Xiaohongshu Collector

Work on Xiaohongshu post/comment collection, cookie handling, refresh flows, and browser plugin integration in the forbidden_company repo.

Overview

Use this skill when working on Xiaohongshu collection in forbidden_company, especially for post bodies, comment pagination, cookie updates, single-URL refreshes, or browser-plugin integration.

What To Use

Prefer the existing repo implementation instead of inventing a new flow:

• scripts/collect_xiaohongshu.py
• scripts/admin_server.py
• scripts/run_xiaohongshu_collection.sh
• browser-extension/xhs-collector/
• docs/xiaohongshu-collector.md
• docs/xhs-plugin-api.md

Core Rules

• Keep cookies private. Never repeat them in final output.
• comment_limit=0 means collect all available comments.
• Comment collection must paginate.
• If the direct comment API returns a login/account error, use the browser-rendered fallback.
• Do not rely on Firecrawl for comment pagination.

Workflow

1. Confirm whether the task is batch collection or single-URL refresh.
2. Load the saved cookie from data/xiaohongshu-cookie.txt unless a newer cookie is provided.
3. Run or update scripts/collect_xiaohongshu.py with the requested URL(s), --db, --refresh-url, and --comment-limit 0 when full comments are needed.
4. For browser plugin work, wire the popup/background scripts to the local backend endpoints in scripts/admin_server.py.
5. Verify that post rows, comment rows, and exported artifacts are written correctly.

Endpoint Map

Use these backend endpoints when integrating the browser plugin:

• GET/POST /api/xhs-cookie
• GET /api/xhs-plugin/status
• POST /api/xhs-plugin/collect
• POST /api/xhs-plugin/refresh

Validation Notes

• Refresh mode must delete the old note rows before writing the new ones.
• The plugin should expose downloadable CSV and JSON artifacts.
• When debugging, check whether the failure is cookie-related, pagination-related, or page-structure related.

Safety Notes

• Do not propose or implement shared-server mass scraping.
• Keep the browser/plugin model user-driven and local-first.
• Preserve source URLs and timestamps for traceability.

Reference

See collector-workflow.md for operational details.