image-generate

name	image-generate
description	Use when a user explicitly asks to generate, draw, create, or illustrate an image from a text prompt.

Image Generate

Overview

Generate an image with DashScope using qwen-image-2.0.

When to Use

• The user explicitly asks for a new image, picture, illustration, drawing, poster, or similar visual output.
• The user provides a prompt describing what the image should contain.

Do not use this skill for ordinary text replies or for image analysis.

Configuration

• OPENCLAW_DASHSCOPE_IMAGE_API_KEY
• OPENCLAW_DASHSCOPE_API_KEY fallback
• OPENCLAW_DASHSCOPE_IMAGE_BASE_URL optional override
• OPENCLAW_DASHSCOPE_IMAGE_MODEL optional override, defaults to qwen-image-2.0
• OPENCLAW_IMAGE_OUTPUT_DIR optional override, defaults to ~/.openclaw/workspace/data/images

Use

1. Confirm the exact image prompt if the request is underspecified.
2. Run:
   • node generate.mjs --prompt "a watercolor mountain village at sunrise" --json
3. Read downloadUrl from the JSON result when it is present. Only fall back to downloadPath or outputPath if downloadUrl is empty.
4. If the current channel supports image attachments, attach the saved image in the reply.
5. If attachment sending is unavailable, prefer returning the public download URL such as https://claw.hatch.yinxiang.com/files/data/images/example.png. Only return the saved local path when no public /files/... URL can be constructed.

Notes

• Keep this skill opt-in only; use it only when the user explicitly asks for image generation.
• The helper saves files into persistent workspace-backed storage so they survive container restarts.
• Keep the reply focused on the generated image and the download URL.