Douyin Video Analyst

抖音账号视频批量采集与文案分析工作流。当用户提供抖音账号主页链接、要求抓取最新 N 条视频链接、提取视频文案（语音转文字）、或对视频内容进行总结归纳时，激活此 skill。依赖 browser 工具（抓取视频列表）和 mcporter + douyin-mcp（文案提取）。

Security Scan

The skill's declared registry metadata omits the API key requirements and its instructions tell the agent to read local mcporter config files for credentials — this mismatch and local-file access are concerning and worth review before installation.

Purpose & Capability

The skill's stated purpose (fetch Douyin video links and extract transcripts via mcporter + douyin-mcp) matches the instructions, but the registry metadata lists no required environment variables or primary credential while SKILL.md clearly depends on either DOUYIN_API_KEY or DASHSCOPE_API_KEY and on a configured douyin-mcp server. The omission in metadata is an incoherence.

Instruction Scope

Runtime instructions explicitly run a Python snippet to open and read user-local configuration files (~/.cursor/mcp.json and ~/.claude.json) to detect versions and environment keys. That means the agent will access files in the user's home directory (potentially containing API keys). Reading those configs is not explicitly declared in the registry and expands the skill's data access beyond what a simple 'fetch-and-analyze' description implies.

Install Mechanism

This is an instruction-only skill with no install spec or bundled code. The SKILL.md suggests the user install mcporter via npm if missing; the skill itself does not download or write code. That minimizes installer-related risk.

Credentials

The skill requires an API key (DOUYIN_API_KEY or DASHSCOPE_API_KEY) to call douyin-mcp, and it instructs checking local config files that may contain those keys — yet the registry lists no required env vars or primary credential. Asking the agent to read files that may contain secrets is disproportionate unless explicitly declared and authorized by the user.

Persistence & Privilege

The skill does not request always:true or modify other skills; it is user-invocable and may be invoked autonomously (normal default). However, autonomous invocation combined with the skill's ability to read local config files and use API keys increases the practical blast radius — consider that when granting runtime permissions.

What to consider before installing

Before installing or invoking this skill, consider the following: (1) The skill reads user config files (~/.cursor/mcp.json and ~/.claude.json) to find douyin-mcp configuration and API keys — review those files and remove or redact any secrets you do not want the agent to access. (2) The registry metadata does not declare the required environment variables (DOUYIN_API_KEY or DASHSCOPE_API_KEY); ask the publisher to include these in the metadata so you can make an informed consent decision. (3) Only run this skill if you trust the mcporter/douyin-mcp binaries and servers it will call; these external tools will receive your API key and may transmit data externally. (4) If you prefer safer testing, run the workflow in an isolated environment (VM/throwaway account) or use dummy API keys to validate behavior first. (5) If you are uncomfortable with the agent reading home-directory config files, modify those files to remove credentials and instead provide the minimal API key explicitly and only for the duration of the call. If you want, ask the skill author to remove automatic local-file inspection and to declare required env vars in the skill metadata.

从抖音账号主页批量采集最新视频链接，通过 douyin-mcp 提取视频语音文案，去除口语化后输出精简原始文案，并进行综合总结。

前置依赖

• browser 工具：用于渲染抖音 SPA 页面并抓取视频列表（web_fetch 无效，抖音为纯 JS 渲染）
• mcporter CLI：which mcporter 确认可用
• douyin-mcp server：在 mcporter list 中应显示 douyin-mcp (5 tools)

mcporter 会自动加载 ~/.cursor/mcp.json 和 ~/.claude.json 两个配置文件中的 MCP server。

版本与 Key 对照

版本	API 平台	环境变量
≤ 1.1.0	硅基流动	DOUYIN_API_KEY
≥ 1.2.0	阿里云百炼	DASHSCOPE_API_KEY

两个版本均受支持，根据实际配置选择对应环境变量。配置指南见 references/setup.md。

工作流步骤

Step 1：检查配置

# 确认 mcporter 和 douyin-mcp 可用
mcporter list 2>&1 | grep douyin-mcp

读取配置文件判断版本和 Key：

python3 << 'EOF'
import json, os
# mcporter 同时加载这两个配置
for path in ['~/.cursor/mcp.json', '~/.claude.json']:
    p = os.path.expanduser(path)
    if not os.path.exists(p): continue
    d = json.load(open(p))
    cfg = d.get('mcpServers', {}).get('douyin-mcp')
    if cfg:
        print(f"Found in {path}")
        print('  args:', cfg.get('args'))
        print('  env keys:', list(cfg.get('env', {}).keys()))
        break
EOF

• env 含 DOUYIN_API_KEY → v1.1.0，硅基流动
• env 含 DASHSCOPE_API_KEY → v1.2.0+，阿里云百炼

若未配置，参考 references/setup.md 完成初始化。

Step 2：用 browser 抓取视频列表

抖音是纯 JS SPA，必须用 browser 工具，不可用 web_fetch。

browser(action="open", profile="openclaw", targetUrl="<账号主页URL>")
browser(action="screenshot", ...)   # 等页面渲染，确认视频列表出现
browser(action="snapshot", ...)     # 提取 listitem 中的 /video/XXXX 链接

从 snapshot 的 aria tree 中提取 link[href=/video/XXXXXXXXXXXXXXXXX]，取前 N 条，拼接完整链接：https://www.douyin.com/video/<video_id>

Step 3：并发提取视频文案

根据 Step 1 检测到的版本，选择对应的环境变量名，并发调用所有视频（不要串行等待）：

# v1.1.0（硅基流动）
DOUYIN_API_KEY="<key>" mcporter call douyin-mcp.extract_douyin_text \
  share_link="https://www.douyin.com/video/<video_id>" 2>&1

# v1.2.0+（阿里云百炼）
DASHSCOPE_API_KEY="<key>" mcporter call douyin-mcp.extract_douyin_text \
  share_link="https://www.douyin.com/video/<video_id>" 2>&1

同时发起所有 exec 调用（放在同一个工具调用块中），设置 timeout=180，yieldMs=120000。

若出现错误，参考 references/troubleshooting.md。

Step 4：整理与输出

按视频顺序（从新到旧）输出，每条视频包含：

1. 视频标题（来自 snapshot 的 link 文本）
2. 视频链接
3. 精简文案：对原始语音文案去除口语化表达（去掉语气词、重复啰嗦、"大家"/"我们"/"你们"等口语填充词），保留核心信息和论点，以书面化段落呈现，不做摘要压缩，完整保留所有要点

最后附综合总结：跨视频归纳主线逻辑、核心主题、一致的风险提示。

输出格式参考

## 📊 [账号名] 最近 N 条视频文案整理

---

### 视频1 | <标题>
🔗 <链接>

<精简后的书面化文案，完整保留所有要点，去除口语填充>

---

### 视频2 | <标题>
...

---

## 🗂️ 综合总结

| 主线 | 核心逻辑 | 重点方向 |
|------|----------|----------|
| ... | ... | ... |

注意事项

• 抖音登录弹窗出现时：直接读取 snapshot，视频列表通常仍可在 aria tree 中获取
• 第一条视频可能需要登录才能访问，从第二条起往往正常
• extract_douyin_text 耗时较长（30~90s/条），务必并发执行
• 若文案提取失败，可先用 parse_douyin_video_info 获取下载链接，再手动处理