Token导航 LogoToken导航TokenDH.com
研究检索执行命令clawhub未标认证来源可访问clear审计通过

duoplus-agent双重 Agent

Agent Skill

duoplus-agent 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

18,861

周安装

763

GitHub Stars

2

下载量

5,921
OpenClaw

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT-0

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:duoplus-agent(双重 Agent)
来源仓库:https://github.com/duoplusofficial/duoplus-agent
安装命令:
openclaw skills install duoplus-agent
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install duoplus-agent

简介

通过 ADB 控制 Android 云手机,支持点击、滑动、截图等操作。

  • 适用于自动化测试或远程设备管理场景。duoplus-agent 属于研究检索类 Skill,可作为该场景下的辅助能力补充。
  • 需在 DuoPlus CloudPhone 服务上运行并配置相关环境。
  • 使用前应确认云手机服务状态和网络连接稳定性。
  • 注意权限边界,避免误操作影响设备安全。

SKILL.md

name
duoplus-agent
displayName
DuoPlus CloudPhone Agent
description
Control Android cloud phones via ADB broadcast commands - tap, swipe, type, screenshot, read UI elements. Requires DuoPlus CloudPhone service running on the device.
version
1.0.9
license
MIT-0
metadata
clawdbot
emoji
📱
requires
bins
["adb", "cwebp"]
changelog
Restructure SKILL.md, add cwebp compression, use uiautomator for UI analysis

DuoPlus CloudPhone Agent

Control and automate DuoPlus cloud phones using ADB broadcast commands.

For more information, visit DuoPlus Official Website.

Connecting Devices

Wireless Connection

adb connect <IP>:<PORT>
adb devices -l

All subsequent commands use -s <DEVICE_ID> to target a specific device.

Common Workflows

Launching an App

scripts/send_command.sh <DEVICE_ID> '{"action_name":"OPEN_APP","params":{"package_name":"com.tencent.mm"}}'

Analyzing the UI

Dump and pull the UI hierarchy to find element coordinates and attributes:

adb -s <DEVICE_ID> shell uiautomator dump /sdcard/view.xml && adb -s <DEVICE_ID> pull /sdcard/view.xml ./view.xml

Then grep for text or resource IDs to find bounds="[x1,y1][x2,y2]".

Interacting with Elements

All interactions are sent via the helper script as JSON commands:

  • Tap coordinate: scripts/send_command.sh <DEVICE_ID> '{"action_name":"CLICK_COORDINATE","params":{"x":500,"y":500}}'
  • Tap element by text: scripts/send_command.sh <DEVICE_ID> '{"action_name":"CLICK_ELEMENT","params":{"text":"Login"}}'

- Optional params: resource_id, class_name, content_desc, element_order (0-based index)

  • Long press: scripts/send_command.sh <DEVICE_ID> '{"action_name":"LONG_COORDINATE","params":{"x":500,"y":500,"duration":1000}}'
  • Double tap: scripts/send_command.sh <DEVICE_ID> '{"action_name":"DOUBLE_TAP_COORDINATE","params":{"x":500,"y":500}}'
  • Type text: scripts/send_command.sh <DEVICE_ID> '{"action_name":"INPUT_CONTENT","params":{"content":"Hello","clear_first":true}}'

- Must tap the input field first to focus it

  • Keyboard key: scripts/send_command.sh <DEVICE_ID> '{"action_name":"KEYBOARD_OPERATION","params":{"key":"enter"}}'

- Supported keys: enter, delete, tab, escape, space

  • Swipe: scripts/send_command.sh <DEVICE_ID> '{"action_name":"SLIDE_PAGE","params":{"direction":"up","start_x":487,"start_y":753,"end_x":512,"end_y":289}}'

- direction: up/down/left/right (required). Coordinates are optional.

  • Home: scripts/send_command.sh <DEVICE_ID> '{"action_name":"GO_TO_HOME","params":{}}'
  • Back: scripts/send_command.sh <DEVICE_ID> '{"action_name":"PAGE_BACK","params":{}}'
  • Wait: scripts/send_command.sh <DEVICE_ID> '{"action_name":"WAIT_TIME","params":{"wait_time":3000}}'
  • Wait for element: scripts/send_command.sh <DEVICE_ID> '{"action_name":"WAIT_FOR_SELECTOR","params":{"text":"Loading complete","timeout":10000}}'
  • End task (only when stuck): scripts/send_command.sh <DEVICE_ID> '{"action_name":"END_TASK","params":{"success":false,"message":"reason"}}'

All action commands are fire-and-forget — they do NOT return results. Take a screenshot after each action to verify.

Visual Verification

Take a screenshot, compress with cwebp, and pull to local for analysis:

# Take screenshot on device
adb -s <DEVICE_ID> shell screencap -p /sdcard/screen.png

# Pull to local
adb -s <DEVICE_ID> pull /sdcard/screen.png ./screen.png

# Compress to WebP for smaller file size (optional, recommended for vision model)
cwebp -q 60 -resize 540 0 ./screen.png -o ./screen.webp

If cwebp is not available, use the PNG directly.

How Commands Work (Internal)

Commands are sent as Base64-encoded JSON via ADB broadcast. The helper script scripts/send_command.sh handles this automatically:

# Usage: scripts/send_command.sh <DEVICE_ID> <ACTION_JSON>
scripts/send_command.sh 192.168.1.100:5555 '{"action_name":"CLICK_ELEMENT","params":{"text":"Login"}}'

The script builds the full payload (task_type, task_id, md5, etc.), Base64-encodes it, and sends via:

adb -s <DEVICE_ID> shell am broadcast -a com.duoplus.service.PROCESS_DATA --es message "<BASE64>"

Typical Workflow

1. Analyze UI    → uiautomator dump to find elements, or screenshot for visual analysis
2. Execute action → send_command.sh with the appropriate action JSON
3. Wait 1-3s     → Let the action take effect
4. Verify        → Screenshot + cwebp compress, or uiautomator dump again
5. Repeat 2-4 until all requested steps are completed

Tips

  • Coordinates: The coordinate system is 0-1000 relative (top-left=0,0, bottom-right=1000,1000), NOT pixels.
  • Element matching: Use CLICK_ELEMENT (by text) when possible; fall back to CLICK_COORDINATE when text matching fails.
  • Input: Must tap the input field first (CLICK_COORDINATE or CLICK_ELEMENT) to focus, then INPUT_CONTENT.
  • Submit: After typing, use KEYBOARD_OPERATION(key="enter") to submit.
  • Wait: Use sleep 1-3 between commands to allow the UI to update. Do NOT use shell sleep on the device.
  • Swipe coordinates: Must use irregular integers, avoid round numbers (500, 800). Vary coordinates between consecutive swipes.

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenClaw

85.13%
按下载量换算5,041

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 openclaw skills install duoplus-agent 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills