xiaohongshu-search

Xiaohongshu Search

Extract searchable content from Xiaohongshu using one local script.

When to use this skill

• The user wants to search Xiaohongshu content by keyword
• The user wants to batch-extract image links from search results
• The user wants to export search result titles for subsequent topic selection or analysis
• The user needs a repeatable local scraping workflow

Scripts overview

Script	Purpose	Dependencies
scripts/search.py	Open the Xiaohongshu search page and extract image links and titles	playwright

Setup

1. Verify Python

python --version

2. Install dependencies (first run)

pip install playwright
playwright install chromium

Critical error recovery: if a command fails with missing dependency/browser errors, install the missing dependency, then rerun the exact same command.

Usage

Basic search

python scripts/search.py "AI"

Limit extracted items

python scripts/search.py "AI" --count 30

Save output files

python scripts/search.py "AI" --count 30 --save ./output

Parameters

• keyword (required): search keyword
• --count (optional): maximum number of items to extract, default 20
• --save (optional): local output directory

Output

Terminal output

Script prints JSON at the end:

{
  "keyword": "AI",
  "images": ["https://..."],
  "titles": ["..."],
  "count": 20
}

Files (when --save is used)

• <keyword>_links.txt: list of image links
• <keyword>_titles.txt: list of titles (written when titles are available)

Execution behavior

• The script opens a browser and visits the Xiaohongshu search page
• If not logged in, it prompts for manual login and waits
• It automatically scrolls the page to load more results before extracting
• It only collects links/titles at the search results level and does not scrape comment threads on detail pages

Limits and edge cases

• Changes to Xiaohongshu's page structure may cause selectors to fail, requiring updates to scripts/search.py
• Login, CAPTCHAs, and risk control policies may affect the scraping success rate
• Different keywords return inconsistent structures; titles may be empty or fewer in number than images
• The script uses a visual browser workflow, not a pure headless API call

Out of scope

This skill does not provide the following capabilities in current repository state:

• Publish image-text/video posts
• Like, comment, reply, and bookmark
• MCP server tool calls
• Login scripts, cover generators, smart publishing scripts

Script reference

• search.py: Xiaohongshu keyword search extractor