A sophisticated vision analysis skill that uses the GLM-4.6V model to interpret images, diagrams, and videos through simple CLI commands.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install zai-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 zai-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 zai-vision?
zai-vision is a powerful multi-modal extension for Openclaw Skills designed to give AI agents a pair of eyes. By leveraging the advanced GLM-4.6V model, this skill enables programmatic analysis of visual assets ranging from static screenshots and technical diagrams to short video clips. It is built to handle complex visual reasoning tasks that standard LLMs cannot, such as identifying UI components in a mockup or diagnosing system errors from a screen capture.
This skill is particularly valuable for developers building automated testing or documentation pipelines. By integrating zai-vision into your Openclaw Skills workspace, you gain the ability to transform raw visual data into structured insights, facilitating better decision-making for autonomous agents and streamlining technical workflows.
zai-vision Use Cases
- Automated OCR and text extraction from document scans or code screenshots.
- Technical diagram interpretation to explain system architectures or flowcharts.
- UI/UX auditing to identify design components and layout structures from mockups.
- Visual error diagnosis by analyzing screenshots of console logs or application crashes.
- Video scene understanding and activity summarization for short media clips.
How zai-vision Works
- The user or agent provides a local path to an image or video file along with a natural language prompt.
- The system invokes a Python-based script that interfaces with the Z.AI SDK using a secure API key.
- For video content, the skill processes the file frame-by-frame to ensure comprehensive scene coverage within size limits.
- The GLM-4.6V model analyzes the visual input based on the specified parameters like temperature and max tokens.
- The skill returns the analysis as either a natural language description or a structured JSON object for programmatic use.
zai-vision Setup
To get started with this vision skill, install the SDK and configure your environment:
pip install zai-sdk
export ZAI_API_KEY='your-api-key'
You can then run an analysis directly from the CLI:
python3 /root/clawd/zai-vision/scripts/vision_analyze.py screenshot.png "Describe this UI"
zai-vision Data Schema & Taxonomy
The skill processes visual inputs and generates descriptive outputs. It supports the following parameter schema for refined control:
| Parameter | Type | Default | Description |
|---|---|---|---|
--model |
String | glm-4.6v | The vision model used for inference |
--max-tokens |
Integer | 2000 | Maximum response length |
--temperature |
Float | 0.5 | Controls randomness (0.0 for factual) |
--json |
Boolean | false | Returns structured JSON output |
zai-vision Advanced Features
- Support for clawd-run integration providing automatic backups and timeout protection.
- Batch processing capabilities to analyze entire directories of images in a single loop.
- Structured JSON output mode for direct integration into data pipelines and other Openclaw Skills.
- Frame-by-frame video processing optimized for short-form technical demos and clips.
- Configurable inference settings to balance between factual accuracy and creative description.
SKILL.md
Loading
Related Openclaw Skills
VibeCoding AI Programming Workflow
A high-performance, standardized AI coding framework developed by Google to transform requirements into validated code through a rigorous multi-step lifecycle.
JD.com Product Comments API
Fetch JD.com product reviews and ratings for market research and sentiment analysis using this Openclaw Skills integration.
Kuaishou User Profile API
A specialized tool for fetching detailed Kuaishou user profiles, audience metrics, and account metadata via JustOneAPI.
PyPI Package Changelog Generator
An automated analyzer that generates structured evidence of changes between PyPI package versions to simplify upgrade risk assessments.
aiXplain Agent Builder
A specialized skill for planning and deploying aiXplain agents with built-in safety gates and read-only discovery.
LocalsBnb 路客云 MCP
A powerful MCP server integration that allows AI agents to manage hotel room availability, pricing, and orders through the LocalsBnb (Lukeyun) official API.
Featured*
