts/keyboard_mouse.py key_hotkey ctrl alt t

# Type text
python3 scripts/keyboard_mouse.py type_text "Hello World"
python3 scripts/keyboard_mouse.py type_text "你好世界" --interval 0.05

Screenshot

# Save a screenshot (primary screen)
python3 scripts/keyboard_mouse.py screenshot /tmp/screenshot.png

# Windows example
python scripts/keyboard_mouse.py screenshot "E:\\temp\\screenshot.png"

Screenshot notes:

• Supported formats: PNG (recommended), JPG, BMP, etc.
• Scope: primary monitor (in multi-monitor setups)

Region Screenshot

# Screenshot specific region (x1, y1, x2, y2)
python3 scripts/keyboard_mouse.py screenshot_region region.png 100 100 500 500

# Windows example - capture QQ chat window area
python scripts/keyboard_mouse.py screenshot_region qq_window.png 2800 300 3800 1200

Parameters:

• x1, y1

  : Top-left corner coordinates

• x2, y2

  : Bottom-right corner coordinates
• Order doesn't matter (automatically calculated)

Copy & Paste

# Copy text to clipboard
python3 scripts/keyboard_mouse.py copy "Text to copy"

# Paste from clipboard (Ctrl+V)
python3 scripts/keyboard_mouse.py paste

# Copy and paste in one command (fastest way to input text)
python3 scripts/keyboard_mouse.py copy_paste "Text to input directly"

Use cases:

• copy_paste

  is faster than

  type_text

  for long text
• Use

  copy_paste

  when you want to skip typing animation
• Use

  type_text

  when you need to simulate realistic typing

Common key names

• Letters:

  a

  b

  c

  ...
• Numbers:

  0

  1

  2

  ...
• Function keys:

  f1

  f2

  ...

  f12

• Modifiers:

  ctrl

  alt

  shift

  win

• Others:

  enter

  esc

  tab

  space

  backspace

  delete

  up

  down

  left

  right

Safety

⚠️ Important:

1. Make sure the target window is focused before executing actions
2. Be careful with system hotkeys to avoid unintended actions
3. Add delays when needed to give yourself time to interrupt
4. Moving the mouse to the top-left corner (0, 0) triggers PyAutoGUI failsafe

Cross-platform notes

• Windows: Full support; admin permission may be needed in some environments
• Linux: Requires X11; Wayland may not work
• macOS: Grant Accessibility permission to Terminal/Python in System Settings

Example scenarios

Open Calculator (Windows)

python3 scripts/keyboard_mouse.py key_hotkey win r
python3 scripts/keyboard_mouse.py type_text "calc"
python3 scripts/keyboard_mouse.py key_press enter

Auto-fill a form

python3 scripts/keyboard_mouse.py mouse_click_at 500 300 left
python3 scripts/keyboard_mouse.py type_text "example@email.com"
python3 scripts/keyboard_mouse.py key_press tab
python3 scripts/keyboard_mouse.py type_text "password123"

Batch clicking

python3 scripts/keyboard_mouse.py mouse_click_at 100 100 left
python3 scripts/keyboard_mouse.py mouse_click_at 200 200 left
python3 scripts/keyboard_mouse.py mouse_click_at 300 300 left

Included scripts

• scripts/keyboard_mouse.py

  - Mouse/keyboard control

• scripts/image_utils.py

  - Image utilities

• scripts/draw_overlay.py

  - Screen overlay markers

• scripts/draw_on_image.py

  - Draw markers on images

• scripts/image_finder.py

  - Image locating (template + OCR)

• scripts/cleanup.py

  - Cleanup tool

Image utilities

Image info

python3 scripts/image_utils.py info screenshot.png
python3 scripts/image_utils.py size photo.jpg

Crop image

python3 scripts/image_utils.py crop screenshot.png 100 100 500 500
python3 scripts/image_utils.py crop screenshot.png 100 100 500 500 -o output.png

Output example

$ python3 scripts/image_utils.py info screenshot.png
{
  "path": "screenshot.png",
  "filename": "screenshot.png",
  "size": {
    "width": 3840,
    "height": 2160
  },
  "format": "PNG",
  "mode": "RGB",
  "file_size_bytes": 2097152,
  "file_size_kb": 2048.0
}

Image fields

width

height

format

mode

file_size_bytes

file_size_kb

Coordinate system

Screen coordinates:

• Origin (0, 0) is the top-left corner
• X increases to the right
• Y increases downward

Crop coordinates:

• x1, y1

  : top-left corner of crop

• x2, y2

  : bottom-right corner of crop
• Cropped size = (x2 - x1) × (y2 - y1)

