A small collection of reusable skills for AI coding agents.
| Skill | Description |
|---|---|
| android-background-test | Test apps on a real Android phone over ADB on a hidden background display: screenshot, UI tree, tap by element or coordinate, swipe, keys, Chinese text, and step-by-step journeys, without disturbing the user. |
| web-shader-extractor | Extract WebGL/Canvas/Shader effects from websites and port them into standalone native JS projects. |
Install the whole repository:
npx skills add https://github.com/lixiaolin94/skillsInstall a specific skill:
npx skills add https://github.com/lixiaolin94/skills --skill web-shader-extractor
npx skills add https://github.com/lixiaolin94/skills --skill android-background-testStatus: beta (0.2.0-dev). Verified on nubia NX888J (Android 17) and OPPO PLG110 (ColorOS, Android 17). Other devices and ROMs are untested; reports with model and ROM version are welcome.
Requirements:
adbon the computer and Python 3.9+ (standard library only)- USB debugging enabled and the computer authorized on the phone
- Android 12 (API 31) or newer is the build target; the lowest working version has not been tested
- The phone must stay awake and unlocked while testing; locking pauses the session
- Some ROMs (for example Xiaomi) may need an extra "USB debugging (security settings)" switch before taps work; untested
Apps that police automation (payment, banking, social, e-commerce) are only opened after the user accepts the account risk; the skill stops and explains when an app blocks it, and never tries to evade detection. The UI tree, element taps and these consent flows are new in this build and not yet verified on a device.
Nothing is installed on the phone. runtime/server.jar is a DEX built from the source in workspace/android-background-test; rebuild it with python3 workspace/android-background-test/scripts/build.py and compare the sha256 in android-background-test/manifest.json.
| Contributor | Contributions |
|---|---|
| Huazi | web-shader-extractor — 2D Canvas extraction report that led to 2D Canvas identification, porting strategy, and framework shell-removal rules |
MIT
