AI-driven YouTube Video Summarizer
BibiGPTAI 音视频助理
博客首页

Blog Post

BibiGPT Team

OpenClaw Skills 101: bibigpt-skill for Full-Platform Video Summary (Bilibili, YouTube, Podcasts)

Table of Contents

In 2026, OpenClaw has hit 280K+ GitHub stars — the fastest-growing open-source AI agent ever. But ask it to summarize a Bilibili video and you get:

"I'm sorry, I cannot directly access Bilibili video content."

The native summarize command only recognizes YouTube. bibigpt-skill is the essential OpenClaw Skill for full-platform video summarization. One command, 30 seconds — your Agent gains Bilibili, Xiaohongshu, Douyin, YouTube, podcasts, and local file support.

This is the complete guide to bibigpt-skill — from the Skills mechanism to advanced CLI usage. It's the methodology overview for the OpenClaw + bibigpt-skill platform-specific series.

What Are Claude Code Skills?

Claude Code Skills (also called OpenClaw Skills) are Anthropic's Agent capability extension mechanism. Think of it this way:

AnalogyExplanation
iPhone AppEach app adds new features to your phone
Claude Code SkillEach Skill adds new capabilities to your AI Agent
bibigpt-skillAdds "full-platform audio/video summarization" to your Agent

Skills are installed via npx skills add <author>/<skill-name>. Once installed, your Agent can call the bibi command natively. Key advantages:

  1. Zero-code integration — No API wiring required; the Agent understands it natively
  2. Cross-platform — Works with both OpenClaw and Claude Code
  3. Extensible — ClawHub already has hundreds of community Skills; bibigpt-skill is among the most downloaded utility Skills

bibigpt-skill CLI help screen

What Problem Does bibigpt-skill Solve?

OpenClaw's native AI has systematic gaps when it comes to Chinese video platforms:

PlatformOpenClaw Nativebibigpt-skill
YouTube⚠️ Basic (no highlights/collections)✅ Full + AI Highlight Notes
Bilibili❌ Not supported✅ Full support
Xiaohongshu❌ Not supported✅ Full support
Douyin❌ Not supported✅ Full support
Podcasts❌ Not supported✅ Full support
Local files❌ Not supported✅ Full support (MP4/MP3/M4A)

One sentence summary: bibigpt-skill is your AI Agent's passport to the Chinese video world.

30-Second Install: Zero to Full-Platform Video Summarization

Prerequisites

  1. BibiGPT Desktop (macOS / Windows) — the host for the bibi CLI tool
  2. OpenClaw or Claude Code — the Agent execution environment

Installation Steps

Step 1: Install BibiGPT Desktop

# macOS (recommended via Homebrew)
brew install --cask bibigpt

# Windows (winget)
winget install BibiGPT.BibiGPT

# Or download directly from
# https://bibigpt.co/download/desktop

Step 2: Install bibigpt-skill

npx skills add JimmyLv/bibigpt-skill

Step 3: Verify installation

bibi auth check
# Output: ✅ BibiGPT authenticated. Ready to summarize.

That's it — just three steps. After this, in Claude Code or OpenClaw you can simply say:

"Summarize this Bilibili video for me: <bilibili-url>"

ClawHub Skill page

Supported Platforms and Command Reference

After installing bibigpt-skill, your Agent supports all these platforms:

Video Platforms

# Bilibili
bibi https://www.bilibili.com/video/BVxxxxxxxx

# YouTube
bibi https://www.youtube.com/watch?v=xxxxxxxxxx

# Xiaohongshu
bibi https://www.xiaohongshu.com/explore/xxxxxxxxxx

# Douyin
bibi https://www.douyin.com/video/xxxxxxxxxx

Audio Platforms

# Xiaoyuzhou podcast (URL)
bibi https://www.xiaoyuzhoufm.com/episode/xxxxxxxxxx

# Apple Podcasts RSS
bibi https://podcasts.apple.com/...

# Local MP3/MP4/M4A files
bibi /path/to/local-audio.mp3

Advanced Flags

# Specify output language
bibi <url> --lang zh-CN

# Summary only (no transcript)
bibi <url> --summary-only

# Chapter-by-chapter output
bibi <url> --chapters

OpenClaw vs. Claude Code: Usage Differences

Both platforms can use bibigpt-skill, but the workflows differ:

