A powerful multimodal tool for generating detailed captions and interpretations for visual media and documents via the ZhiPu GLM-V model.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install glmv-caption
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 glmv-caption 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 GLM-V Caption?
The GLM-V Caption skill is a specialized multimodal integration for Openclaw Skills that leverages the ZhiPu GLM-V model series. It allows AI agents to "see" and interpret various media types, including images, videos, and complex documents like PDFs or PowerPoints. By interfacing directly with the GLM-V API, it provides highly accurate visual descriptions, video summaries, and document insights.
This skill is designed for precision, supporting custom prompts for specific architectural analysis, textual extraction, or comparative studies across multiple files. It ensures that the AI agent provides authentic model-generated insights rather than internal estimations, maintaining technical truth and high-fidelity output for developer workflows.
GLM-V Caption Use Cases
- Describing and captioning individual or batches of images (up to 50 at once).
- Summarizing video content from URLs to understand key events or themes.
- Interpreting complex documents like PDF, DOCX, and PPTX for rapid information extraction.
- Comparing multiple images to identify differences or commonalities.
- Automating the generation of alt-text or metadata for digital asset management.
How GLM-V Caption Works
- The user provides a media input (URL or local path) and an optional custom prompt.
- The skill validates the input type (Image, Video, or File) and checks for the required ZHIPU_API_KEY.
- A request is sent to the ZhiPu GLM-V API (defaulting to glm-4.6v) using the specified parameters like temperature and max tokens.
- The model processes the visual or document data and returns a structured JSON response.
- The raw caption or summary is displayed to the user exactly as returned by the model to ensure data integrity.
GLM-V Caption Setup
To use this contribution to Openclaw Skills, you must configure your ZhiPu API key.
- Get your API Key from the Zhipu Open Platform.
- Configure the environment variable using one of these methods:
Set in your openclaw.json:
"glmv-caption": {
"enabled": true,
"env": {
"ZHIPU_API_KEY": "YOUR_API_KEY_HERE"
}
}
Or via your shell:
export ZHIPU_API_KEY="YOUR_API_KEY_HERE"
GLM-V Caption Data Schema & Taxonomy
The skill returns a structured JSON object containing the model's interpretation and metadata:
| Field | Type | Description |
|---|---|---|
success |
Boolean | Indicates if the API request was successful. |
caption |
String | The primary text output containing the description or summary. |
usage |
Object | Includes prompt_tokens, completion_tokens, and total_tokens. |
warning |
String | Present if content was flagged by safety filters. |
error |
String | Detailed error message if the request fails. |
GLM-V Caption Advanced Features
- Thinking Mode: Enable reasoning-based processing for complex visual queries using the
--thinkingflag. - Batch Processing: Support for up to 50 images or documents in a single command execution.
- Custom Prompts: Override default behavior to focus on specific details (e.g., "Describe the architectural style").
- Streaming Output: Real-time response generation for long-form summaries using the
--streamflag. - Output Persistence: Directly save the JSON results to a local file using the
--outputparameter.
SKILL.md
Loading
Related Openclaw Skills
Load Balancer Reference Skill
A comprehensive, zero-dependency reference tool for mastering load balancer concepts, patterns, and performance tuning.
Wordcraft
A comprehensive developer reference tool providing instant access to Wordcraft implementation patterns and devtool best practices.
Qiniu Cloud Upload Skill (tianshu-qiniu-upload)
A professional utility for AI agents to upload local files to Qiniu Cloud storage and return publicly accessible web URLs.
Agent Persona Analyzer
A specialized skill for analyzing and generating persistent AI agent personalities including MBTI types, gender roles, and professional inclinations.
Bozo Writer
Bozo Writer is a professional script expansion tool that converts complex AI technical topics into engaging, viral-ready short video spoken scripts.
Skill Inventory Expert
An essential framework for AI agents to discover, categorize, and report their installed capabilities using structured JSON inventory systems.
Featured*