Example:

python3 scripts/image_utils.py crop screenshot.png 1520 880 1920 1080

Typical workflows

Analyze positions in a screenshot

python3 scripts/image_utils.py size screenshot.png
python3 scripts/image_utils.py crop screenshot.png 3440 1960 3840 2160 -o bottom_right.png

Batch image sizing

for img in *.png; do
    echo -n "$img: "
    python3 scripts/image_utils.py size "$img"
done

Capture a region of the screen

python3 scripts/keyboard_mouse.py screenshot full.png
python3 scripts/image_utils.py crop full.png 500 300 1000 800 -o region.png

Screen overlay markers

Draw temporary markers on the screen for coordinate verification. Useful for:

• Calibrating coordinates
• Confirming the real position of a button/element
• Debugging automation scripts

Draw a marker

python3 scripts/draw_overlay.py marker cross 500 300
python3 scripts/draw_overlay.py marker target 800 600 --duration 10
python3 scripts/draw_overlay.py marker circle 500 300 --color blue --text "Send button"
python3 scripts/draw_overlay.py marker arrow 1000 800 --direction down --color yellow
python3 scripts/draw_overlay.py marker square 600 400 --color green --size 40

Draw a rectangular area

python3 scripts/draw_overlay.py area 3028 276 3832 2098 --label "Window" --duration 8
python3 scripts/draw_overlay.py area 3744 2062 3832 2098 --label "Send button" --color red

Marker types

cross

circle

square

arrow

target

Colors

red

green

blue

yellow

cyan

magenta

white

orange

Coordinate calibration example

python3 scripts/keyboard_mouse.py screenshot screen.png
python3 scripts/draw_overlay.py marker target 3788 2080 --text "Send button" --duration 10
python3 scripts/draw_overlay.py marker target 3790 2090 --text "Send button (adjusted)" --duration 10
python3 scripts/keyboard_mouse.py mouse_click_at 3790 2090 left

Draw markers on images

Draw persistent markers into image files. Useful for:

• Annotating recognized positions on a screenshot
• Producing reference images
• Batch marking candidates for comparison
• Keeping calibration records

Draw a marker

python3 scripts/draw_on_image.py screenshot.png marker cross 500 300
python3 scripts/draw_on_image.py screenshot.png marker target 800 600 -o marked.png
python3 scripts/draw_on_image.py screenshot.png marker circle 500 300 --color red --text "Send button"
python3 scripts/draw_on_image.py screenshot.png marker arrow 1000 800 --direction down --color yellow
python3 scripts/draw_on_image.py screenshot.png marker point 600 400 --color green --size 10

Draw a rectangular area

python3 scripts/draw_on_image.py screenshot.png area 3028 276 3832 2098 --label "Window"
python3 scripts/draw_on_image.py screenshot.png area 3744 2062 3832 2098 -o button_marked.png --label "Send button"

Batch marking workflow

python3 scripts/keyboard_mouse.py screenshot screen.png
python3 scripts/draw_on_image.py screen.png marker target 3788 2080 --text "Send button" -o step1.png
python3 scripts/draw_on_image.py step1.png marker target 3790 2090 --text "Adjusted" -o step2.png
python3 scripts/draw_on_image.py step2.png marker circle 3000 1500 --text "Avatar area" -o final.png

Screen overlay vs drawing on image

Recommended coordinate calibration (cost-saving)

python3 scripts/keyboard_mouse.py screenshot screen.png
python3 scripts/image_utils.py size screen.png

python3 scripts/draw_on_image.py screen.png marker target 3788 2080 --text "Candidate 1" -o marked1.png
python3 scripts/draw_on_image.py screen.png marker target 3790 2090 --text "Candidate 2" -o marked2.png
python3 scripts/draw_on_image.py screen.png marker target 3785 2085 --text "Candidate 3" -o marked3.png

python3 scripts/draw_overlay.py marker target 3790 2090 --duration 3
python3 scripts/keyboard_mouse.py mouse_click_at 3790 2090 left

Image locating

Built on OpenCV template matching and RapidOCR. Supports locating UI elements by image and by text.

Install dependencies

pip install opencv-python numpy rapidocr_onnxruntime

Note: RapidOCR models are ~15MB and are downloaded automatically on first use.

