- Home
- Skills
- Agents & Orchestration
- Android Agent Bridge
Works with the AI Agents you already use
Android Agent Bridge
Control, automate, and inspect physical or emulated Android devices via ADB, Shizuku, and Root.
$12
Android Agent Bridge
Example session with this skill installed
Switch my Android phone to Wireless ADB so I can unplug the USB cable, audit its battery and privilege level, then open Settings, search for "Developer options", tap the result, and show me the active screen controls.
- Read your context and instructions
- Compiled the android agent bridge
I have transitioned your device to Wireless ADB, audited its hardware status, navigated into Developer Options via search, and grounded the active toggles. You can now safely unplug the USB cable.
1. Wireless ADB Handoff
- Discovered Wi-Fi IP:
192.168.1.x - Port Mode: Switched ADB daemon to TCP/IP port
5555 - Connection:
connected to 192.168.1.x:5555 - Status: 🟢 Wireless ADB active. Physical USB cable is now optional.
2. Device & Privilege Telemetry
- Hardware: Android Test Device (Android 12, API 31)
- Serial:
[REDACTED]| Battery: 95% - Privilege Level: Tier 1: ADB Shell (UID 2000) (
shell) - Shizuku Server: 🟢 Active (Tetherless on-device execution ready)
- OEM Unlock Status: Supported (Toggle in Developer Options available)
3. Multi-Step Navigation & Text Input
- Launched Settings:
am start -a android.settings.SETTINGS - Perception: Grounded
Search settingsinput at center coordinate(360, 185). - Search Injected: Tapped input and typed
Developer%soptions. - Result Selection: Located
Developer optionsresult in list at(360, 342)and executed tap.
4. Active Screen Grounding (Developer Options)
ui_inspector.py parsed the live hierarchy into 18 interactable toggles & items:
| Idx | Center (X, Y) | View Class | Element Label / Control | State |
|---|---|---|---|---|
[1] | (620, 210) | Switch | Developer options master toggle | ON |
[2] | (360, 415) | TextView | Memory status | Info |
[3] | (360, 580) | TextView | Bug report | Action |
[4] | (620, 745) | Switch | OEM unlocking (Allow bootloader unlock) | OFF |
[5] | (620, 910) | Switch | USB debugging (ADB authorized) | ON |
[6] | (620, 1075) | Switch | Wireless debugging | ON |
[7] | (360, 1240) | TextView | Revoke USB debugging authorizations | Action |
5. Verification & State Delta
- Active Focus:
com.android.settings/.SubSettings - Confirmed Screen:
DevelopmentSettingsDashboardActivity
Result: ✅ Device is fully un-tethered on Wi-Fi, Developer Options is open, and all toggle coordinates are mapped for immediate interaction.
Connects securely to your tools. The creator never sees your data.
About this skill
Control Android Devices Directly From Your AI Agent
Android Agent Bridge equips AI coding agents (Claude Code, Antigravity, Cursor, Codex) and Cloud Agents (Grok Bot, Dots, Muse) with an intelligent, full-stack device control harness.
Whether you need to test mobile app UIs, automate daily phone workflows, debug via Wireless ADB, or perform authorized device diagnostics, this skill bridges the gap between language models and physical Android hardware.
Key Capabilities
- Three-Tier Escalation Model:
- Tier 1 (Universal ADB Shell): No root or data wipe needed for common tasks (tapping, typing, screen capture, launching apps, granting permissions).
- Tier 2 (Tetherless Shizuku): Run ADB-privileged automations directly on-device (via Shizuku/rish in Termux) without needing a computer attached.
- Tier 3 (Superuser Root): Guarded diagnostic checks for internal app inspection (
/data/data/*), SELinux checks, and comprehensive OEM unlocking/Magisk reference playbooks.
- Screen Perception & Coordinate Grounding: Extracts live Android view hierarchies (
uiautomator dump) into element center coordinates so agents don't guess coordinates. - Wireless ADB Handoff: Automated transition from USB debugging to Wireless ADB over TCP/IP port 5555.
- Flexible Execution (Python + Direct ADB): Includes zero-dependency Python 3 tools plus direct
adbCLI commands for sandboxed agent environments. - Guarded Root Diagnostics: Root access is restricted to fixed read-only checks (
id,getenforce,uptime,whoami, app-data listing), with input validation against shell injection.
Why This Beats Prompting Raw ADB
Writing ad-hoc ADB prompts leads to coordinate guessing, missed click targets, and shell escaping bugs. This skill uses a deterministic perception loop that grounds UI elements into verified coordinates, verifies state changes via a 3-part contract (Device Status → Grounded Action → Verified Result), and strictly guards root commands.
Included in this Product Bundle
SKILL.md— Core orchestration playbook, 3-part output contract, and direct CLI fallbacksscripts/device_doctor.py— Device telemetry, battery, and privilege auditorscripts/ui_inspector.py— Live screen perception and element coordinate grounding parserscripts/wireless_connect.py— Wireless ADB and TCP/IP connection managerscripts/root_exec.py— Guarded su command runner & root diagnosticreferences/rooting_playbook.md— Device-agnostic OEM unlocking (Pixel, Motorola, Xiaomi, Samsung) & Magisk guidereferences/adb_patterns.md— High-frequency intents, dumpsys filters, and keyeventsreferences/shizuku_rish.md— On-device tetherless execution with Shizuku and Termuxexamples/README.md— Real-world worked examples and input/output contractsREADME.md— Complete customer installation and onboarding guide
Quick Start
Connect your phone via USB or Wi-Fi, open your AI Agent, and prompt:
"Connect to my Android device via ADB, check its battery and privilege level, and show me what buttons are on the screen."
Compatibility & Limitations
- Supported: Physical devices and emulators running Android 11+.
- Not supported: iOS or Apple mobile devices.
- Note: Android 14+ blocks system CA certificate injection into system directories (APEX). Carrier-locked bootloaders may restrict OEM unlocking.
No iOS support. Android 14+ restricts system CA cert installation. Requires ADB or Shizuku setup. Bootloader operations blocked on carrier-locked devices.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 9 days ago
Requires