← 返回 Skills 市场
124
总下载
1
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install modelscope-img-generator
功能描述
Generate images with ModelScope API. Use for image generation requests. Supports text-to-image + image-to-image; configurable models; use --input-image.
安全使用建议
This skill appears to do what it says: it will upload your prompt and any input image to ModelScope's official API and save the returned image in your current working directory. Before installing, confirm you trust ModelScope with any image content or prompts you send (sensitive images may be retained by the service). Provide an API key scoped to your account, and be prepared to revoke it if misused. Note the script will download whatever image URL the API returns (typically ModelScope storage/CDN); if you need stricter guarantees about where assets are hosted, review ModelScope's service docs. Also verify you are comfortable installing/using the 'uv' runner and the Python dependencies (requests, pillow).
功能分析
Type: OpenClaw Skill
Name: modelscope-img-generator
Version: 1.2.0
The ModelScope-Img skill is a legitimate tool for generating and editing images using the ModelScope API. The implementation in `scripts/generate_img.py` follows security best practices by hardcoding the official API endpoint (https://api-inference.modelscope.cn/) to prevent API key redirection or data exfiltration to untrusted servers. The skill handles image-to-image tasks by converting local files to base64 data URLs for the official API and includes clear instructions in `SKILL.md` for the AI agent to manage filenames and prompts without any signs of malicious intent or prompt injection attacks.
能力评估
Purpose & Capability
Name/description, manifest, SKILL.md, and the Python script all align: the tool needs a MODELSCOPE_API_KEY, accepts prompts and optional input images, supports model selection and LoRA, and saves generated images locally. No unrelated credentials, binaries, or capabilities are requested.
Instruction Scope
Instructions and code stay within the stated purpose (send prompts and optional input images to ModelScope and save the returned image). Important privacy note: any provided input image and the prompt are uploaded (the script encodes input images as base64 and posts them). The script downloads the result from the URL returned by the API — SKILL.md and manifest assert the API endpoint is fixed, but the final image URL returned by the API could point to a storage/CDN host (the script will fetch whatever URL ModelScope returns). This is expected for the task but is worth the user's awareness.
Install Mechanism
There is no install spec (instruction-only with a single script). Dependencies are standard Python packages (requests, pillow) and running via 'uv' is documented. Nothing is downloaded from arbitrary or untrusted URLs and no extracted archives or install scripts are present.
Credentials
Only MODELSCOPE_API_KEY is required and it is declared as the primaryEnv in manifest. The code only reads the provided API key and does not request unrelated secrets or system credentials.
Persistence & Privilege
The skill does not request permanent/always-on inclusion and does not modify other skills or system-wide settings. It writes only the generated image file to the user's working directory (expected behavior).
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install modelscope-img-generator - 安装完成后，直接呼叫该 Skill 的名称或使用
/modelscope-img-generator触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.2.0
**Security hardening: Endpoint is now fixed to the official ModelScope API**
- The API endpoint is now locked to `https://api-inference.modelscope.cn/` and cannot be customized.
- All ability to set a custom base URL has been removed, eliminating risks of sending your API key or images to untrusted servers.
- Documentation updated to clarify endpoint restrictions and reinforce user security.
- All other usage, prerequisites, and prompt/filename conventions remain unchanged.
v1.1.0
- Adds manifest.yaml for improved skill metadata and distribution.
- SKILL.md updated with clearer prerequisites and detailed security warnings, especially regarding the --base-url parameter and image input.
- Explicit instructions for required Python version, package manager (uv), and package dependencies.
- Existing features and usage remain unchanged.
v1.0.0
- Initial release of ModelScope-Img skill for image generation and editing using ModelScope community models.
- Supports both text-to-image and image-to-image tasks, with flexible model selection via the `--model` flag.
- Implements LoRA adapter support, including multiple LoRAs and weighting.
- API key can be provided via command line or environment variable.
- Guides on filename generation, typical error handling, and secure use from any working directory.
- Provides usage examples for generating and editing images, as well as prompt templates for better results.
元数据
常见问题
modelscope_img_generator 是什么？
Generate images with ModelScope API. Use for image generation requests. Supports text-to-image + image-to-image; configurable models; use --input-image. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 124 次。
如何安装 modelscope_img_generator？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install modelscope-img-generator」即可一键安装，无需额外配置。
modelscope_img_generator 是免费的吗？
是的，modelscope_img_generator 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
modelscope_img_generator 支持哪些平台？
modelscope_img_generator 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 modelscope_img_generator？
由 ruifeng（@windandliberty）开发并维护，当前版本 v1.2.0。
推荐 Skills