MiniMax MMX

MiniMax MMX-CLI 命令行工具，用于 AI Agent 调用 MiniMax 全模态能力（图片生成、视频生成、语音合成、音乐创作、图片理解、网络搜索、文本对话等）。当用户需要生成图片、视频、语音旁白、音乐，或者需要理解图片内容、执行搜索时使用此技能。安装：npm install -g mmx-cli；...

Security Scan

The skill is an instruction-only wrapper describing how to use a MiniMax CLI; its requirements and instructions are mostly consistent with that purpose, with only minor documentation gaps (no install spec and the CLI's API key usage isn't declared as an environment requirement).

Purpose & Capability

The skill describes use of an mmx CLI (install via npm is mentioned in prose) and all commands in SKILL.md are consistent with a multimodal CLI. However, the package installation is only mentioned in description text (npm install -g mmx-cli) and there is no declared install spec or required binary in the metadata. Also the instructions reference an API key-based auth flow but the skill does not declare any required environment variables for credentials. These are documentation/integration gaps rather than indicators of unrelated capabilities.

Instruction Scope

SKILL.md instructs the agent to run mmx CLI commands (image/video/music/speech/search/vision/text). Commands reference user-provided file paths for image/vision operations and an auth command to supply an API key; they do not instruct the agent to read unrelated system files, aggregate extraneous data, or send data to unexpected endpoints. Scope stays within the CLI's functionality.

Install Mechanism

There is no install spec or code included (instruction-only). The README mentions npm install -g mmx-cli, but the skill does not require or supply an install mechanism. This is not dangerous by itself but means the agent environment must already have mmx-cli available (or the operator must install it), which is a deployment consideration.

Credentials

The SKILL.md shows an API-key based auth (mmx auth login --api-key <key>), implying a credential is needed, yet the skill metadata does not list any required environment variables or a primary credential. Requesting an API key via the CLI is reasonable, but the absence of declared credential requirements is an inconsistency the user should be aware of.

Persistence & Privilege

The skill does not request always:true and has no install-time persistence or cross-skill config changes. It relies on a local CLI which may persist its own auth tokens (normal for a CLI), but the skill itself does not request elevated platform privileges.

Assessment

This skill is an instruction-only guide for using a third-party mmx CLI. Before installing or using it: 1) Confirm the mmx-cli package exists on npm and is from a trusted publisher (the skill's source/homepage is not provided). 2) Understand that the CLI requires an API key (mmx auth login) — verify where that key will be stored (local CLI config) and avoid reusing high-privilege keys. 3) The agent will need access to the mmx binary or you must install it (npm install -g mmx-cli) in your environment. 4) Commands like vision describe operate on local files you supply; avoid passing sensitive system paths. 5) If you need higher assurance, request the package repository URL or vendor documentation and inspect the mmx-cli code before installing.

MMX-CLI - MiniMax 全模态命令行工具

快速开始

# 鉴权（如未配置）
mmx auth login --api-key <your-api-key>

# 查看配额
mmx quota show

核心命令

图片生成 (image-01)

mmx image generate --prompt "描述" --aspect-ratio 16:9 --out-dir ./output --quiet --output json

• --n: 生成数量 (1-4)
• --subject-ref: 角色一致性参考

语音合成 (speech-hd)

mmx speech synthesize --text "文本" --voice "中文主播" --out audio.mp3 --quiet --output json

• 查看可用声音：mmx speech voices

音乐创作 (music-2.5)

mmx music generate --prompt "描述" --out music.mp3 --quiet --output json

视频生成 (MiniMax-Hailuo-2.3)

# 同步（等待完成）
mmx video generate --prompt "描述" --out video.mp4 --duration 6

# 异步（不阻塞）
mmx video generate --prompt "描述" --out video.mp4 --async
mmx video task get <task-id>  # 查询状态
mmx video download <task-id>   # 下载

图片理解

mmx vision describe --image path/to/image.jpg --quiet --output json

网络搜索

mmx search query --query "关键词" --quiet --output json

文本对话

mmx text chat --prompt "问题" --quiet --output json

Agent 优化参数

参数	作用
--quiet	抑制进度条、彩色字符，只输出结果
--output json	输出纯 JSON，避免解析干扰
--async	异步模式，长任务转后台执行
--non-interactive	禁用交互式提示，CI/Agent 模式

错误处理（Exit Codes）

Code	含义	处理
0	成功	-
1	通用错误	查看 error.message
401	鉴权失败	重新 mmx auth login
400	参数错误	检查命令参数
408	请求超时	重试
429	速率限制	等待后重试

详细命令参考：See references/commands.md