MiniMax Media

MiniMax media skill for voice, image, video, and music generation. Use when the user asks for MiniMax, TTS, text-to-speech, voice, image generation, image-to...

Security Scan

The skill's instructions expect a local CLI and API credentials that are not declared or bundled with the skill, creating a mismatch and potential risk — verify the script and endpoints before installation.

Purpose & Capability

The SKILL.md describes a media-generation CLI that uses MINIMAX_API_KEY and MINIMAX_API_HOST, which is consistent with a media service. However the registry metadata declares no required environment variables and there is no bundled code or install that provides the referenced scripts/minimax. This mismatch makes it unclear how the skill would function and why the metadata omits the API credential it appears to need.

Instruction Scope

The runtime instructions tell the agent to invoke a local executable path (scripts/minimax) and to read/write files (input images, --output paths). Because the skill includes no script or install, invoking the command would rely on an existing binary in the agent's environment; that gives the agent broad filesystem and network capability depending on what that binary does. The instructions do not themselves solicit unrelated secrets, but they implicitly assume access to API credentials and will read/write arbitrary paths specified by the user.

Install Mechanism

There is no install spec and no code files — lowest install risk. However without an install step the SKILL.md's required CLI is not provided, making the skill non-self-contained. This is potentially sloppy or incomplete design rather than direct maliciousness, but it reduces transparency: users cannot inspect an included script because none is bundled.

Credentials

The SKILL.md documents environment variables (MINIMAX_API_KEY, MINIMAX_API_HOST, and model identifiers). Those variables are reasonable for a media-generation integration, but the package metadata declares no required env vars and no primary credential. The discrepancy between documentation and declared requirements is concerning and should be resolved before trusting secrets to the skill.

Persistence & Privilege

The skill does not request permanent 'always' inclusion and declares no config paths or elevated privileges. Autonomous invocation is allowed (platform default), but that alone is not a red flag here.

What to consider before installing

This skill's README instructs the agent to run a local CLI (scripts/minimax) and to use an API key/host, but the package contains no script, no install instructions, and the registry metadata does not declare the API credential it documents. Before installing or providing secrets: 1) Ask the publisher for the scripts/minimax source or an install package you can inspect. 2) Verify MINIMAX_API_HOST is a legitimate, documented endpoint (avoid personal servers or unknown domains). 3) Do not supply MINIMAX_API_KEY until you can review the client code or a trusted installer. 4) If you must test, run the CLI in a sandboxed environment and inspect network calls (e.g., with an outbound firewall or proxy) and the script's source. 5) Prefer a skill that either bundles its client code or declares its required env vars/primary credential in metadata and provides a verifiable install mechanism. If the publisher cannot provide code or a trusted install, treat this skill as incomplete and avoid giving it secrets.

统一入口：scripts/minimax

能力

• tts：语音合成
• image：文生图 / 图生图
• video：文生视频 / 图生视频
• music：文生音乐

环境变量

MINIMAX_API_KEY=...
MINIMAX_API_HOST=https://api.minimaxi.com
MINIMAX_TTS_MODEL=speech-2.8-hd
MINIMAX_TTS_VOICE_ID=English_expressive_narrator
MINIMAX_IMAGE_MODEL=image-01
MINIMAX_VIDEO_MODEL=MiniMax-Hailuo-2.3
MINIMAX_MUSIC_MODEL=music-2.5+

快速用法

scripts/minimax tts "突突你好啊，我是柱子。" --output /tmp/voice.mp3
scripts/minimax image "黑白花狸花猫在电脑前敲代码" --output /tmp/cat.jpg
scripts/minimax image "把这张图改成夜景霓虹风" --input-image ./ref.jpg --output /tmp/i2i.jpg
scripts/minimax video "黑白花狸花猫在夜晚的办公室里写代码" --output /tmp/video.mp4
scripts/minimax video "把静态图做成轻微推进镜头" --input-image ./ref.jpg --output /tmp/i2v.mp4
scripts/minimax music "lofi chill, warm piano, night city vibe" "歌词"

规则

• 默认优先使用 MINIMAX_API_HOST=https://api.minimaxi.com
• 图 / 视频支持参考图输入；不记字段名，直接走 scripts/minimax
• 任务结果文件会写到当前工作目录或 --output 指定路径
• 需要排查时，可加 --print-json
• 遇到额度不足、鉴权失败、超时、返回结构异常，会直接中文报错，不会让人干等

你该怎么用

当用户要：

• 生成语音
• 生成图片
• 做图生图
• 生成视频
• 做图生视频
• 生成音乐

直接调用 scripts/minimax，不要让人类记那些 API 细节。