phone-elements

Run the elements.py script with --text followed by the button label and the --tap flag. It dumps the UI hierarchy via uiautomator, parses the XML, computes the element's center coordinates, and issues an input tap command through Shizuku or ADB.

uiautomator dump returns structured XML with exact element bounds in about 1 second, while screenshot plus vision analysis takes 7-8 seconds. Use uiautomator when you know the target text or ID; use vision only for unknown layouts or scene understanding.

Yes, but MIUI has a known bug where uiautomator dump returns a non-zero exit code even when the XML output is valid. The script handles this by checking the XML content rather than the exit code, so parsing succeeds on MIUI and HyperOS devices.

Some OEMs like vivo OriginOS do not support dumping to /dev/tty. The script automatically falls back to a file-based dump at /sdcard/ui_dump.xml and reads it back, so it works across AOSP, Samsung One UI, vivo, and Xiaomi devices.

You need either Shizuku with the rish command or ADB connected over TCP at 127.0.0.1:5555, plus Python 3. The XML parser uses xml.etree.ElementTree from the Python standard library, so no third-party packages are required.