← 返回 Skills 市场

Douyin Super Agent

Name: Douyin Super Agent
Author: pry520okgpt

作者 PRY520OKGPT · GitHub ↗ · v1.1.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install douyin-super-agent

功能描述

免费抖音处理工具。视频解析、音频提取、语音识别、文案纠错，全链路自动化，零付费依赖。

使用说明 (SKILL.md)

🚀 douyin-super-agent

一句话： 丢一个抖音链接，自动给你提取文案。纯免费，零付费依赖。

安全声明

✅ 无恶意代码：不上传数据、不执行敏感命令
✅ 所有外部调用均已声明（见下方依赖表）
✅ 跨技能调用均为可选降级，缺失不影响核心功能
✅ 数据流透明：链接 → 解析 → 下载 → 音频 → 识别 → 文案 → 桌面输出

核心能力

功能	说明
抖音视频解析	标题、ID、下载链接（MCP） ✅
音频转写	任意音频 → 语音识别 ✅
三级 ASR 降级	qwen-asr（远程优先）→ whisper-medium（本地降级）→ 腾讯云（备选） ✅
文案自动纠错	擎天柱/铁疙瘩/Grok 等 10+ 常见错别字 ✅
视频下载	MCP 直链下载 ⚠️ 依赖 MCP 稳定性

完整依赖表

必选

依赖	用途	网络	凭据
Python 3.10+	运行环境	安装时	无
faster-whisper	本地 ASR	首次下载模型	无
ffmpeg	音频提取	安装时	无

可选（自动检测，缺失则降级/跳过）

依赖	用途	调用方式	降级处理
mcporter	抖音解析	子进程	提示安装
uv	qwen-asr 运行时	子进程	跳过远程，用本地
qwen-asr skill	远程 ASR	子进程脚本	本地 whisper
tencentcloud-asr	云备选	子进程脚本	报错提示
memory-manager	记忆存储	子进程脚本	静默忽略

外部二进制

工具	调用方式	用途
`mcporter`	子进程	MCP 客户端
`curl`	Python subprocess	视频直链下载
`ffmpeg`	Python subprocess	音频提取
`uv`	Python subprocess	qwen-asr 环境

数据流

抖音短链接
  ↓ mcporter parse
标题 + ID + 下载链接
  ↓ curl / mcporter download
视频文件 (mp4)
  ↓ ffmpeg
音频文件 (mp3)
  ↓ qwen-asr / whisper / tencentcloud
识别文本
  ↓ simplify_text
精简文案 → 保存桌面

文件写入

路径	内容
`~/Desktop/douyin-super-agent/`	视频/音频/文案
`~/.cache/whisper/`	whisper 模型

快速开始

安装

chmod +x setup.sh && ./setup.sh

日常使用

# 完整流程：链接 → 文案
python3 scripts/douyin.py video "https://v.douyin.com/xxx/"

# 仅音频转写
python3 scripts/douyin.py audio audio.mp3

# 查看能力统计
python3 scripts/douyin.py stats

输出文件

~/Desktop/douyin-super-agent/ 自动保存：

dy_\x3CID>.mp4 视频
dy_\x3CID>.mp3 音频
transcript_\x3CID>.txt 精简文案
result_\x3CID>.json 结构化结果

自动纠错

已知 ASR 错误自动修正：

晴天柱 → 擎天柱
铁哥 → 铁疙瘩
住进/注进 → 注入
这特曼 → 这特么
AI减4 / AI加4 → AI-FSD
零言池 → 零延迟
Grogg → Grok
几倍发凉 → 脊背发凉

技术依赖

必选： Python 3.10+, faster-whisper, ffmpeg 可选： uv (qwen-asr), mcporter/douyin-mcp (抖音解析), tencentcloud-asr, memory-manager 模型： whisper-medium (~1.5GB，首次运行自动下载，无需手动干预) 费用： 全部免费，无付费依赖

文件结构

douyin-super-agent/
├── SKILL.md           ← 技能文档（安全声明）
├── README.md          ← 详细使用指南
├── requirements.txt   ← Python 依赖
├── setup.sh           ← 一键安装脚本
└── scripts/
    └── douyin.py      ← 主程序

