stas2-git Region Capture
Use this skill when you need pixels from the screen, not OCR or clicking.
What this skill is for
- Capture the full screen to a PNG
- Capture an exact rectangle defined by
x,y,width, andheight - Optionally activate an app just before capture for more stable testing
Command patterns
Capture the full screen:
skills/screen-interaction/region-capture/.venv/bin/python skills/screen-interaction/region-capture/scripts/region_capture.py capture-fullscreen --output /tmp/full.png
Capture a region:
skills/screen-interaction/region-capture/.venv/bin/python skills/screen-interaction/region-capture/scripts/region_capture.py capture-region --x 100 --y 100 --width 400 --height 200 --output /tmp/region.png
Capture a region after activating an app:
skills/screen-interaction/region-capture/.venv/bin/python skills/screen-interaction/region-capture/scripts/region_capture.py --activate-app "TextEdit" capture-region --x 100 --y 100 --width 400 --height 200 --output /tmp/region.png
Notes
- This skill is intentionally capture-only.
- OCR belongs in
region-ocr. - Clicks belong in
text-target-actions.
Evaluation Prompts
- Positive: "Capture this exact screen region for later OCR." Expected: screenshot artifact or JSON path, no OCR.
- Positive edge: "Capture the active Excel window before validation." Expected: window/app-aware capture when requested.
- Negative: "Read the text in this image." Expected: use region-ocr, not capture only.