A powerful image recognition and analysis skill that enables AI agents to detect objects, extract text, and moderate content using Google Cloud Vision.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install google-cloud-vision
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install google-cloud-vision 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 Google Cloud Vision?
The Google Cloud Vision skill for Openclaw Skills provides a seamless bridge between your AI agent and Google's sophisticated machine learning models for image analysis. It allows developers and agents to programmatically understand image content, perform high-accuracy Optical Character Recognition (OCR), and identify labels or landmarks within a secure environment.
By leveraging the Membrane framework, this skill eliminates the complexity of manual authentication. It handles credential refreshing and OAuth flows in the background, allowing your agent to focus on processing visual data rather than managing API keys. Whether you are building a document processor or an automated content moderator, this skill provides the robust infrastructure needed for professional-grade computer vision tasks.
Google Cloud Vision Use Cases
Primary Use Cases
- Automated Document Processing: Extract dense text from scanned documents or PDFs using specialized OCR models.
- Content Moderation: Automatically detect explicit or unsafe material to maintain platform safety.
- Inventory Management: Use object detection to identify and localize products within images for retail or logistics.
- Accessibility Enhancements: Generate descriptive labels and identify faces or emotions to provide context for visually impaired users.
- Brand Monitoring: Identify company logos and brand marks across social media or web imagery.
How Google Cloud Vision Works
Operational Workflow
- Initialization: The agent initializes the Openclaw Skills environment and ensures the Membrane CLI is authenticated.
- Connection Establishment: A connection to Google Cloud Vision is ensured via a specific app URL, which manages the secure handshake.
- Intent Discovery: The agent searches for specific actions (like Detect Labels or Detect Objects) based on the user's natural language request.
- Image Processing: The agent sends image data to the selected action. Membrane handles the headers, authentication, and request formatting.
- Data Analysis: Google Cloud Vision processes the image and returns a structured JSON response containing annotations, confidence scores, and bounding boxes.
Google Cloud Vision Setup
Installation and Configuration
First, install the Membrane CLI globally to enable Openclaw Skills communication:
npm install -g @membranehq/cli@latest
Authenticate your agent with the appropriate tenant and client name:
membrane login --tenant --clientName=openclaw
Create or find a connection to the Google Cloud Vision API:
membrane connection ensure "https://cloud.google.com/vision/docs" --json
If the state is BUILDING, wait for it to become READY:
membrane connection get <id> --wait --json
Google Cloud Vision Data Schema & Taxonomy
Data Organization and Schema
The skill utilizes structured JSON outputs for all image analysis tasks. Below are the primary data structures returned by the core actions:
| Feature | Data Type | Description |
|---|---|---|
| Label Detection | EntityAnnotation[] |
List of identified objects/categories with confidence scores. |
| Text Detection | TextAnnotation[] |
OCR results including detected words and their locations. |
| Face Detection | FaceAnnotation[] |
Bounding boxes for faces, including emotional likelihoods (Joy, Sorrow, etc.). |
| Safe Search | SafeSearchAnnotation |
Likelihood ratings for adult, spoof, medical, violence, and racy content. |
| Crop Hints | CropHintsAnnotation |
Optimized aspect ratio coordinates for automated image resizing. |
Google Cloud Vision Advanced Features
Power-User Capabilities
- Proxy Requests: For edge cases not covered by pre-built actions, send raw HTTP requests directly to any Google Cloud Vision endpoint while Membrane manages the auth headers.
- Batch Processing: Utilize BatchAnnotateImages to process multiple images in a single call, optimizing token usage and latency.
- Real-time Polling: Use the --wait flag in long-running connection setups to automate agent wait-states during initial configuration.
- JSON-Native Output: Every command supports a --json flag, ensuring your AI agent can parse and act on raw data without text manipulation.
SKILL.md
Loading
Related Openclaw Skills
AI VoC Review Intelligence
A deep semantic intelligence agent that transforms raw customer reviews into actionable product-market fit signals and marketing copy.
CQ Novel Writer
A professional-grade AI writing assistant designed for continuous long-form novel creation with built-in plot and character memory systems.
Content Product Analyzer
A professional-grade analysis tool that converts raw content, product pages, and social profiles into strategic commercial teardowns and competitive insights.
Polymarket Minimal Trading Python
A streamlined Python trading client for Polymarket that enables authenticated market and limit orders via a decentralized wallet private key.
Image Deduplicator Pro
A high-performance recursive image scanner that identifies and safely moves duplicate images to the trash while preserving the highest quality versions.
Leap
Leap is an automation powerhouse that enables AI-driven media generation and workflow orchestration through a unified integration layer.
Featured*