安全使用建议

This skill's core functionality (download → extract → ASR → correct → save) matches its description, but pay attention to the following before installing: - Contradictory security claim: SKILL.md states "no data upload", yet the code can use remote ASR (qwen-asr and Tencent Cloud). If you care about privacy, assume audio may be uploaded when remote engines are used. Prefer the local faster-whisper engine if you want to avoid network transmission. - Cross-skill calls: The script invokes other skills' scripts under ~/.openclaw/skills/ (qwen-asr, tencentcloud-asr, memory-manager) and will call mcporter/uv/curl/ffmpeg. Inspect those integrations (and any credentials/config for tencentcloud-asr) before enabling. - Local persistence and disk usage: models (~1–3GB) are downloaded to ~/.cache/whisper/ and outputs saved to ~/Desktop/douyin-super-agent/. Ensure you have disk space and are comfortable with transcriptions being stored in that folder. - Review the code of the invoked remote scripts (qwen-asr, tencentcloud-asr, memory_manager) if present, and audit any network behavior of those components. If you do not trust remote ASR, run with --engine whisper or remove/disable qwen/tencent integrations. - Run in a sandbox or non-sensitive environment first. If you need absolute assurance of no uploads, do not run remote ASR and remove/inspect any code that calls external services. If you want, I can: (a) highlight exact lines in scripts/douyin.py that call remote services and memory-manager, (b) suggest a minimal safe configuration (local-only), or (c) produce commands to run the skill in a restricted environment.

功能分析

Type: OpenClaw Skill Name: douyin-super-agent Version: 1.1.0 The douyin-super-agent skill is a legitimate tool for downloading Douyin videos and performing speech-to-text transcription. It utilizes a multi-tiered ASR approach (Qwen, Whisper, and Tencent Cloud) and handles media processing via standard tools like ffmpeg and curl. The code in scripts/douyin.py follows safe subprocess practices to prevent shell injection and restricts file operations to the user's Desktop and cache directories, aligning perfectly with its stated functionality.

能力评估

✓ Purpose & Capability

Name/description align with the code and dependencies: the code parses Douyin links, downloads videos (mcporter/curl), extracts audio (ffmpeg), runs ASR (qwen-asr remote, faster-whisper local, tencentcloud fallback), and post-processes text. Optional cross-skill calls (mcporter, qwen-asr, tencentcloud-asr, memory-manager) are coherent with multi-engine ASR/parse design.

⚠ Instruction Scope

SKILL.md and code claim "no data upload", but the code explicitly supports remote ASR (qwen-asr via 'uv run' and a tencentcloud-asr script) which will send audio to remote services, and it invokes other skill scripts under ~/.openclaw/skills/. The skill also calls a memory-manager script to 'add' content if present. These behaviors can transmit or persist user audio/text outside the local machine despite the security claim.

ℹ Install Mechanism

No registry install spec but the bundle includes a setup.sh and pip requirements. setup.sh runs pip install -r requirements.txt, may install ffmpeg via brew, and triggers a run of scripts/douyin.py stats which can download large Whisper model files (~1–3GB). No suspicious external download URLs were observed, but model downloads and pip/npm installs will write substantial data to disk.

ℹ Credentials

The skill declares no required env vars or primary credential. It optionally invokes other skills (qwen-asr, tencentcloud-asr, memory-manager) that in practice may require credentials/configuration. Those credentials are not requested here, which is reasonable for optional fallback behavior, but it means the skill may rely on other skills' credentials if present — review those integrations before use.

ℹ Persistence & Privilege

always:false and no automatic persistent agent privileges. The skill writes outputs and models to ~/Desktop/douyin-super-agent/ and ~/.cache/whisper/, and will call a memory-manager to store short snippets if available. It does not appear to modify other skills' configs, but it does invoke other skills' scripts and can cause data to be stored or transmitted by them.

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 douyin-super-agent
安装完成后，直接呼叫该 Skill 的名称或使用 /douyin-super-agent 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

v1.1.0

完整安全声明：声明所有外部依赖，可选降级，无恶意代码，纯免费

v1.0.0

首发：抖音视频解析+音频提取+语音识别+文案纠错，纯免费零付费依赖

元数据

Slug douyin-super-agent

版本 1.1.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 2

常见问题