技能详情（站内镜像，无评论）
作者：sniper-one @aqbjqtd
许可证：MIT-0
MIT-0 ·免费使用、修改和重新分发。无需归因。
版本：v2.3.0
统计：⭐ 0 · 70 · 0 current installs · 0 all-time installs
⭐ 0
安装量（当前） 0
🛡 VirusTotal ：可疑 · OpenClaw ：可疑
Package：aqbjqtd/ffmpeg-master-pro
安全扫描（ClawHub）
- VirusTotal ：可疑
- OpenClaw ：可疑
OpenClaw 评估
The skill appears to implement a plausible FFmpeg-based video-processing tool, but its declared metadata and runtime instructions mismatch what the included scripts actually require and the SKILL.md forces background execution and outbound chat notifications — review before installing.
目的
The name/description (FFmpeg video processing) matches the code and docs: the package includes many analyzers, builders, and workflows for ffmpeg/ffprobe usage. However the skill metadata declares no required binaries or credentials while the code and runtime docs clearly call external binaries (ffmpeg, ffprobe, nvidia-smi, amdgpu-info, vainfo, etc.). That metadata omission is inconsistent and misleading.
说明范围
SKILL.md enforces direct background execution of ffmpeg via exec background:true, requires immediate user-facing 'started' reply, and mandates sending a chat/Telegram notification on completion using the platform message tool (omitting channel). It also instructs copying inputs from /mnt to /tmp and cleaning /tmp afterwards. Those instructions grant the skill broad filesystem access and require it to post completion messages (including paths) …
安装机制
No external install/download steps are declared; files are bundled with the skill (many Python modules and docs). There is no suspicious external URL or archive-based install. That's lower install risk, but the presence of many executable scripts means they could be run by the agent at runtime.
证书
No environment variables or credentials are declared, which superficially looks safer — but the code writes to and reads from the user's home config (e.g., ~/.config/ffmpeg-master/presets), accesses arbitrary file paths (e.g., /mnt, /tmp), and calls system binaries. Requesting no credentials is coherent, but the combination of filesystem access, subprocesses invoking system GPU tools, and mandated chat notifications is broader than what the me…
持久
always is false and autonomous invocation is allowed (platform default). The skill will create user config/preset files under the home directory and runs long-running background ffmpeg processes per SKILL.md. That is expected for a processing tool, but background execution plus required automatic messaging increases its operational blast radius — consider restricting usage or running in an isolated environment.
安装（复制给龙虾 AI）
将下方整段复制到龙虾中文库对话中，由龙虾按 SKILL.md 完成安装。
请把本段交给龙虾中文库（龙虾 AI）执行：为本机安装 OpenClaw 技能「Ffmpeg Master Pro」。简介：FFmpeg Master Pro - 全能视频处理技能。当用户需要视频处理时使用，支持视频转换、视频压缩、视频编辑等。当用户要求视频处理, 视频转换, 视频…。
请 fetch 以下地址读取 SKILL.md 并按文档完成安装：https://raw.githubusercontent.com/openclaw/skills/refs/heads/main/skills/aqbjqtd/ffmpeg-master-pro/SKILL.md
（来源：yingzhi8.cn 技能库）SKILL.md
---
name: ffmpeg-master-pro
description: "FFmpeg Master Pro - 全能视频处理技能。当用户需要视频处理时使用，支持视频转换、视频压缩、视频编辑等。当用户要求视频处理, 视频转换, 视频压缩, FFmpeg, 视频编辑, 视频转码, 视频剪辑, 字幕处理, 视频优化, 批量视频处理, GIF转换, 视频翻转, 速度调节, 音频提取, 视频合并, 宽高比, 水印, 截图, 抽帧时使用此技能。"
