skill-pack

A curated collection of skills for AI coding assistants, compatible with Claude Code, Codex CLI, OpenCode, and OpenClaw.

中文文档

What to Use First

Start with link-to-note. It is the recommended URL-to-note workflow for YouTube, Bilibili, Apple Podcasts, Xiaoyuzhou, and other compatible audio links.

video-to-note is kept for backward compatibility with the old video-only workflow. Prefer link-to-note for new usage.

Skills

Quick Start

curl -fsSL https://raw.githubusercontent.com/LjyYano/skill-pack/main/install.sh | bash

Restart your AI coding assistant, then run one of the installed skill commands:

/link-to-note https://www.youtube.com/watch?v=xxx
/link-to-note https://www.bilibili.com/video/BV1rxqmBhE91/
/link-to-note https://podcasts.apple.com/cn/podcast/.../id1552904790?i=1000755467027
/link-to-html AI/Podcasts/example-note.md
/article-to-anki https://example.com/article

For Obsidian notes, run your assistant from the root of your Obsidian vault when possible. The skills write files relative to the current working directory, so running from the vault root lets Obsidian index the generated AI/... files automatically.

Installation

One-click install

Auto-detect installed tools and install into their skill directories:

curl -fsSL https://raw.githubusercontent.com/LjyYano/skill-pack/main/install.sh | bash

Install for a specific tool

curl -fsSL https://raw.githubusercontent.com/LjyYano/skill-pack/main/install.sh | bash -s -- 

Manual install

Clone this repository, then copy the skills you want into your assistant’s skills directory:

SKILLS_DIR="$HOME/.codex/skills"  # Change this for Claude Code, OpenCode, or OpenClaw.
mkdir -p "$SKILLS_DIR"
cp -r skills/link-to-note    "$SKILLS_DIR"/
cp -r skills/link-to-html    "$SKILLS_DIR"/
cp -r skills/article-to-anki "$SKILLS_DIR"/

Common skill directories:

Copy skills/video-to-note too only if you need the legacy workflow.

Requirements

ASR configuration

link-to-note uses video subtitles whenever they are available. When a video has no subtitles, or when the input is a podcast / audio URL, it uses Alibaba Cloud DashScope paraformer-v2 for async speech recognition.

Get an API key from the Alibaba Cloud DashScope Console, then configure it for your OS and shell.

macOS / Linux:

# Current terminal only
export ALIYUN_API_KEY="sk-xxxxxxxxxxxxxxxxxxxx"

# Persistent for zsh, the default shell on recent macOS versions
echo 'export ALIYUN_API_KEY="sk-xxxxxxxxxxxxxxxxxxxx"' >> ~/.zshrc
source ~/.zshrc

# Persistent for bash
echo 'export ALIYUN_API_KEY="sk-xxxxxxxxxxxxxxxxxxxx"' >> ~/.bashrc
source ~/.bashrc

If your shell reads another startup file, such as ~/.bash_profile, add the same export ALIYUN_API_KEY=... line there instead.

Windows PowerShell:

# Current PowerShell session only
$env:ALIYUN_API_KEY = "sk-xxxxxxxxxxxxxxxxxxxx"

# Persistent for the current Windows user
[Environment]::SetEnvironmentVariable("ALIYUN_API_KEY", "sk-xxxxxxxxxxxxxxxxxxxx", "User")

Open a new terminal and restart your AI coding assistant after setting the environment variable.

Install yt-dlp if you use YouTube, podcasts, or generic audio URLs:

# macOS
brew install yt-dlp
yt-dlp --version

# Windows PowerShell
py -m pip install -U yt-dlp
yt-dlp --version

Install the Python dependency if it is missing:

# macOS / Linux
python3 -m pip install requests

# Windows PowerShell
py -m pip install requests

link-to-note does not require ffmpeg; paraformer-v2 processes the full audio file directly.

Examples

link-to-note

/link-to-note https://podcasts.apple.com/cn/podcast/.../id1552904790?i=1000755467027
/link-to-note https://www.xiaoyuzhoufm.com/episode/xxx
/link-to-note https://www.youtube.com/watch?v=xxx
/link-to-note https://www.bilibili.com/video/BV1rxqmBhE91/

link-to-note detects the platform, fetches transcript data from subtitles when possible, falls back to ASR when needed, and writes a structured Obsidian note with summary, takeaways, mindmap, chapters, highlights, detailed points, personal thoughts, and full transcript.

link-to-html

/link-to-html https://podcasts.apple.com/cn/podcast/%E5%95%86%E4%B8%9A%E5%B0%8F%E6%A0%B735-%E9%9C%8D%E5%B0%94%E6%9C%A8%E5%85%B9%E6%B5%B7%E5%B3%A1%E4%B8%8A%E7%9A%84%E7%89%B9%E6%AE%8A%E4%BF%9D%E9%99%A9/id1552904790?i=1000755467027
/link-to-html AI/Podcasts/example-note.md

Paste a podcast / video URL, or pass an existing .md note generated by link-to-note. The skill creates a standalone HTML viewer with sidebar navigation, 6 tabs, dark/light theme, transcript search, highlights, shownotes, and an interactive markmap mindmap.

article-to-anki

/article-to-anki https://mp.weixin.qq.com/s/Ld_NbZZaYd2z9qpfMxP_aQ

Paste an article URL or local article path. The skill extracts the content, splits it into independent knowledge cards, and writes import-ready Markdown Anki cards.

Notes

• Generated files are written under the current working directory.
• link-to-note does not add speaker labels; transcript lines use timestamp plus text only.
• link-to-note uses Bilibili REST APIs instead of yt-dlp, because current yt-dlp Bilibili extraction may return HTTP 412.
• link-to-html outputs an .html file next to the source .md file.