Hailuo Gyh

MiniMax 海螺视频生成技能。支持文生视频、图生视频、首尾帧视频、主体参考视频四种模式。使用前需设置环境变量 MINIMAX_API_KEY。

Security Scan

The skill is plausibly a video-generation wrapper for MiniMax, but multiple contradictory and unexplained items (missing declared env var, metadata claiming an embedded API key, and a different API host domain) make the package inconsistent and worth extra scrutiny before use.

Purpose & Capability

The name, SKILL.md, and the included Python script all implement video-generation calls to a MiniMax-like API — requiring python3 and the requests library is consistent with that purpose. However registry metadata did not declare the MINIMAX_API_KEY requirement while both SKILL.md and the script do, which is an inconsistency.

Instruction Scope

SKILL.md instructs the agent/user to run the included Python script with expected arguments and to set MINIMAX_API_KEY. Examples reference local image file paths (expected for image-to-video use). The instructions do not ask for unrelated files or network exfiltration beyond calling the remote video API.

Install Mechanism

There is no install spec beyond running a Python script and installing requests via pip; no external download or archive extraction is performed by the skill itself. This is low-risk from an install/execution mechanism perspective.

Credentials

The package requires an API key (MINIMAX_API_KEY) in practice, but the registry metadata lists no required env vars and _meta.json claims an 'API Key 已内置' (API key included). The script actually reads MINIMAX_API_KEY from the environment and exits if not set. Additionally, SKILL.md tells users to obtain a key from platform.minimax.com while the script calls https://api.minimaxi.com (different hostname/typo). These contradictions are unexplained and could indicate sloppy packaging or an attempt to mislead users about credential handling.

Persistence & Privilege

The skill is not marked always:true and does not request persistent system-wide changes. It does not modify other skills or agent config. Autonomous invocation is allowed (platform default) and is not itself a new risk here.

What to consider before installing

Do not provide long-lived or highly privileged API keys to this skill until you resolve the inconsistencies. Actions to consider before installing or running: - Verify the API host: the script uses https://api.minimaxi.com but SKILL.md refers to platform.minimax.com. Confirm the correct official endpoint with the service provider to avoid typo-squatted domains. - Confirm credential handling: _meta.json claims an embedded API key while the script requires MINIMAX_API_KEY; ask the author which is true. Avoid running code that claims to have an 'embedded' key unless you inspect it directly. - If you test, use a limited-scope or disposable API key and run in an isolated environment (container/VM) to limit potential misuse. - If you don't trust the author, inspect and, if necessary, modify the script to point to the known-good API endpoint or to log nothing sensitive. Prefer official SDKs or documented endpoints from the provider. - Ask the publisher to correct registry metadata to list MINIMAX_API_KEY as a required env var and to resolve the conflicting _meta.json description; lack of accurate metadata is a red flag.

MiniMax 海螺视频生成

使用 MiniMax 海螺 API 生成视频。支持 4 种模式：

1. 文生视频：根据文本描述生成视频
2. 图生视频：基于图片 + 文本描述生成视频
3. 首尾帧：首图 + 尾图 + 文本生成视频
4. 主体参考：人脸照片 + 文本，保持人物特征一致

前置要求

• Python 3
• pip3 install requests

使用方法

# 文生视频
python3 {baseDir}/scripts/video_gen.py --mode text --prompt "描述文字"

# 图生视频（推荐用于小说分镜）
python3 {baseDir}/scripts/video_gen.py --mode image --prompt "描述文字" --image "图片URL或本地路径"

# 首尾帧生成
python3 {baseDir}/scripts/video_gen.py --mode start_end --prompt "描述文字" --first "首图URL" --last "尾图URL"

# 主体参考
python3 {baseDir}/scripts/video_gen.py --mode subject --prompt "描述文字" --subject "人脸图片URL"

参数说明

参数	说明	必填
--mode	模式：text/image/start_end/subject	是
--prompt	视频描述文本（建议英文，效果更佳）	是
--image	图生视频的首帧图片 URL（image 模式）	image 模式必填
--first	首尾帧模式的首帧图片 URL	start_end 模式必填
--last	首尾帧模式的尾帧图片 URL	start_end 模式必填
--subject	主体参考模式的人脸图片 URL	subject 模式必填
--duration	视频时长：6 或 10 秒，默认 6	否
--resolution	分辨率：720P / 768P / 1080P，默认 768P	否
--output	输出文件名，默认 output.mp4	否

模型说明

• 文生视频 / 图生视频：MiniMax-Hailuo-2.3（支持 768P / 1080P）
• 首尾帧视频：MiniMax-Hailuo-02（支持 768P / 1080P）
• 主体参考：S2V-01

⚠️ 注意：龚云荷的账户不支持 1080P，建议使用 --resolution 768P（已设为默认）。

示例（小说分镜）

python3 {baseDir}/scripts/video_gen.py \
  --mode image \
  --prompt "Cinematic winter scene in rural northern China. A young mother washes vegetables in icy well water. A little girl crouches nearby. Muted earth tones, film grain, shallow depth of field, melancholic atmosphere." \
  --image "/Users/gyh/Desktop/obsidian_file/GYH_file/AI小说/情节反转小说/女性向/周桂兰.png" \
  --duration 6 \
  --resolution 768P \
  --output "/Users/gyh/Desktop/obsidian_file/GYH_file/AI小说/情节反转小说/女性向/scene1.mp4"

API 配置

使用前需设置环境变量 MINIMAX_API_KEY（从 https://platform.minimax.com 获取）。

export MINIMAX_API_KEY="你的API Key"