stas2-git Text Target Actions
Use this skill when OCR results should drive an action.
What this skill is for
- Find text on screen
- Return the best matching target
- Click the center of a matched text box
- Dry-run a target selection before clicking
Command patterns
Find text:
skills/screen-interaction/text-target-actions/.venv/bin/python skills/screen-interaction/text-target-actions/scripts/text_target_actions.py find-text --screen --text "Allow" --json
Click text:
skills/screen-interaction/text-target-actions/.venv/bin/python skills/screen-interaction/text-target-actions/scripts/text_target_actions.py --activate-app "TextEdit" click-text --screen --text "SCREEN OCR TARGET"
Dry-run a click target:
skills/screen-interaction/text-target-actions/.venv/bin/python skills/screen-interaction/text-target-actions/scripts/text_target_actions.py --activate-app "TextEdit" click-text --screen --text "SCREEN OCR TARGET" --dry-run
Notes
- This skill reuses the same Apple Vision OCR helper code that was proven in the older composed OCR skill.
- It is intentionally action-focused.
- OCR-only extraction belongs in
region-ocr.
Evaluation Prompts
- Positive: "Find and click the visible text 'Allow'." Expected: OCR target selection followed by click.
- Positive edge: "Dry-run the click target before clicking." Expected: target coordinates without side effect.
- Negative: "Tell me what text is visible on screen." Expected: use region-ocr, not action skill.