AI Agent Skills 中文导航平台

Users want to extract Xiaohongshu post content. Please follow these steps to process:

Constant Definitions

Cookies File:
```
~/cookies.json
```
(Xiaohongshu cookies exported from Chrome)
Obsidian Save Directory:
```
~/Documents/Obsidian Vault/xhs
```
Whisper Model:
```
mlx-community/whisper-large-v3-turbo
```

Input

Xiaohongshu link provided by user: $ARGUMENTS

Extraction Process

Step 0: Check Cookies

Check if
```
~/cookies.json
```
exists

If it does not exist, inform the user to export cookies from Chrome:

Open xiaohongshu.com in Chrome and confirm you are logged in
Open DevTools Console and run the following code to copy cookies to clipboard:

javascript

copy(JSON.stringify(document.cookie.split('; ').map(c => {
  const [name, ...rest] = c.split('=');
  return { name, value: rest.join('='), domain: '.xiaohongshu.com', path: '/',
    expires: Date.now()/1000 + 86400*30, size: name.length + rest.join('=').length,
    httpOnly: false, secure: false, session: false, priority: 'Medium',
    sameParty: false, sourceScheme: 'Secure', sourcePort: 443 };
})))

Save the clipboard content to
```
~/cookies.json
```
Then terminate the process and rerun after the user completes the above steps

Step 1: Parse the Link

Extract the post ID (24-digit hexadecimal string) and xsec_token parameter from the URL.

Step 2: Retrieve Post Content

Use a Python script to request the post page HTML via Cookies, and parse all post data from

window.__INITIAL_STATE__

python

import json, urllib.request, ssl, re

with open('<Cookies 文件>') as f:
    cookies = json.load(f)
cookie_str = '; '.join(f"{c['name']}={c['value']}" for c in cookies)

ctx = ssl.create_default_context()
ctx.check_hostname = False
ctx.verify_mode = ssl.CERT_NONE

req = urllib.request.Request('<帖子URL>')
req.add_header('Cookie', cookie_str)
req.add_header('User-Agent', 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36')

resp = urllib.request.urlopen(req, timeout=15, context=ctx)
html = resp.read().decode('utf-8', errors='ignore')

m = re.search(r'window\.__INITIAL_STATE__\s*=\s*(\{.+?\})\s*</script>', html, re.DOTALL)
raw = m.group(1).replace('undefined', 'null')
data = json.loads(raw)

# Post data is located at: data['note']['noteDetailMap'][<key>]['note']
# Contains: title, desc, type, time, user, imageList, video, interactInfo, ipLocation

If the request fails (redirected to 404/error page), it means the cookies have expired. Prompt the user to re-export cookies as per Step 0.

Step 3: Video Transcription (Video Posts Only)

If the post type is video, perform the following sub-steps:

3a. Extract Video URL

Parse the video stream from the data obtained in Step 2:

note['video']['media']['stream'] -> Take the masterUrl of the first stream in the priority order of h264 > h265 > av1

3b. Download Video and Extract Audio

bash

curl -L -o /tmp/xhs_{post_id}.mp4 -H "Referer: https://www.xiaohongshu.com/" <视频URL>
ffmpeg -y -i /tmp/xhs_{post_id}.mp4 -vn -acodec pcm_s16le -ar 16000 -ac 1 /tmp/xhs_{post_id}.wav

3c. Speech Transcription

python

import mlx_whisper
result = mlx_whisper.transcribe("/tmp/xhs_{post_id}.wav",
    path_or_hf_repo="mlx-community/whisper-large-v3-turbo", language="zh", verbose=False)

3d. Clean Transcribed Text

Remove trailing repeated characters (background music noise)
Split sentences by semantics, add punctuation and paragraphs
If there is a step/point structure, format it with Markdown

3e. Clean Temporary Files

bash

rm -f /tmp/xhs_{post_id}.mp4 /tmp/xhs_{post_id}.wav

Step 4: Organize Output and Save

Organize the content into a Markdown file and save it to

<Obsidian Save Directory>/{YYYY-MM-DD} {Short Title}.md

File name format:
```
{Post Date} {Short Title}.md
```
. The short title should be no more than 15 characters, which is a concise summary of the core insight
The date prefix ensures sorting by time
Do not create subdirectories; all post md files are placed directly in the xhs folder

Media files are uniformly stored in

<Obsidian Save Directory>/img/

<Obsidian Save Directory>/video/

Writing Style: Peter Thiel-style — Direct, counterintuitive, and deliver judgments in one sentence. Notes are decision-making tools, not knowledge bases. Users can decide at a glance whether to dig deeper or skip.

File Structure (No YAML frontmatter):

markdown

# One-sentence Core Insight (Counterintuitive judgment, not descriptive title)

Core argument, 2-3 sentences. Directly give a judgment like "Most people think X, but actually Y".
No nonsense, no foreshadowing, just like Thiel speaking in a board meeting.

**Relevance to Me:** One sentence. Read the user's memory (user and project-type memory under ~/.claude/projects/*/memory/) to understand the user's background, research direction, and current work, and clarify how this content relates to the user. If memory is unavailable, start from the perspective of general personal development/tools/methodology.

**Is it worth digging deeper?:** Yes/No. One-sentence reason.

> [!tip]- Details
> Structured organization of the post's core content (collapsed state, visible only when expanded):
> - Extract from desc and video transcription, clean up `#xxx[topic]#` tags
> - Divide into sections by logical structure, retain key data and conclusions
> - Embed images using `![Image N](urlDefault)`
> - For video posts, place the organized transcribed content here

> [!info]- Note Attributes
> - **Source**: Xiaohongshu · Author Name
> - **Post ID**: xxx
> - **Link**: Original Link
> - **Date**: YYYY-MM-DD
> - **Type**: image/video
> - **Engagement**: N Likes / N Saves / N Comments
> - **Tags**: Tag 1, Tag 2, ...

Key Constraints:

Visible content outside collapsed areas must not exceed 6 lines
The title must be an insight/judgment, not "Summary of XX Post"
Use the URL from the
```
urlDefault
```
field for images

xhs

NPX Install

Tags

SKILL.md Content (Chinese)