Device Streaming and Android skills - available in Android CLI

Google announced updates to its Android CLI command-line tooling that let AI agents access real physical devices remotely through Android Device Streaming over a secure ADB over SSL connection, enabling device spin-up, build deployment, log and trace collection, and headless screenshots from the terminal. The company also expanded its Android skills repository to over 20 structured SKILL.md instruction files, including a new Wear Compose Material 3 skill, and added Android CLI commands such as `android init`, `android skills list`, `android skills add wear-compose-m3 --project=.`, and `android skills update --all`.

As Android developers, you have many choices when it comes to the agents, LLMs, tools, and command-line interfaces CLI you use for app development. Our goal is to help you build beautiful, high-quality Android apps, no matter how you choose to build. We’re announcing updates to our Android CLI https://developer.android.com/tools/agents/android-cli command-line tooling, including the ability to access real devices through Android Device Streaming. We’re also sharing more Android skills https://developer.android.com/tools/agents/android-skills , and giving you a deep dive into the Wear OS Compose Material 3 skill. Android CLI is our tool to make command-line interface Android development easier. It supports any AI agent or tool in building more efficiently for Android. Along with the android-cli skill, your agents use Android CLI to create, build, test, and manage Android projects, help you set up the development environment, create and run emulators, and execute test runs. It’s important to test your app on real devices to catch issues that are hardware- or OS-specific, but sometimes you don’t have access to the ones you need. Android Device Streaming gives you access to real physical devices, remotely. Your agent can now use Android Device Streaming anywhere, through Android CLI. The agent can interact with physical devices as if they were plugged in over USB over a secure ADB over SSL connection. This enables spinning up devices, deploying builds, collecting logs and traces, and even capturing screenshots headlessly—all through the terminal. To get started, link your project, instruct your agent to list available remote devices and decide which device you want next. Read the documentation https://developer.android.com/tools/agents/android-cli/commands/device remote to learn more, and see the Android CLI release notes https://developer.android.com/tools/agents/android-cli/release-notes . To bridge the gap between LLMs’ default knowledge and platform-specific standards and updates, we keep growing our Android skills repository . Android skills are structured instructions SKILL.md files that ground AI agents with our official guidance from developer.android.com http://developer.android.com . Instead of relying on a model’s training cutoff, skills provide more precise and fresher data, API references and samples, and architectural patterns directly into your agent’s context. With over 20 skills available https://developer.android.com/tools/agents/android-skills/browse , you can now equip your AI agents to handle more complex and specialized development tasks such as: Our Android skills are thoroughly evaluated. To understand the philosophy and methodology behind this project, as well as why all skills should come with evals, make sure to read Inside Android Skills - Built for deprecation https://developer.android.com/blog/posts/inside-android-skills-built-for-deprecation . Managing skills across projects and individual agent directories is pretty straightforward with Android CLI: android init to install the android-cli skill android skills list android skills add wear-compose-m3 --project= . android skills update --all android skills update wear-compose-m3 for an individual skills Skills are designed to be environment-agnostic . From writing code in Android Studio and Antigravity, to pairing with third-party agents like Claude and Codex, our Android skills work across your entire setup. Building for Wear OS means distinct design and development decisions: round viewports, rotary input, ambient display mode, minimizing power consumption, preferring TransformingLazyColumn , and using the AppScaffold and ScreenScaffolds containers. Without explicit guidance, LLMs lack the understanding and knowledge of these distinct patterns that make the Wear apps really stand out and shine. To help agents with this, we released the Wear Compose Material 3 skill wear/wear-compose-m3 . Check out this video for more information on how powerful this skill is: Early adopters are already seeing significant productivity impact with this skill. The engineering team at FotMob used it for tasks like modernizing their existing Wear M3-based app, migrating multiple lists to TransformingLazyColumn with ScreenScaffold content padding, ListHeader titles, SurfaceTransformation on cards and buttons, theme typography, and Wear previews. The result? The changes compiled successfully and were verified on the emulator for scrolling, rotary, edge morphing and RTL. This enabled the team to delete their legacy wrapper, and all rotary and focus boilerplate. The skill caught mistakes that the underlying model missed, such as forgetting to forward ScreenScaffold 's contentPadding into the list, and using theme typography over hardcoded sp. “One skill, one afternoon, eight lists migrated and a pile of custom rotary code gone ” - Roy Solberg, Android Tech Lead at FotMob. When you're ready to get started, install Android CLI with just one command https://developer.android.com/tools/agents . You can learn more about Android CLI by reading the developer documentation https://developer.android.com/tools/agents/android-cli , and as well as checking out the latest updates in the Android CLI release notes https://developer.android.com/tools/agents/android-cli/release-notes . After installing Android CLI, run android init to configure your agent, then browse the latest Android skills available for use and install them with android skills add. Also check out the Android Device Streaming documentation https://developer.android.com/studio/run/android-device-streaming to learn more about how this feature allows you to access real devices on the cloud inside both Android Studio and via your agents.