mcpbeat Sign in

Interacting With Android Device Skill for Claude

Instructions for capturing UI state, comparing with mocks, and interacting with an Android device using MCP tools backed by ADB.

2k tokens
context cost
the whole folder, loaded on every use
1
files
instructions only
0
copies elsewhere
how many repositories repackaged it
9164
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/bitwarden/android --skill interacting-with-android-device

What it tells the agent to use

found in the instruction text
Bash runs shell commands — read the instruction before connecting

The instruction itself

18 sections, as written by the author

Interacting with Android Device

Quick Start: MCP Tools

The android-device MCP server provides 6 tools for device interaction. These replace the previous shell scripts with proper XML parsing, structured dumpsys parsing, and native obstruction detection.

Available tools:

  • capture — Capture UI hierarchy XML and/or screenshot. Params: { xml?: boolean, screenshot?: boolean }. Default: both.
  • find_element — Find element by text or content-desc, return coordinates with obstruction detection. Params: { text: string }. Returns JSON with coordinates, bounds, and obstruction status.
  • tap_at — Tap at specific coordinates, wait, capture screenshot. Params: { x, y, waitSeconds? }.
  • tap_element — Find, tap, and capture in one call (recommended). Params: { text, waitSeconds? }. Auto-adjusts coordinates when obstructed.
  • navigate — Navigation actions: home, back, app-drawer. Params: { action, waitSeconds? }. Captures screenshot after action.
  • input_text — Type text into the focused field. Params: { text, clear? }. Set clear: true to erase existing content first.

Use these MCP tools instead of raw ADB commands to save tokens, get structured results, and benefit from automatic obstruction detection.

1. Capturing Current State

To understand what is currently on the device, use the capture tool:

  • It saves view.xml (UI hierarchy) and screen.png (screenshot) to the working directory
  • Read view.xml to find coordinates (bounds) and properties (like text or resource-id) of UI elements
  • Use screen.png for visual verification against design mocks

2. Interacting with the Device

  • Find and tap an element by text — use tap_element:

This finds the element, detects obstructions, taps (with adjusted coordinates if needed), and captures a screenshot — all in one call.

  • Tap at specific coordinates — use tap_at:

When you already have coordinates from find_element or manual inspection.

  • Navigate (home, back, app-drawer) — use navigate:

Performs the action and captures a screenshot.

  • Find element without tapping — use find_element:

Returns coordinates and full element info. Useful when you need to inspect before acting.

  • Type text into a field — use input_text:

Types text into the currently focused field. Set clear: true to erase existing content first. Tap the field before calling this if it isn't already focused.

Raw ADB Commands (When MCP Tools Aren't Sufficient)

  • Key Events:
  • Back: adb shell input keyevent 4
  • Home: adb shell input keyevent 3
  • Enter: adb shell input keyevent 66
  • Scrolling/Swiping: Use adb shell input swipe <x1> <y1> <x2> <y2> <duration_ms> where:
  • (x1, y1) = starting point
  • (x2, y2) = ending point
  • duration_ms = duration in milliseconds (1000ms is typical; adjust for speed/distance)
  • Note: For expanding containers/drawers, use large distances (e.g., 2400->300 for a 2992px tall screen)

3. Obstruction Detection

The find_element and tap_element tools automatically detect when another element would intercept the tap. This catches:

  • System overlays (Layer 1): TalkBack floating menu, PiP windows, accessibility services — detected via dumpsys window windows touchable regions
  • In-app elements (Layer 2): FABs, dialogs, bottom sheets, snackbars — detected by finding the topmost clickable element at the tap point in the UI hierarchy

When obstruction is detected:

  • Coordinates are auto-adjusted to the center of the largest unobstructed strip (top/bottom/left/right of the obstructor)
  • The response includes the obstructor identity, bounds, and visible region info
  • If fully obscured (no visible region), the original center is returned as best-effort
  • Compose parent wrapper pattern (identical bounds) is recognized as non-obstruction

4. Verification Workflow

Follow these steps for a complete UI test:

  • Build and Install: Ensure the latest version of the app is running: ./gradlew installDebug.
  • Inspect: Use capture to dump the UI hierarchy and take a screenshot.
  • Compare: Check the current UI against any mock image files in the project.
  • Interact: Use tap_element to tap a UI element by text. The tool handles coordinate calculation and obstruction detection automatically.
  • Verify: Use capture again to confirm the UI has updated as expected (e.g., a new screen is shown, or a success message appeared).