Template matching (find by image)

python3 scripts/image_finder.py image button.png
python3 scripts/image_finder.py image button.png --all
python3 scripts/image_finder.py image button.png --threshold 0.95
python3 scripts/image_finder.py image button.png --mark
python3 scripts/image_finder.py image button.png --click

Output example:

✅ Match found: position (3788, 2080), similarity: 98.50%

OCR text locating (find by text)

python3 scripts/image_finder.py text "Send"
python3 scripts/image_finder.py text "OK" --click
python3 scripts/image_finder.py text "Send" --mark-on-image checked.png
python3 scripts/image_finder.py text-all
python3 scripts/image_finder.py text "Login" --confidence 0.9

Output example:

✅ Found 2 candidates containing 'Send':
  [1] Text: 'Send', position: (3788, 2080), confidence: 95%
  [2] Text: 'Send to all', position: (2100, 1500), confidence: 88%

Recommended automation workflows

Template matching (most accurate):

python3 scripts/image_finder.py image qq_send_button.png --threshold 0.9
python3 scripts/draw_on_image.py marker screen.png target 3788 2080 --text "Candidate 1" -o check1.png
python3 scripts/keyboard_mouse.py mouse_click_at 3788 2080 left

OCR text locating (when no template is available):

python3 scripts/image_finder.py text "Send"
python3 scripts/keyboard_mouse.py mouse_click_at 3548 1462 left

Important principle:

1. OCR returns accurate screen coordinates; do not modify the returned coordinates
2. If there are multiple candidates, mark them on an image to visually choose the correct one
3. Once you choose the right candidate, click using the original coordinates

Template matching vs OCR

Why this is better than guessing coordinates

1. High precision and repeatability (pixel-level)
2. Local compute with no API cost
3. Fast response
4. Easy to debug via marked outputs

Workflow Patterns

Standard workflow patterns for reliable UI automation. These patterns combine the tools above into proven strategies.

Core Principles

1. Step-by-step confirmation: Pause, check, and adjust at each step
2. Human-friendly: Support manual intervention at any time
3. Failure recovery: Retry or skip on failure
4. Flexible composition: Combine strategies as needed

Pattern 1: Locate-Verify-Click

Workflow:

1. Screenshot / region screenshot
2. OCR/image finder to locate target
3. Mark all candidates on image
4. AI/human selects the correct one
5. Draw overlay on screen to confirm position
6. Click (or manual click)

Use cases: Button clicks, menu selections

Command sequence:

# Region screenshot to reduce analysis scope
python3 scripts/keyboard_mouse.py screenshot_region check.png 2800 300 3800 1200

# OCR to find candidates
python3 scripts/image_finder.py text "Send" --mark-on-image candidates.png

# Overlay confirmation on screen
python3 scripts/draw_overlay.py marker target 3548 1462 --duration 3

# Click after confirmation
python3 scripts/keyboard_mouse.py mouse_click_at 3548 1462 left

Pattern 2: Search-Input-Verify

Workflow:

1. Check if target already exists
2. If not, open search
3. Type search keyword
4. Screenshot to verify search results
5. Click the correct result
6. Confirm target appeared

Use cases: Finding contacts, search functions

Command sequence:

# 1. Check if exists
python3 scripts/image_finder.py text "John" --mark-on-image check1.png

# 2. Click search box (estimated or located)
python3 scripts/keyboard_mouse.py mouse_click_at 2950 250 left

# 3. Quick input via clipboard
python3 scripts/keyboard_mouse.py copy_paste "John"

# 4. Verify results
python3 scripts/keyboard_mouse.py screenshot_region result.png 2800 400 3200 800
python3 scripts/image_finder.py text "John" --mark-on-image check2.png

# 5. Click result
python3 scripts/keyboard_mouse.py mouse_click_at 3000 600 left

# 6. Confirm target appeared
python3 scripts/image_finder.py text "John" --mark-on-image verify.png

Pattern 3: Form Filling

Workflow:

1. Locate first field
2. Click input box
3. Enter value (copy_paste)
4. Locate next field (Tab or click)
5. Repeat until all fields complete
6. Screenshot to confirm
7. Click submit

Use cases: Form filling, configuration settings

Command sequence:

# Process each field
python3 scripts/keyboard_mouse.py mouse_click_at 1000 500 left  # Field 1
python3 scripts/keyboard_mouse.py copy_paste "value1"

python3 scripts/keyboard_mouse.py mouse_click_at 1000 600 left  # Field 2
python3 scripts/keyboard_mouse.py copy_paste "value2"

# Or use Tab to navigate
python3 scripts/keyboard_mouse.py key_press tab
python3 scripts/keyboard_mouse.py copy_paste "value3"

# Final verification and submit
python3 scripts/keyboard_mouse.py screenshot verify.png
python3 scripts/keyboard_mouse.py mouse_click_at 1200 800 left  # Submit button

Pattern 4: Regional Precise Location

Workflow:

1. Roughly locate region (e.g., know QQ is on the right)
2. Region screenshot to narrow scope
3. Precise locate within small region
4. Calculate actual screen coord = region top-left + relative coord
5. Click

Use cases: When you know approximate position but need precision

Command sequence:

# 1. Region screenshot (QQ window area)
python3 scripts/keyboard_mouse.py screenshot_region qq_area.png 2800 200 3840 1500

# 2. Find within small region
python3 scripts/image_finder.py text "Send" --mark-on-image local_find.png
# Returns relative coords like (500, 1300)

