minimax-image

MiniMax 图片生成技能 - 支持文生图(Text-to-Image)、图生图(Image-to-Image)。支持多种宽高比(1:1/16:9/9:16/4:3/3:4)，返回 URL 或 Base64 格式，可下载保存到本地。

Security Scan

The skill's code is a straightforward MiniMax image client, but the package/registry metadata omits required environment secrets and dependency information and the skill will upload local images to an external API — this mismatch warrants caution.

Purpose & Capability

Name/description match the code: it implements text→image and image→image using a MiniMax API. Requesting an API key and region is coherent with that purpose, but the registry metadata incorrectly lists no required environment variables or primary credential, which is an inconsistency.

Instruction Scope

SKILL.md and scripts/image.py limit actions to calling the remote /v1/image_generation endpoint, downloading returned URLs, and reading local image files when doing image-to-image. Reading local files and uploading them (base64-embedding) is explicit in the instructions and code.

Install Mechanism

Instruction-only install (no install spec) is low risk. The code imports the third-party 'requests' package but the skill metadata does not declare this dependency — users must ensure 'requests' is available.

Credentials

The runtime requires MINIMAX_API_KEY (and optionally MINIMAX_REGION) and will place the key in an Authorization: Bearer header. The registry declared no required env vars/primary credential, which is misleading. Also, local images supplied for editing are base64-encoded and uploaded to the remote API; sensitive images could be exfiltrated to the service.

Persistence & Privilege

No always:true, no special persistence or modifications to other skills or system settings. The skill runs network requests and file writes only within its stated scope.

What to consider before installing

This skill appears to implement the advertised MiniMax image generation features, but the registry metadata is inconsistent: the code and SKILL.md require MINIMAX_API_KEY and MINIMAX_REGION even though the registry lists none. Before installing, verify the MiniMax API host (api.minimaxi.com / minimax.io) is legitimate for your use, provide a dedicated API key (avoid using high-privilege or long-lived secrets), and do not upload sensitive images because the skill will base64-encode and send local files to the remote service. Ensure the Python 'requests' package is installed in the runtime. If you cannot confirm the skill's source or the API endpoints, run it in an isolated environment or container, and consider reaching out to the publisher for corrected metadata (required env vars and dependencies) before use.

MiniMax Image Skill

使用 MiniMax API 进行图片生成和图片编辑。支持文生图、图生图多种宽高比，返回 URL 或 Base64 格式，可下载保存到本地。

环境配置

{
  "MINIMAX_API_KEY": "your-api-key",
  "MINIMAX_REGION": "cn" | "int"
}

• MINIMAX_API_KEY: MiniMax API 密钥
• MINIMAX_REGION: 区域设置，cn 为中国，int 为国际（默认 cn）

可用函数

generate_image(prompt, aspect_ratio, response_format)

文生图 - 根据文本描述生成图片

参数:

• prompt: 图片描述文本
• aspect_ratio: 宽高比（默认: 1:1）
  • 1:1 - 正方形
  • 16:9 - 宽屏
  • 9:16 - 竖屏
  • 4:3 - 标准
  • 3:4 - 竖版标准
• response_format: 返回格式（默认: url）
  • url - 返回图片URL
  • base64 - 返回base64编码

返回: 图片URL或base64编码

示例: generate_image("一只可爱的猫咪", "1:1", "url")

generate_image_from_image(prompt, image_file, image_url, aspect_ratio)

图生图 - 根据参考图和描述生成新图片

参数:

• prompt: 图片修改描述
• image_file: 参考图文件路径（本地文件）
• image_url: 参考图URL
• aspect_ratio: 宽高比

返回: 生成的图片URL或base64编码

示例: generate_image_from_image("把这只猫变成老虎", image_file="cat.png")

download_image(image_url, output_path)

下载图片到本地

参数:

• image_url: 图片URL
• output_path: 保存路径

返回: 保存的文件路径

save_base64_image(base64_data, output_path)

保存base64编码的图片

参数:

• base64_data: base64编码的图片数据
• output_path: 保存路径

返回: 保存的文件路径

使用示例

文生图

python scripts/image.py generate "日出时分的海边风景" -o sunset.png -r 16:9

图生图

python scripts/image.py edit "把这张照片变成油画风格" -i photo.jpg -o painting.png

使用参考图URL

python scripts/image.py edit "给这幅画加上蓝天白云" -i https://example.com/painting.jpg -o new_painting.png

返回格式说明

• response_format=url: 返回的图片URL有效期为1小时
• response_format=base64: 直接返回图片数据，可直接保存

注意事项

1. prompt 描述越详细，生成效果越好
2. 支持中文描述
3. 图生图时参考图越大越好（建议 > 512px）
4. 生成的图片版权由用户自行承担