5. Examples

Example: Navigate to Settings and Check for Updates

# Go to home screen
navigate({ action: "home" })

# Open app drawer
navigate({ action: "app-drawer" })

# Find and tap through settings
tap_element({ text: "Settings", waitSeconds: 2 })
tap_element({ text: "System", waitSeconds: 2 })
tap_element({ text: "Software updates", waitSeconds: 2 })
tap_element({ text: "Check for update", waitSeconds: 5 })

Example: Swiping

For swipe gestures not covered by the navigate tool, use raw ADB:

adb shell input swipe 672 2800 672 500 1000 && sleep 1 && adb shell screencap -p /sdcard/screen.png && adb pull /sdcard/screen.png .

6. Best Practices

Coordinate Calculation

  • Prefer find_element or tap_element over manual coordinate calculation — they handle bounds parsing, center computation, and obstruction detection automatically
  • When multiple instances of an element exist (e.g., in prediction row and full list), check the find_element response to verify you're targeting the correct one
  • Verify after each interaction: Don't assume an action succeeded — use capture after interactions to confirm the UI changed as expected
  • Check both visual and structural state: Use screenshot for visual verification, XML dump for structural confirmation (element presence, text content, state changes)
  • Identify navigation failures early: If a tap opened the wrong screen, use navigate({ action: "back" }) to recover immediately

Interaction Patterns

  • Scrolling before interaction: When looking for an element, check if it's visible on screen first. If not, scroll using swipe gestures to reveal it
  • Use consistent scroll direction: For vertical scrolling in lists/settings, use downward swipes (higher Y -> lower Y) to scroll down
  • Handle app crashes gracefully: Don't retry the same action — use back button and try an alternative approach
  • Check Accessibility: Use the content-desc and text properties in the UI hierarchy to ensure the UI is accessible for screen readers

7. Troubleshooting

Device Not Connected

If tools report ADB errors:

  • Check USB connection or emulator status
  • Enable USB debugging on the device (Settings > Developer Options > USB Debugging)
  • Accept the RSA key prompt on the device if asked
  • Restart the device or disconnect/reconnect the USB cable
  • Run adb devices to verify the device is visible

MCP Server Not Available

If tools are not listed in /mcp:

  • Ensure Node.js 18+ is installed
  • The server auto-builds on first use via .mcp.json at the project root
  • Check .claude/mcp/android-device-server/ exists with package.json
  • Try manual build: cd .claude/mcp/android-device-server && npm install && npm run build

Other skills for the same job

different authors, same section of the catalogue
Protocolsio Integration
by christophacham
×4

Integration with protocols.io API for managing scientific protocols. This skill should be used when working with protocols.io to search, create, update, or publish protocols; manage protocol steps and materials; handle discussions and comments; organize workspaces; upload and manage files; or integrate protocols.io functionality into workflows. Applicable for protocol discovery, collaborative protocol development, experiment tracking, lab protocol management, and scientific documentation.

16k tokens
Tailored Resume Generator
by frostant
×4

Analyzes job descriptions and generates tailored resumes that highlight relevant experience, skills, and achievements to maximize interview chances

3k tokens
Excalidraw Diagram Generator
by github
vendor ×3

Generate Excalidraw diagrams from natural language descriptions. Use when asked to "create a diagram", "make a flowchart", "visualize a process", "draw a system architecture", "create a mind map", or "generate an Excalidraw file". Supports flowcharts, relationship diagrams, mind maps, and system architecture diagrams. Outputs .excalidraw JSON files that can be opened directly in Excalidraw.

36k tokens scripts
Expo Dev Client
by openai
vendor ×3

Build and distribute Expo development clients locally or via TestFlight

961 tokens
Executing Plans
by ZhanlinCui
×3

Use when you have a written implementation plan to execute in a separate session with review checkpoints

542 tokens
Anndata
by christophacham
×3

Data structure for annotated matrices in single-cell analysis. Use when working with .h5ad files or integrating with the scverse ecosystem. This is the data format skill—for analysis workflows use scanpy; for probabilistic models use scvi-tools; for population-scale queries use cellxgene-census.

16k tokens
Benchling Integration
by christophacham
×3

Benchling R&D platform integration. Access registry (DNA, proteins), inventory, ELN entries, workflows via API, build Benchling Apps, query Data Warehouse, for lab data management automation.

14k tokens
Biopython
by christophacham
×3

Comprehensive molecular biology toolkit. Use for sequence manipulation, file parsing (FASTA/GenBank/PDB), phylogenetics, and programmatic NCBI/PubMed access (Bio.Entrez). Best for batch processing, custom bioinformatics pipelines, BLAST automation. For quick lookups use gget; for multi-service integration use bioservices.

24k tokens

How to use it

Copy the folder

Take bitwarden/interacting-with-android-device from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.

Install what it needs

The instructions reference npm. Without those the skill loads but fails at the first command.