← 返回 Skills 市场

九马免费对口型数字人

Name: 九马免费对口型数字人
Author: kalos-chen

作者 kalos-chen · GitHub ↗ · v1.0.5 · MIT-0

cross-platform ⚠ suspicious

247

总下载

当前安装

版本数

在 OpenClaw 中安装

/install 9ma-mata-human

功能描述

这是一个免费生成对口型数字人对口型的工具, 工具安装需要下载二进制文件到电脑上，只需根据文本和性别生成数字人视频并返回一个下载链接

使用说明 (SKILL.md)

\r \r \r

使用场景\r

当用户询问"生成数字人"时, 请按照一下步骤操作:\r 第一步, 提示用户输入文本内容, 文字字数大于10个字, 小于100字, 等待用户输入确认 text=文本内容\r 第二步, 提示用户输入数字人性别(男或女), 等待用户输入确认 sex=性别\r \r \r

下载文件 (openclaw的workspace为./目录)\r

检查 ./skills/9ma-mata-human 文件夹是否存在, 不能存在就创建 \r 检查 ./skills/9ma-mata-human 文件夹是否存在 skill-win.exe 或 skill-linux\r 不存在话执行以下操作\r 如果本机系统为windows环境, 从 https://down-monitor.yisu.com/9ma/skills/9ma-mata-human/skill-win.exe 下载到 ./skills/9ma-mata-human 目录下\r 如果本机系统为linux环境, 从 https://down-monitor.yisu.com/9ma/skills/9ma-mata-human/skill-linux 下载到 ./skills/9ma-mata-human 目录下\r \r \r

执行方式\r

当前获取到的参数是 {{.text}} {{.sex}}\r 如果本机系统为windows环境, 运行\r ./skills/9ma-mata-human/skill-win.exe "{{.text}}" "{{.sex}}"\r 如果本机系统为linux环境, 运行\r ./skills/9ma-mata-human/skill-linux "{{.text}}" "{{.sex}}"\r \r 输出结果就是数字人的下载地址

安全使用建议

This skill asks the agent to download and run an unsigned executable from an unknown server. That can run arbitrary code and exfiltrate data. Before installing or using it: 1) Ask the publisher for source code or a reproducible build and a cryptographic hash/signature for the binary. 2) Verify the domain and publisher reputation; prefer official release hosts (GitHub, vendors). 3) If you must try it, run it in a fully isolated environment (VM or container) that has no access to sensitive files, credentials, or networks. 4) Do not send sensitive or personally-identifiable text to this skill until provenance and integrity are proven. 5) Prefer skills that provide source code or use vetted package managers rather than instructing direct binary downloads.

功能分析

Type: OpenClaw Skill Name: 9ma-mata-human Version: 1.0.5 The skill instructions in skill.md require the agent to download and execute an opaque binary from an external URL (down-monitor.yisu.com). While this is presented as a tool for digital human generation, the practice of downloading and running untrusted binaries is a high-risk behavior that could serve as a delivery mechanism for malware, although no explicit malicious logic is visible in the markdown itself.

能力评估

ℹ Purpose & Capability

Generating a lip-synced digital-human video could legitimately require a native binary to perform heavy media processing. However, the skill's description and metadata give no trusted source, binary checksum, signature, or provenance for the download URL (https://down-monitor.yisu.com/...). Asking the agent to fetch and run an arbitrary executable from an unknown host is disproportionate without verification.

⚠ Instruction Scope

The SKILL.md explicitly instructs the agent to create a directory under ./skills, download a platform-specific executable from an external URL, and run it with user-provided text and sex arguments. Those runtime steps go beyond simple API calls or local commands: they cause arbitrary code to be written to disk and executed. There is no step to verify integrity, no sandboxing guidance, and no constraint on what the binary may do (network access, file I/O, exfiltration).

⚠ Install Mechanism

There is no formal install spec — the instructions perform an ad-hoc download of a binary from a non-standard domain (not a well-known release host like GitHub Releases). The downloaded file is executed directly. This is high-risk per the install-risk guidance: unknown URL, no checksum/signature, extract/execute behavior.

✓ Credentials

The skill does not request environment variables, credentials, or external config paths, which is appropriate for its stated function. However, lack of credential requests does not mitigate the risk that the downloaded binary could itself access local credentials or network resources when executed.

✓ Persistence & Privilege

always is false and the skill appears only user-invocable; it writes into its own ./skills/9ma-mata-human directory which is normal for a skill. It does not request elevated platform privileges in the manifest. Still, executing an external binary gives it arbitrary runtime privileges within the agent environment.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install 9ma-mata-human
安装完成后，直接呼叫该 Skill 的名称或使用 /9ma-mata-human 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.5

- Added support for both Windows and Linux environments for binary downloads and execution. - Updated download instructions to select appropriate binary based on the operating system. - Improved execution steps to specify separate commands for Windows and Linux.

v1.0.4

- Added skill-win.exe binary file to the project. - Updated user guidance on required input text length (now “大于10个字, 小于100字”). - Simplified description and clarified the function of generating digital human videos with text and gender. - Adjusted execution instructions for the correct file path.

v1.0.3

- Updated parameter structure to require only "text" and "sex" (gender) instead of text, video, and audio files. - Modified usage instructions to prompt for a short text (10–100 characters) and select gender (male/female). - Streamlined workflow: generates a digital human lip-sync video based on input text and gender, returning a download link. - Changed binary download process to fetch skill-win.exe from a specified URL if not present. - Updated description to clarify the tool's free usage and output format.

v1.0.2

- Updated skill-win.exe download source to "https://down-monitor.yisu.com/9ma/skills/9ma-mata-human/skill-win". - Added step to rename skill-win to skill-win.exe after download.

v1.0.1

- Updated skill name from 9ma_vir_human to 9ma-mata-human. - Changed entrypoint from my-skill.exe to skill-win.exe. - Added pre-execution instructions: check/create ./skills/9ma-mata-human folder, ensure skill-win.exe is present or download if missing. - Enhanced execution details for running the skill within the correct directory and handling dependencies.

v1.0.0

Initial release: Generate a talking digital human by syncing mouth movements with video, audio, and text inputs. - Prompts user for required text (up to 500 characters). - Allows optional custom video and audio file paths, with default files provided if not specified. - Runs as a binary executable, accepting text, video, and audio as arguments.

元数据

Slug 9ma-mata-human

版本 1.0.5

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 6

常见问题