小红书视频下载器

SKILL.md

---
name: xiaohongshu-downloader
description: Download and summarize Xiaohongshu (小红书/RedNote) videos. Produces a full resource pack with video, audio, subtitles, transcript, and AI summary. This skill should be used when the user asks to "download xiaohongshu video", "下载小红书视频", "save rednote video", "download from xiaohongshu", "小红书视频下载", "总结小红书视频", "summarize xiaohongshu video", or mentions downloading/summarizing content from xiaohongshu.com or xhslink.com.
version: 2.0.0
---

# Xiaohongshu Video Downloader & Summarizer

Download videos from Xiaohongshu (小红书/RedNote) and optionally generate a full resource pack: video + audio + subtitles + transcript + AI summary.

## Workflow

Follow these 6 steps in order:

### Step 1: Check Dependencies

Verify required tools are installed:

```bash
yt-dlp --version
ffmpeg -version
```

Both must be available. If missing, instruct the user to install:
- `brew install yt-dlp` (macOS) or `pip install yt-dlp`
- `brew install ffmpeg` (macOS)

### Step 2: Get Video Information

```bash
/opt/homebrew/Caskroom/miniconda/base/envs/myenv/bin/python scripts/download_xiaohongshu.py "URL" --list-formats
```

This shows available formats and verifies the URL works with cookie authentication.

### Step 3: Download Video + Extract Audio

For **basic download** (video only, backward compatible with v1.0):

```bash
/opt/homebrew/Caskroom/miniconda/base/envs/myenv/bin/python scripts/download_xiaohongshu.py "URL"
```

For **full resource pack** (video + audio + subtitles + transcript):

```bash
/opt/homebrew/Caskroom/miniconda/base/envs/myenv/bin/python scripts/download_xiaohongshu.py "URL" --full
```

For **full resource pack + AI summary preparation**:

```bash
/opt/homebrew/Caskroom/miniconda/base/envs/myenv/bin/python scripts/download_xiaohongshu.py "URL" --summary
```

The `--full` flag creates a folder `~/Downloads/<video title>/` containing:
- `video.mp4` — original video
- `audio.mp3` — extracted audio
- `subtitle.vtt` — WebVTT subtitles (via 3-tier strategy)
- `transcript.txt` — plain text transcription

The `--summary` flag implies `--full` and additionally saves `.meta.json` for AI summary generation.

### Step 4: Subtitle Acquisition (Automatic — 3-Tier Strategy)

The script automatically tries these strategies in order:
1. **Manual subtitles** — `yt-dlp --write-subs --sub-lang zh,en,zh-Hans,zh-CN`
2. **Auto-generated subtitles** — `yt-dlp --write-auto-subs`
3. **Whisper local transcription** — Falls back to `parallel_transcribe.py` using faster-whisper

### Step 5: Generate Transcript (Automatic)

The script automatically strips timestamps from VTT to produce `transcript.txt`.

### Step 6: AI Summary (Claude generates summary.md)

If the user requested a summary (via `--summary` flag or by asking to "summarize"):

1. Read the transcript file:
   ```
   ~/Downloads/<video title>/transcript.txt
   ```

2. Read the metadata file:
   ```
   ~/Downloads/<video title>/.meta.json
   ```

3. Read the summary prompt template:
   ```
   reference/summary-prompt.md
   ```

4. Replace the template placeholders with actual values:
   - `{{TITLE}}` — from .meta.json
   - `{{URL}}` — from .meta.json
   - `{{DURATION}}` — from .meta.json
   - `{{PLATFORM}}` — "Xiaohongshu (小红书)"
   - `{{TRANSCRIPT}}` — contents of transcript.txt

5. Generate the summary following the template structure.

6. Save the result to:
   ```
   ~/Downloads/<video title>/summary.md
   ```

## Options Reference

| Option | Description | Default |
|--------|-------------|---------|
| `-o, --output` | Output directory | `~/Downloads` |
| `-q, --quality` | Video quality (`best`, `1080p`, `720p`, `480p`) | `best` |
| `--browser` | Browser for cookies (`chrome`, `firefox`, `safari`, `none`) | `chrome` |
| `-a, --audio-only` | Download audio only as MP3 | `false` |
| `--list-formats` | List available formats | `false` |
| `--full` | Full resource pack mode | `false` |
| `--summary` | AI summary mode (implies `--full`) | `false` |

## Output Structure

### Basic mode (default)
```
~/Downloads/
└── <title> [<id>].mp4
```

### Full resource pack mode (`--full` or `--summary`)
```
~/Downloads/<video title>/
├── video.mp4          # Original video
├── audio.mp3          # Extracted audio
├── subtitle.vtt       # WebVTT subtitles
├── transcript.txt     # Plain text transcript
├── .meta.json         # Video metadata (--summary only)
└── summary.md         # AI-generated summary (--summary only, written by Claude)
```

## Supported URL Formats

| Format | Example |
|--------|---------|
| Explore link | `https://www.xiaohongshu.com/explore/676a35670000000013002578` |
| Discovery link | `https://www.xiaohongshu.com/discovery/item/676a35670000000013002578?xsec_token=TOKEN` |
| Short link | `http://xhslink.com/a/xxxxx` |

## Troubleshooting

| Problem | Solution |
|---------|----------|
| `No video formats found` | Log into xiaohongshu.com in browser first, use `--browser chrome` |
| `Unable to extract initial state` | CAPTCHA triggered — open URL in browser, solve it, retry |
| Link expired | Copy fresh share link (tokens expire) |
| No subtitles found | Script will fall back to Whisper transcription automatically |
| Whisper not available | Install uv (`brew install uv`) for automatic dependency management |

## Important Notes

- Always use the **full share URL** (with `xsec_token`) for best results
- Log into xiaohongshu.com in your browser before downloading
- Maximum video quality is typically 1080p (platform limitation)
- Whisper transcription requires `uv` for automatic dependency management, or `faster-whisper` installed manually
- Respect copyright and Xiaohongshu's terms of service