metadata:
{
"openclaw":
{
"emoji": "🎬",
"requires":
{ "bins": ["ffmpeg", "ffprobe", "python3"] },
"install":
[
{
"id": "apt-ffmpeg",
"kind": "apt",
"packages": ["ffmpeg", "python3"],
"bins": ["ffmpeg", "ffprobe", "python3"],
"label": "Install FFmpeg + Python (apt)",
},
{
"id": "brew-ffmpeg",
"kind": "brew",
"formula": "ffmpeg",
"bins": ["ffmpeg", "ffprobe"],
"label": "Install FFmpeg (brew)",
},
],
},
}
---
# FFmpeg Master Pro
全能视频处理技能，提供 14 个核心工作流和智能决策系统。
## 快速开始
首次使用前，确认 FFmpeg 已安装：`ffmpeg -version`。详见 [快速入门](references/quickstart.md)。
## 核心工作流
所有工作流的完整命令、参数说明和代码示例见 [详细工作流](references/detailed_workflows.md)。
### 转码与优化
| # | 工作流 | 触发词 | 详细文档 |
|---|--------|--------|----------|
| 1 | 智能转码与压缩 | 压缩、转码、格式转换、减小文件 | [→](references/detailed_workflows.md#工作流1智能转码与压缩) |
| 2 | 智能参数优化 | 自动识别6种视频类型并优化编码 | [→](references/detailed_workflows.md#工作流2智能参数优化) |
| 3 | 预设模板系统 | YouTube/B站/微信/抖音等平台预设 | [→](references/detailed_workflows.md#工作流3预设模板系统) |
| 4 | 精确文件大小控制 | 指定目标大小，两遍编码偏差<5% | [→](references/detailed_workflows.md#工作流4精确文件大小控制) |
### 剪辑与分析
| # | 工作流 | 触发词 | 详细文档 |
|---|--------|--------|----------|
| 5 | Smart Cut 混合剪辑 | 剪辑、裁剪、cut | [→](references/smart_cut_guide.md) |
| 6 | 关键帧分析 | 关键帧、最佳剪辑点 | [→](references/keyframe_analysis.md) |
### 编辑与特效
| # | 工作流 | 触发词 | 详细文档 |
|---|--------|--------|----------|
| 7 | 字幕处理 | 提取/嵌入/烧录字幕 | [→](references/detailed_workflows.md#工作流7字幕处理) |
| 8 | 滤镜与特效 | 水印、旋转、翻转 | [→](references/detailed_workflows.md#工作流8滤镜与特效) |
| 9 | GIF 转换 | GIF、动图 | [→](references/detailed_workflows.md#工作流9gif转换) |
| 10 | 翻转与镜像 | 镜像、左右/上下翻转 | [→](references/detailed_workflows.md#工作流10翻转与镜像) |
### 变换与合并
| # | 工作流 | 触发词 | 详细文档 |
|---|--------|--------|----------|
| 11 | 速度调节 | 加速、慢动作、倍速 | [→](references/detailed_workflows.md#工作流11速度调节) |
| 12 | 音频提取与移除 | 提取音频、去音频、静音 | [→](references/detailed_workflows.md#工作流12音频提取与移除) |
| 13 | 视频合并 | 合并、拼接、concat | [→](references/detailed_workflows.md#工作流13视频合并) |
| 14 | 宽高比调整 | 16:9、9:16、竖屏、横屏 | [→](references/detailed_workflows.md#工作流14宽高比调整) |
## 智能决策系统
技能会自动执行以下决策，无需手动干预：
- **内容类型识别**：自动检测电影/动漫/屏幕录制/体育/音乐视频/老旧视频，选择最优编码参数。详见 [优化指南](references/optimization_guide.md)
- **GPU 加速**：自动检测 NVIDIA NVENC / AMD AMF / Intel QSV 并优先使用
- **质量验证**：自动校验输出文件的分辨率、时长、码率，可选 VMAF/SSIM/PSNR 评估
## 预设模板
内置 8+ 平台预设（`assets/presets/`）：youtube、bilibili、wechat、douyin、social_media、archival、preview、web_optimized。使用时提及平台名称即可自动匹配。
## 参考文档
- **详细工作流**：[detailed_workflows.md](references/detailed_workflows.md) — 所有14个工作流的完整命令和代码
- **Smart Cut 指南**：[smart_cut_guide.md](references/smart_cut_guide.md)
- **关键帧分析**：[keyframe_analysis.md](references/keyframe_analysis.md)
- **优化算法**：[optimization_guide.md](references/optimization_guide.md)
- **最佳实践**：[best_practices.md](references/best_practices.md)
- **API 参考**：[api_reference.md](references/api_reference.md)
- **快速入门**：[quickstart.md](references/quickstart.md)
- **故障排除**：[troubleshooting.md](references/troubleshooting.md)
## 工作流程
1. 分析用户需求，匹配触发词到对应工作流
2. 用 `ffprobe` 分析输入视频（格式、编码、分辨率、码率、时长）
3. 如有脚本可用（`scripts/`），优先执行脚本；否则按详细工作流文档构建 ffmpeg 命令
4. 执行转换，验证输出质量
5. 批量任务参考 detailed_workflows.md 附录中的批量处理器