← 返回 Skills 市场
Send Video Message
作者
Yining Shi
· GitHub ↗
· v1.0.2
· MIT-0
181
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install send-video-message
功能描述
Send the user a video message with an AI avatar that speaks any text, using Runway Character API. For async updates, explanations, or anything better said th...
安全使用建议
This skill appears to do what it says: it uses your RUNWAY_API_SECRET to call Runway APIs to create avatars, generate audio, and produce avatar videos. Before installing: 1) Be willing to share any text, image URLs, or generated-image prompts you pass to Runway (they go to api.dev.runwayml.com). 2) Ensure you trust Runway and that your RUNWAY_API_SECRET has appropriate scope / rotation policy. 3) The brew install only provides the 'uv' CLI; you'll still need Python (>=3.10) plus the runwayml and httpx packages and, if you use --vertical or --square, ffmpeg on PATH. 4) The skill saves an avatar ID to ~/.openclaw/runway-avatar.json — if you want to remove that later, delete that file and/or any avatars you created via the Runway API. 5) Note small metadata omissions: the scripts honor several optional env vars (RUNWAY_AVATAR_ID, RUNWAY_AVATAR_PRESET, RUNWAY_VOICE_PRESET, SEND_VIDEO_MESSAGE_VERTICAL, SEND_VIDEO_MESSAGE_SQUARE) which are not enumerated in the registry metadata; none are credentials but be aware they can influence behavior.
功能分析
Type: OpenClaw Skill
Name: send-video-message
Version: 1.0.2
The skill is a legitimate integration with the Runway Character API for generating AI avatar videos. It uses standard Python libraries (httpx, runwayml) and subprocess calls to ffmpeg for video cropping. Configuration is stored locally in ~/.openclaw/runway-avatar.json, and temporary files are handled in /tmp. No evidence of data exfiltration, malicious execution, or prompt injection attacks was found.
能力评估
Purpose & Capability
Name/description (send a Runway avatar video) align with the code and runtime instructions. The skill legitimately needs a Runway API key and the uv runner. Default avatar image URLs point to runway-owned S3 assets and the scripts call Runway endpoints.
Instruction Scope
SKILL.md and the scripts confine activity to creating avatars, TTS, requesting avatar_videos, downloading the resulting MP4, and saving avatar metadata to ~/.openclaw/runway-avatar.json. The code will send only the text, optional description/image URL, and any generated image to the Runway API as expected. Note: SKILL.md mentions MEDIA: attachment behavior (OpenClaw platform integration), but the included scripts themselves only save files to /tmp and print paths — attaching/uploading to other services is not performed by these scripts.
Install Mechanism
Install spec only installs the 'uv' brew formula (reasonable because scripts are invoked via 'uv run'). There are no downloads from untrusted hosts. However, Python dependencies (runwayml, httpx) and optional ffmpeg are required by the scripts but not installed by the install spec or documented as automatic installs; users must ensure Python packages and ffmpeg are available.
Credentials
Primary credential RUNWAY_API_SECRET is appropriate and necessary. The code also reads several optional environment variables that are not declared in metadata: RUNWAY_AVATAR_ID, RUNWAY_AVATAR_PRESET, RUNWAY_VOICE_PRESET, SEND_VIDEO_MESSAGE_VERTICAL, and SEND_VIDEO_MESSAGE_SQUARE. These are not secrets but you should be aware the scripts will respect those env vars if present. The skill does not request unrelated credentials.
Persistence & Privilege
always:false and no privileged/autostart behavior. The skill writes only its own config file (~/.openclaw/runway-avatar.json) and output videos to /tmp — expected one-time avatar persistence. It does not modify other skills or system-wide agent settings.
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 send-video-message - 安装完成后，直接呼叫该 Skill 的名称或使用
/send-video-message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
v1.0.2
- Added support for square crop (`--square`) for Telegram in addition to existing vertical cropping, with clear flag and environment variable usage.
- Updated cropping instructions to distinguish between Telegram (1:1) and other mobile messaging (9:16).
- Clarified that `--square` and `--vertical` are mutually exclusive options for video format.
- Improved documentation on how to prepare videos for different chat apps.
v1.0.1
- Clarified avatar setup: Default is now "Openclaw Lobster 2" with user-chosen or generated avatars as options; Runway preset avatars are only for fallback use.
- Added clear guidance to always reuse a saved avatar ID from previous runs, skipping unnecessary setup.
- Expanded setup instructions to emphasize avatar selection order and fallback procedures.
- Added instructions for generating vertical (9:16) videos for mobile/Telegram with the `--vertical` flag.
- Improved clarity and ordering of steps for avatar creation and video generation.
v1.0.0
- Initial release of the "send-video-message" skill.
- Allows users to create an AI avatar (with custom or preset image) via Runway Character API.
- Generates and sends video messages with lip-synced speech, using selectable TTS voices.
- Detailed setup instructions for avatar creation and video generation.
- Video files are saved to `/tmp` and auto-attached to chat via OpenClaw.
元数据
常见问题
Send Video Message 是什么？
Send the user a video message with an AI avatar that speaks any text, using Runway Character API. For async updates, explanations, or anything better said th... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 181 次。
如何安装 Send Video Message？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install send-video-message」即可一键安装，无需额外配置。
Send Video Message 是免费的吗？
是的，Send Video Message 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Send Video Message 支持哪些平台？
Send Video Message 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Send Video Message？
由 Yining Shi（@yining1023）开发并维护，当前版本 v1.0.2。
推荐 Skills