Phone Agent Skill for Openclaw

A mobile automation skill that enables AI agents to control Android devices via natural language for testing, data collection, and UI interaction.

gaojizhou
v1.0.0
Mar 4, 2026
0
1.3k
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install phone-agent-skill

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install phone-agent-skill using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is Phone Agent Skill?

The AutoGLM Phone Agent Skill allows developers to drive Android devices or emulators using natural language instructions. By integrating with the AutoGLM Phone Agent SDK, this skill can perform complex mobile operations such as tapping, typing, swiping, and capturing screenshots. It is a critical component for developers using Openclaw Skills who need to automate end-to-end testing or interact with mobile applications that lack traditional APIs.

This skill transforms high-level goals into executable UI actions, making it possible to navigate complex app interfaces without manual scripting. Whether you are conducting user journey research or automating routine mobile tasks, this integration provides a robust interface between AI logic and physical device interaction.

Phone Agent Skill Use Cases

  • Automated end-to-end testing of Android applications.
  • Automated data collection from mobile-only apps and services.
  • Simulation of user journeys for UI/UX research and bug reproduction.
  • Verification of app states through automated screenshot capture and text analysis.
  • Remote control of mobile devices for productivity and task automation workflows.

How Phone Agent Skill Works

  1. The user provides a natural language instruction to the agent.
  2. The Openclaw Skills framework processes the intent and identifies the necessary mobile interaction steps.
  3. The skill communicates with the AutoGLM Phone Agent backend via a designated endpoint.
  4. Instructions are translated into specific ADB commands or accessibility actions on the Android device.
  5. The device executes the UI interaction (e.g., clicking a button or entering text).
  6. The skill retrieves feedback, such as action logs or screenshots, to confirm the result.

Phone Agent Skill Setup

  1. Ensure an Android device or emulator has USB debugging enabled and is recognized by your system:
adb devices
  1. Start the AutoGLM Phone Agent service following the SDK documentation.
  2. Configure the environment variable for the agent runtime to point to your service URL:
export PHONE_AGENT_ENDPOINT=http://127.0.0.1:5000
  1. Grant the necessary overlay and accessibility permissions on the device when prompted to ensure Openclaw Skills can interact with the UI.

Phone Agent Skill Data Schema & Taxonomy

Data Component Description
Action Logs Detailed records of tap, swipe, scroll, and input actions performed.
UI Screenshots Visual verification of the device screen state in image format.
Structured Observations Metadata describing UI elements and text recognized on the screen.
Error Feedback Real-time troubleshooting data returned from the Phone Agent backend.

Phone Agent Skill Advanced Features

  • Natural language UI anchoring: Identify buttons and fields by their on-screen text.
  • Multi-step task decomposition: Break down complex goals into a series of logical UI operations.
  • Remote endpoint support: Connect to devices across a network using the PHONE_AGENT_ENDPOINT configuration.
  • Safety-first execution: Provides visual confirmation through screenshots before or after risky UI changes.
  • Full ADB integration: Leverages standard Android debugging protocols for maximum device compatibility.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*