OpenClaw Mode (autonomous Agent execution):

You: Summarize all of LeCun's YouTube videos from this past week
OpenClaw: Sure, let me search and summarize each one...
          [Auto-invokes bibi command, batch processing]

Claude Code Mode (conversational execution):

You: Summarize this Bilibili video: https://...
Claude Code: [Calls bibi command, displays summary result]

The key difference: OpenClaw can autonomously schedule bibigpt-skill within an Agent workflow, enabling tasks like "summarize this podcast for me every day." Claude Code is better for real-time, interactive single summarizations.

For the OpenClaw + Bilibili deep-dive workflow, see OpenClaw + Bilibili Collection Summary.

Hands-On Demo: Developer First-Hand Experience

First-hand experience review (tested by BibiGPT founder JimmyLv)

In a Claude Code environment set up with bibigpt-skill, I ran a complete test:

Scenario: Summarize the latest 3 technical lecture videos from the official OpenClaw YouTube channel into a research report

Action:

Me: Summarize the latest 3 videos from the OpenClaw channel,
    compile key points into a report
Claude Code: Sure, processing...
             [bibi https://youtube.com/watch?v=xxx1]
             [bibi https://youtube.com/watch?v=xxx2]
             [bibi https://youtube.com/watch?v=xxx3]
             
             Here's the comprehensive report:
             1. Video 1 key insights: ...
             2. Video 2 key insights: ...
             3. Video 3 key insights: ...

Result: From entering the URLs to getting a complete report — under 3 minutes. Compared to manual work (opening each video, taking notes by hand), this saved roughly 40 minutes.

Key stats: BibiGPT supports videos up to 4 hours long. Subtitle extraction success rate for Bilibili 4K videos is over 98%.

GitHub Skill installation page

Advanced bibi CLI Usage

For developers who live in the terminal, the bibi command itself is a powerful standalone tool:

# View all supported commands
bibi --help

# Check authentication status
bibi auth check

# Output summary directly in terminal (no GUI)
bibi <url> --output terminal

# Save as Markdown file
bibi <url> --output markdown --save ./notes/

# Pipe output (use with other commands)
bibi <url> --output json | jq '.summary'

This makes bibigpt-skill not just an Agent tool, but a powerful CLI utility for power users.

FAQ

Q1: What's the difference between bibigpt-skill and BibiGPT's web version?

A: The features are identical. The difference is how you use them: the web/desktop version uses a GUI; bibigpt-skill lets AI Agents call the same backend capabilities via command line. Essentially, the bibi command calls BibiGPT's local desktop API.

Q2: Do I need a BibiGPT subscription?

A: Basic features (summarization, subtitle extraction) are free. Advanced features like AI Highlight Notes and Collection Summary require a Pro subscription. The free tier has a daily usage limit.

Q3: Why doesn't OpenClaw natively support Bilibili?

A: OpenClaw is designed as a general-purpose AI Agent framework, not deeply adapted for specific video platforms. Bilibili requires specialized subtitle API handling (covering three sources: danmaku, CC subtitles, and AI-generated subtitles) — this is exactly the 6 years of specialized capability BibiGPT has built up.

Q4: Is multi-language summarization supported?

A: Yes. bibi <url> --lang en can summarize a Chinese video in English, and vice versa. Ideal for researchers who need cross-language knowledge integration.

Q5: Does it work on Windows?

A: Yes. BibiGPT Desktop supports both macOS and Windows (install via winget). The bibi command works on both platforms.

Series Navigation

This is the methodology overview for the "OpenClaw × bibigpt-skill" series. Platform-specific deep-dive articles:

Start your AI Agent audio/video summarization journey now:

BibiGPT Team