minimax-video

MiniMax 视频生成技能 - 支持文生视频(Text-to-Video)、图生视频(Image-to-Video)、首尾帧视频(Frame-to-Frame)、主体参考视频(Subject-to-Video)。使用模型 MiniMax-Hailuo-2.3，默认生成 5 秒 720P MP4 视频，自动下载保...

Security Scan

The skill's code and docs match a MiniMax video client and only need an API key and region, but the registry metadata omits those required environment variables and the package will call external APIs and write files — this mismatch and external-network behavior deserve caution.

Purpose & Capability

The name/description (MiniMax video generation) aligns with the included code and SKILL.md: the script posts to a MiniMax video-generation API and downloads resulting MP4s. Requiring an API key and region is reasonable for this purpose. However, the registry metadata declares no required environment variables or primary credential while SKILL.md and the script both require MINIMAX_API_KEY and optionally MINIMAX_REGION — an incoherence between manifest and implementation.

Instruction Scope

SKILL.md instructions and examples match the script: they instruct creating tasks, polling status, and downloading video files. The instructions explicitly use MINIMAX_API_KEY and MINIMAX_REGION and direct the agent to call external MiniMax endpoints and save files locally. There is no scope creep (no hidden file/system scanning or unrelated credential access) inside the provided instructions.

Install Mechanism

No install spec is provided (instruction-only with an included Python script). No external archives or third-party package downloads are executed as part of an installer, so the installation risk is low. The script uses the requests library but does not attempt to install packages itself.

Credentials

The script and SKILL.md legitimately require MINIMAX_API_KEY (and optionally MINIMAX_REGION). Those credentials are proportional to the stated functionality. The concern is the registry metadata lists no required env vars/primary credential — this mismatch could cause surprising runtime failures or indicate incomplete metadata. The skill only reads MINIMAX_API_KEY and MINIMAX_REGION; it doesn't access other secrets or system config paths.

Persistence & Privilege

The skill does not request persistent/always-on presence (always: false) and does not modify other skills or system-wide settings. It performs network calls and writes downloaded video files to disk, which is expected for its function but should be noted.

What to consider before installing

This skill appears to implement a reasonable MiniMax video client, but before installing: 1) note that the script requires MINIMAX_API_KEY and optional MINIMAX_REGION even though the registry metadata does not declare them — treat the metadata as incomplete. 2) Verify the API endpoints (api.minimaxi.com / api.minimax.io) are the legitimate service you expect. 3) Provide a scoped or disposable API key (monitor usage) rather than a long-lived high-privilege key. 4) Because the skill will make network requests and write downloaded video files, run it in an isolated environment if you have concerns and inspect network logs. 5) If you need stronger assurance, ask the publisher for provenance (homepage/repo) or request that required env vars be added to the registry metadata and for a checksum/signature for the script.

MiniMax Video Skill

使用 MiniMax API 进行视频生成和视频编辑。支持文生视频、图生视频、首尾帧视频、主体参考视频。使用模型 MiniMax-Hailuo-2.3，默认生成 5 秒 720P MP4 视频，支持自动下载到本地。

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_video(prompt, model)

文生视频 - 根据文本描述生成视频

参数:

• prompt: 视频描述文本
• model: 视频模型（默认: MiniMax-Hailuo-2.3）

返回: 任务ID列表

示例: generate_video("一只猫在草地上奔跑")

generate_video_with_image(prompt, image_url, model)

图生视频 - 根据参考图和描述生成视频

参数:

• prompt: 视频描述文本
• image_url: 参考图片URL
• model: 视频模型

返回: 任务ID列表

示例: generate_video_with_image("让这幅画动起来", "https://example.com/painting.jpg")

generate_video_with_frames(prompt, start_image_url, end_image_url, model)

首尾帧视频 - 根据起始帧和结束帧生成过渡视频

参数:

• prompt: 视频描述文本
• start_image_url: 起始帧图片URL
• end_image_url: 结束帧图片URL
• model: 视频模型

返回: 任务ID列表

示例: generate_video_with_frames("花朵绽放的过程", "seed.jpg", "flower.jpg")

generate_video_with_subject(prompt, subject_image_url, model)

主体参考视频 - 根据主体参考图和描述生成视频

参数:

• prompt: 视频描述文本
• subject_image_url: 主体参考图片URL
• model: 视频模型

返回: 任务ID列表

示例: generate_video_with_subject("让这个人物跳舞", "https://example.com/person.jpg")

query_video_task(task_id)

查询视频生成任务状态

参数:

• task_id: 任务ID

返回: 任务状态信息 {status, video_url, ...}

状态值: Pending, Processing, Success, Fail

wait_for_video(task_id, poll_interval, max_wait)

等待视频生成完成（轮询）

参数:

• task_id: 任务ID
• poll_interval: 轮询间隔秒数（默认: 10）
• max_wait: 最大等待时间（默认: 600）

返回: 最终任务状态信息

get_file_info(file_id)

获取文件信息

参数:

• file_id: 文件ID

返回: 文件信息 {file: {file_id, status, download_url}}

download_video(file_id, output_path)

下载视频到本地

参数:

• file_id: 文件ID
• output_path: 保存路径

返回: 保存的文件路径

使用示例

文生视频

python scripts/video.py generate "日出时分的海边，海浪轻轻拍打沙滩"

图生视频

python scripts/video.py from-image "让这幅风景画动起来" -i https://example.com/landscape.jpg

首尾帧视频

python scripts/video.py frames "人物从室内走到室外" -s indoor.jpg -e outdoor.jpg

主体参考视频

python scripts/video.py subject "让这个人物挥手" -i https://example.com/person.jpg

查询任务状态

python scripts/video.py query <task_id>

等待视频完成

python scripts/video.py wait <task_id> -i 10

下载视频

python scripts/video.py download <file_id> -o video.mp4

视频规格

• 默认生成 5 秒视频
• 分辨率: 1280x720 (720P)
• 格式: MP4

注意事项

1. 视频生成是异步的，需要通过 task_id 查询状态
2. 建议使用 wait_for_video 函数自动等待完成
3. 生成的视频有效期有限，请及时下载
4. 视频生成可能需要几分钟时间，请耐心等待