---

name: image-generation description: Image generation skill based on Alibaba Cloud DashScope, supporting the creation of high-quality hand-drawn or standard images from user descriptions. version: 1.0.0 license: MIT

Image Generation Skill

This skill allows agents to automatically generate high-quality images (defaulting to hand-drawn style) based on user intent.

Core Features

• Smart Prompt Optimization: Transforms simple user intent into detailed hand-drawn style prompts.
• Fast Generation: Uses non-streaming interfaces to significantly speed up image generation.
• Auto-Save: Automatically downloads generated images locally and saves metadata and API responses simultaneously.

Prerequisites

Before using this skill, ensure that the DASHSCOPE_API_KEY environment variable is set:

Copyexport DASHSCOPE_API_KEY="Your API Key"

User Guide

1. Refine the Prompt

You need to refine the user's original intent into a prompt suitable for image generation. For hand-drawn versions of architecture or flowcharts, it's recommended to include keywords like "hand-drawn", "sketch", "architectural drawing", etc.

2. Run the Script

Use the following command to call the generation script:

Copynode skills/image-generate/scripts/generate_image.js "Your detailed prompt"

3. View Results

After the script completes, it will generate the following files in the current directory:

• image_YYYY-MM-DDTHH-mm-ss.png: The generated image file.
• metadata_YYYY-MM-DDTHH-mm-ss.json: Metadata including prompt, file size, and duration.
• response_YYYY-MM-DDTHH-mm-ss.json: Raw API response data (for debugging).

Example

User Intent: "Help me draw an architecture diagram of an AI coding assistant."

Recommended Prompt: "A detailed hand-drawn architectural diagram of an AI coding assistant, showing the interaction between the user, the IDE, and the LLM, technical sketch style, clean lines, white background."

Execution Command:

Copynode skills/image-generate/scripts/generate_image.js "A detailed hand-drawn architectural diagram of an AI coding assistant, showing the interaction between the user, the IDE, and the LLM, technical sketch style, clean lines, white background."