qwen-image

What This Skill Does

The Qwen Image Skill is a specialized OpenClaw agent tool designed to bridge the gap between user intent and the Alibaba Cloud DashScope (Model Studio) image generation and editing ecosystem. It provides a standardized interface for interacting with the advanced Qwen-Image 2.0 series and the Wan 2.6 model family. By leveraging the uv toolchain for dependency management, it ensures that developers and users can execute complex image synthesis tasks—such as text-to-image generation, targeted image editing, and image interleaving—with minimal environmental friction. The skill dynamically manages API configurations based on regional requirements (CN, Intl, and US), ensuring that authentication tokens are correctly routed to the appropriate platform infrastructure.

Installation

To install this skill, use the ClawHub command: clawhub install openclaw/skills/skills/awsl1110/aliyun-qwen-image. Ensure you have uv installed (pip install uv) to automatically handle the required dependencies including dashscope and requests. Once installed, locate your script path, typically at ~/.claude/skills/qwen-image/scripts/run.py. Ensure your DASHSCOPE_API_KEY is exported in your environment variables before running the command, matching your specific regional deployment.

Use Cases

• Professional Creative Content: Generating high-fidelity visuals for marketing materials or concept art using the Qwen-Image 2.0-Pro flagship model.
• Image Refinement: Modifying existing assets through the edit subcommands, allowing for precise adjustments and consistency in output quality.
• Advanced AI Interaction: Utilizing the Wan 2.6 engine for complex text-to-image workflows that require high levels of semantic adherence and structural intelligence.
• Global Deployment: Supporting distributed teams by allowing users to toggle between Chinese Mainland, International (Singapore), and US (Virginia) backends for optimal latency and compliance.

Example Prompts

1. "使用qwen-image生成一张冬日雪景中的古典中式庭院图片，要求使用2048*2048分辨率。"
2. "帮我用wan2.6-image编辑这张图片，把背景中的树换成樱花树。"
3. "请调用百炼的千问模型，生成一个赛博朋克风格的城市街道场景。"

Tips & Limitations

• Regional Integrity: Always clarify your region (CN vs Intl vs US) with the user before executing commands. Authentication will fail if you mismatch the key with the regional endpoint.
• Model Selection: Use the Pro models for complex文字渲染 (text rendering) and high-fidelity textures. Use Max models for geometric reasoning and industrial design tasks.
• Dependencies: The skill relies on PEP 723; avoid manual virtual environment management when uv run is available, as it handles isolated execution natively.
• Wan 2.6 Specifics: Be aware that Wan 2.6 interleaving mode is limited to a single input image, and mixing images with the t2i mode will result in errors.