Bilibili Video Summarizer

B站（bilibili）视频字幕下载与总结工具。当用户说"帮我总结这个B站视频"、"B站视频总结"、"总结b站视频"、"这个视频说了什么"、"视频内容是什么"时触发此技能。自动从B站下载字幕（支持官方字幕和AI字幕），解析为纯文本后对视频内容进行总结，支持中英文双语字幕。

Security Scan

The skill is internally consistent: it downloads Bilibili subtitles using yt-dlp and a user-provided SESSDATA cookie, parses them locally, and summarizes the text; no unexpected network endpoints, unrelated credentials, or opaque installers are present.

Purpose & Capability

The skill's name/description (download B站 subtitles and summarize) matches the supplied scripts and instructions. Required artifacts (a Bilibili session cookie in a Netscape cookie file and the yt-dlp tool) are logically needed to fetch authenticated subtitles. No unrelated credentials, binaries, or config paths are requested.

Instruction Scope

Runtime instructions ask the user to extract and save the SESSDATA cookie and then run the included scripts to download and parse subtitles. This is within scope, but it involves handling a sensitive session cookie: the skill instructs the agent to store it at ~/.config/bilibili-cookies.txt. The SKILL.md and references claim the agent can "save it automatically," but there is no provided code that writes the cookie file — that will rely on the agent or the user to perform. The scripts themselves only read the local cookie file and do not reference other system files.

Install Mechanism

There is no installer; this is instruction + two small scripts. The only external dependency mentioned is yt-dlp (recommended via pip). No archive downloads, no URLs to arbitrary executables, and no extract/execute operations from unknown hosts are present.

Credentials

The skill requests no environment variables or cloud credentials. It does require the user's SESSDATA cookie stored in ~/.config/bilibili-cookies.txt — this is sensitive but proportionate to the stated goal (authenticated subtitle downloads). The guidance claims the cookie is 'never sent elsewhere'; the code provided does not send it, but you should assume the agent (or whatever host runs this skill) could access or transmit that file unless you trust the environment.

Persistence & Privilege

always:false and no special privileges are requested. The skill writes subtitle output to /tmp by default and reads a cookie from the user's home config; it does not modify other skills' configs or request permanent platform-wide presence.

Assessment

This skill appears to do what it says, but it requires a Bilibili SESSDATA cookie (a session secret) to download authenticated subtitles. Before installing or using it: 1) Only paste or store SESSDATA on a trusted machine; the cookie grants authenticated access and should be treated like a password. 2) If possible, save the cookie manually (as described) rather than relying on an agent to "save it automatically." 3) Revoke or rotate your Bilibili session (log out / change session) after use if you are concerned. 4) Install yt-dlp from its official source (pip or package manager) to avoid tampered binaries. 5) Be aware the parsed subtitle text is printed to stdout (and therefore will be visible to the agent/LLM you use) — do not summarize videos containing sensitive personal data unless you trust the processing environment.

B站视频字幕下载与智能总结。

首次使用：配置 Cookie

B站字幕下载需要登录认证。首次使用时：

1. 读取 references/cookie-setup.md 查看获取 SESSDATA 的方法
2. 引导用户打开 B站 并登录，按 F12 → Console → 运行 console.log(document.cookie.match(/SESSDATA=([^;]+)/)?.[1])
3. 用户粘贴 SESSDATA 后，保存为 Netscape 格式 cookie 文件到 ~/.config/bilibili-cookies.txt
4. 如果用户已有 cookie 文件，直接使用

使用流程

Step 1：下载字幕

从用户提供的 B站 URL 中提取 BV 号或完整 URL，调用 scripts/download.sh：

bash <skill>/scripts/download.sh "https://www.bilibili.com/video/BV1X4wAzEEMe"

成功输出示例：

VIDEO_TITLE: 视频标题
AVAILABLE_SUBS: zh-CN srt en-US srt ai-zh srt
SUBTITLE_FILE: /tmp/bili-subtitles/视频标题.zh-CN.srt
SUCCESS

失败处理：

• ERROR: Cookie file not found → 引导用户配置 cookie（见上方首次使用）
• ERROR: No subtitles could be downloaded → 说明该视频无字幕，告知用户无法总结
• ERROR: Could not fetch video title → URL 无效，请用户确认链接

Step 2：解析字幕为纯文本

调用 scripts/parse.py：

python3 <skill>/scripts/parse.py "/tmp/bili-subtitles/视频标题.zh-CN.srt"

输出纯文本内容（约数千字）。

Step 3：总结内容

将字幕文本整理为结构化总结，格式：

## 视频标题
**字幕字数：** X,XXX 字符

### 🎯 内容主题
[一句话概括视频核心主题]

### 📝 主要内容
[按视频结构分段总结，每段2-4句话]

### 💡 关键要点
- [要点1]
- [要点2]
- [要点3]

### 📌 结论/彩蛋
[视频结尾信息，如有]

总结原则：

• 忠实于字幕原文，不添加视频未提及的内容
• 保留关键术语和数据
• 中文视频用中文总结，英文视频可用中英文混排
• 语气自然，像人工撰写

技术说明

• 字幕优先级：zh-CN → zh-Hans → zh-Hant → ai-zh → en-US
• 工具：yt-dlp（pip3 install yt-dlp --break-system-packages）
• Cookie 格式：Netscape HTTP Cookie File（~/.config/bilibili-cookies.txt）
• Cookie 有效期：约30天，过期后需重新配置
• 不适用场景：无字幕的纯音乐/舞蹈/游戏实况视频

文件结构

bilibili-video-summarizer/
├── SKILL.md                          # 本文件
├── references/
│   └── cookie-setup.md               # Cookie 配置指南
└── scripts/
    ├── download.sh                    # 字幕下载脚本
    └── parse.py                      # SRT→纯文本解析脚本