rdmptv
rdmptv/AdbAutoPlayer
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/rdmptv/AdbAutoPlayer --skill adb-screen-detection命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | adb-screen-detection |
| description | Screen understanding with OCR and template matching for Android device automation |
| version | 1.0.0 |
| modularized | true |
| scripts_enabled | true |
| tier | 2 |
| category | adb-automation |
| last_updated | "2025-12-01T00:00:00.000Z" |
| compliance_score | 100 |
| dependencies | ["pytesseract>=0.3.10","opencv-python>=4.8.0","pillow>=10.0.0","numpy>=1.24.0"] |
| auto_trigger_keywords | ["adb-screen","ocr","template-match","element-detection","screen-understanding"] |
| scripts | [{"name":"adb-screen-capture.py","purpose":"Capture Android device screen via ADB screencap","type":"python","command":"uv run .claude/skills/adb-screen-detection/scripts/adb-screen-capture.py","zero_context":true,"version":"1.0.0","last_updated":"2025-12-01T00:00:00.000Z"},{"name":"adb-ocr-extract.py","purpose":"Extract text from screen using Tesseract OCR","type":"python","command":"uv run .claude/skills/adb-screen-detection/scripts/adb-ocr-extract.py","zero_context":true,"version":"1.0.0","last_updated":"2025-12-01T00:00:00.000Z"},{"name":"adb-find-element.py","purpose":"Find UI element by template matching or OCR text","type":"python","command":"uv run .claude/skills/adb-screen-detection/scripts/adb-find-element.py","zero_context":true,"version":"1.0.0","last_updated":"2025-12-01T00:00:00.000Z"},{"name":"adb-tap-coordinate.py","purpose":"Tap device screen at specific coordinates","type":"python","command":"uv run .claude/skills/adb-screen-detection/scripts/adb-tap-coordinate.py","zero_context":true,"version":"1.0.0","last_updated":"2025-12-01T00:00:00.000Z"}] |
| color | blue |
Screen Understanding for Android Automation
What It Does: Provides OCR-based text detection and template matching to understand Android device screens. Enables reliable UI automation by verifying screen state before and after actions.
Core Capabilities:
When to Use:
Capture Android device screen and save locally.
# Basic usage
uv run .claude/skills/adb-screen-detection/scripts/adb-screen-capture.py
# Specify device
uv run .claude/skills/adb-screen-detection/scripts/adb-screen-capture.py --device 127.0.0.1:5555
# Custom output path
uv run .claude/skills/adb-screen-detection/scripts/adb-screen-capture.py --output /tmp/screen.png
# JSON output
uv run .claude/skills/adb-screen-detection/scripts/adb-screen-capture.py --json
Output:
{
"device": "127.0.0.1:5555",
"timestamp": "2025-12-01T10:30:45Z",
"local_path": "/tmp/screenshot.png",
"size": [1080, 2400]
Extract all visible text from device screen using Tesseract OCR.
# Basic usage (uses most recent screenshot)
uv run .claude/skills/adb-screen-detection/scripts/adb-ocr-extract.py
# Specify screenshot path
uv run .claude/skills/adb-screen-detection/scripts/adb-ocr-extract.py --image /tmp/screen.png
# Search for specific text
uv run .claude/skills/adb-screen-detection/scripts/adb-ocr-extract.py --search "Login"
# JSON output with coordinates
uv run .claude/skills/adb-screen-detection/scripts/adb-ocr-extract.py --json
Output:
{
"text": ["Login", "Username", "Password", "Submit"],
"detected": true,
"search_found": true,
"search_term": "Login",
"coordinates": {
"Login": [[100, 200, 150, 230]]
}
}
Find UI element by template matching or OCR text search.
