image-reader

Image recognition and understanding tool. Uses a multimodal model (e.g. doubao-seed-2.0-pro, kimi-k2.5) to analyze image content and supports OCR text extraction and image description. Use this skill when a user sends a screenshot or image and needs the text extracted or the image content understood.

simonjoe246@simonjoe246

Install

openclaw skills install @simonjoe246/image-reader

Image Reader Skill

Image recognition and understanding tool that leverages Doubao multimodal models to analyze image content.

Features

Text Extraction (OCR): Extract text from images, suitable for documents, screenshots, posters, menus, etc.
Image Description: Generate detailed descriptions of images, suitable for photos, illustrations, memes, UI screens, etc.
General Analysis: Automatically choose the best analysis strategy based on the image type.

API Configuration

Item	Value
API Endpoint	`https://ark.cn-beijing.volces.com/api/coding/v3`
Model	`doubao-seed-2.0-pro`
Authentication	API Key (configured in config.yaml)

Usage

Command Line

bash

# General analysis
python image_reader.py /path/to/image.png

# Extract text (OCR)
python image_reader.py /path/to/image.png -p "Extract all text from the image"

# Describe the image
python image_reader.py /path/to/image.png -p "Describe this image in detail"

OpenClaw Skill Invocation

Once installed, you can invoke it using natural language:

yaml

Analyze this image
Extract the text from the image
Describe this screenshot

Output

Text-heavy images: Returns all extracted text, preserving original formatting.
Non-text images: Returns a detailed scene description, including objects, people, colors, style, etc.
Mixed content: Provides both text extraction and a visual description.

Technical Details

Uses an OpenAI-compatible API to call Doubao multimodal models
Images are sent as base64-encoded data
The system prompt adapts to the image type to select the most appropriate analysis strategy