# 3. Calculate actual coords
# x = 2800 + 500 = 3300
# y = 200 + 1300 = 1500

# 4. Click
python3 scripts/keyboard_mouse.py mouse_click_at 3300 1500 left

General Strategies

Strategy A: Progressive Confirmation

Instead of executing all at once, confirm at each step:

User: Click the send button for me

AI:
1. Screenshot
2. OCR find "Send"
3. Mark candidates for you to see
4. Ask: Is candidate 2 correct?
5. After your confirmation, draw overlay for 3 seconds
6. Click only when you say "go"

Strategy B: Failure Retry

Handle step failures:

- Can't find target → Expand region and retry
- Multiple matches → Mark all and let user choose
- Click not working → Check if window is focused
- OCR failed → Switch to template matching

Strategy C: Human-Machine Collaboration

Machine does what it's good at, human makes decisions:

Machine: Screenshot, locate, mark, execute clicks
Human: Judge selections, confirm positions, handle exceptions

Common Scenarios

QQ Message Sending:

1. Check if QQ window exists
2. Focus QQ window
3. Check if contact is already open
4. Search for contact (if not open)
5. Type message
6. Find and click Send button
7. Verify message appears in chat

Web Form:

1. Locate form area (scroll if needed)
2. For each field:
   - Find label
   - Click input box
   - copy_paste value
   - Tab or click next
3. Screenshot to confirm
4. Click submit
5. Wait and verify result

Best Practices

1. Prefer region screenshots over full screen

   • Reduces analysis time
   • Improves OCR accuracy
   • Reduces false matches

2. Mark multiple candidates with

   --mark-on-image

   • Let user/AI choose
   • Avoid clicking wrong positions blindly

3. Use copy_paste instead of type_text for long text

   • Faster input
   • Avoids Chinese input issues
   • Good for fixed content

4. Offset clicking when you know base position

   • Base coord + offset = target coord
   • Works for UIs with fixed relative positions

5. Screenshot at each important step

   • Easier to debug
   • Can review the process
   • Supports post-analysis

Error Handling

Cleanup

Analyze disk usage

python3 scripts/cleanup.py analyze .

Clean files

python3 scripts/cleanup.py clean . --days 7
python3 scripts/cleanup.py clean . --days 7 --execute
python3 scripts/cleanup.py clean . --size 1024 --execute
python3 scripts/cleanup.py clean . --execute

Auto cleanup

python3 scripts/cleanup.py auto . --max-files 50 --max-size 100
python3 scripts/cleanup.py auto . --max-files 20 --max-size 50

End-to-end example

python3 scripts/keyboard_mouse.py screenshot screen.png
python3 scripts/draw_on_image.py marker screen.png target 500 300 --text "Button" -o marked.png

python3 scripts/cleanup.py analyze .
python3 scripts/cleanup.py clean . --days 1 --execute
python3 scripts/cleanup.py auto . --max-files 10 --max-size 50

Command quick reference

Mouse/keyboard (

keyboard_mouse.py

)

screen_size

keyboard_mouse.py screen_size

mouse_position

keyboard_mouse.py mouse_position

mouse_move x y

keyboard_mouse.py mouse_move 500 300

mouse_click button

keyboard_mouse.py mouse_click left

mouse_click_at x y button

keyboard_mouse.py mouse_click_at 500 300 left

mouse_double_click x y

keyboard_mouse.py mouse_double_click 500 300

mouse_drag x1 y1 x2 y2

keyboard_mouse.py mouse_drag 500 300 800 600

mouse_scroll amount

keyboard_mouse.py mouse_scroll 5

key_press key

keyboard_mouse.py key_press enter

key_hotkey key1 key2

keyboard_mouse.py key_hotkey ctrl c

type_text text

keyboard_mouse.py type_text "Hello"

screenshot path

keyboard_mouse.py screenshot img.png

Image utilities (

image_utils.py

)

info path

image_utils.py info photo.png

size path

image_utils.py size photo.jpg

crop x1 y1 x2 y2

image_utils.py crop img.png 100 100 500 500

Screen overlay (

draw_overlay.py

)

marker type x y

draw_overlay.py marker target 500 300

area x1 y1 x2 y2

draw_overlay.py area 100 100 500 400

Draw on image (

draw_on_image.py

)

marker type x y

draw_on_image.py img.png marker target 500 300

area x1 y1 x2 y2

draw_on_image.py img.png area 100 100 500 400

Image finder (

image_finder.py

)

image template

image_finder.py image button.png

text str

image_finder.py text "Send"

text-all

image_finder.py text-all

Cleanup (

cleanup.py

)

analyze dir

cleanup.py analyze .

clean dir

cleanup.py clean . --days 7 --execute

auto dir

cleanup.py auto . --max-files 50