# Find by OCR text
uv run .claude/skills/adb-screen-detection/scripts/adb-find-element.py \
--method ocr \
--target "Login Button" \
--threshold 0.8
# Find by template image
uv run .claude/skills/adb-screen-detection/scripts/adb-find-element.py \
--method template \
--template /path/to/template.png \
--threshold 0.8
# JSON output
uv run .claude/skills/adb-screen-detection/scripts/adb-find-element.py \
--method ocr \
--target "Login" \
--json
Output:
{
"found": true,
"method": "ocr",
"target": "Login",
"coordinates": {
"x": 100,
"y": 200,
"width": 150,
"height": 30
},
"confidence": 0.95,
"message": "Element found at (100, 200)"
}
Tap device screen at specific coordinates.
# Tap at coordinates
uv run .claude/skills/adb-screen-detection/scripts/adb-tap-coordinate.py \
--x 100 \
--y 200 \
--device 127.0.0.1:5555
# Tap with verification (check screen after tap)
uv run .claude/skills/adb-screen-detection/scripts/adb-tap-coordinate.py \
--x 100 \
--y 200 \
--verify-text "Next Screen" \
--timeout 5
# JSON output
uv run .claude/skills/adb-screen-detection/scripts/adb-tap-coordinate.py \
--x 100 \
--y 200 \
--json
Output:
{
"device": "127.0.0.1:5555",
"tap": {
"x": 100,
"y": 200
},
"success": true,
"verified": true,
"verify_text": "Next Screen",
"verification_match": true
}
# 1. Capture current screen
adb-screen-capture.py
# 2. Check for expected element
adb-find-element.py --method ocr --target "Login Button"
# 3. If found, tap it
adb-tap-coordinate.py --x 100 --y 200 --verify-text "Welcome"
# 1. Capture screen
adb-screen-capture.py
# 2. Extract all text
adb-ocr-extract.py --search "Settings"
# 3. Get coordinates and tap
adb-find-element.py --method ocr --target "Settings"
adb-tap-coordinate.py --x 150 --y 300
# 1. Have known UI template images in ./templates/
# 2. Capture screen
adb-screen-capture.py
# 3. Match against templates
adb-find-element.py --method template --template ./templates/button.png
# 4. Tap matched location
adb-tap-coordinate.py --x $(jq -r '.coordinates.x') --y $(jq -r '.coordinates.y')
Design Principles:
Dependency Relationship:
adb-screen-capture.py (foundation)
↓
adb-ocr-extract.py (uses capture)
adb-find-element.py (uses capture or templates)
↓
adb-tap-coordinate.py (uses find-element for verification)
Used By:
adb-navigation-base - Wait for elements between actionsadb-magisk - Verify Magisk UI stateadb-karrot - Verify app state during automationadb-workflow-orchestrator - Screen verification in workflowsDependencies:
adb command-line toolbrew install tesseract (macOS) or apt-get install tesseract-ocr (Linux)export TESSDATA_PREFIX=/usr/local/share/tessdata--threshold parameter (0.0-1.0)adb devicesadb connect <device>adb kill-server && adb start-serverThis skill includes TOON-based workflow definitions for automation.
TOON (Task-Oriented Orchestration Notation) is a structured workflow definition language that pairs with Markdown documentation. Each workflow consists of:
This TOON+MD pairing approach is inspired by the BMAD METHOD pattern, adapted to use TOON instead of YAML for better orchestration support.
Workflow files are located in workflow/ directory:
Example Workflows (adb-screen-detection):
workflow/screen-verification.toon - Capture and verify screen stateworkflow/element-detection.toon - Find elements via OCR or template matchingworkflow/screen-monitoring.toon - Continuous screen monitoring and analysisExecute any workflow using the ADB workflow orchestrator:
uv run .claude/skills/adb-workflow-orchestrator/scripts/adb-run-workflow.py \
--workflow .claude/skills/adb-screen-detection/workflow/screen-verification.toon \
--param device="127.0.0.1:5555"
Each workflow includes comprehensive documentation in the corresponding .md file:
See the workflow/ directory for complete TOON file definitions and documentation.
To create custom workflows for this skill:
.toon file in the workflow/ directory.md file with comprehensive documentationFor more information, refer to the TOON specification and the workflow orchestrator documentation.
Version: 1.0.0 Status: ✅ Foundation Tier Scripts: 4 (all MCP-ready) Last Updated: 2025-12-01 Tier: 2 (Foundation)