Turn Xiaohongshu searches into reusable skills

An agent browsed Xiaohongshu notes, compared the most-liked results, and saved the shared patterns as a knowledge-base skill.

Done with Custom agentยท Intermediateยท Oct 2026

What they did
A custom agent used the rednote-bootstrap pack with agent-browser to control Chrome. It broke a question into smaller angles, generated search terms, checked the result page, opened the most-liked notes, and captured screenshots from each one. A vision model then read the image text, and the agent removed duplicates before saving the findings as a sub-skill in a registry. The user scanned a QR code once to sign in.
How it went
For a question about the publishing process, it reviewed 6 top notes; the highest-liked had 4479 likes; and it produced a full guide. It later built another skill about account warm-up from 4 notes. The project is still early-stage, and a run can take about 5-10 minutes.
Tips
This works best when you cap the topic tightly and let the agent search in rounds. Keep in mind the limits: up to 5 search rounds, 30 notes, and 15 minutes, with login state stored locally for reuse.

Try it yourself with Claude in Chrome

Search [platform] for [topic] and split the topic into a few smaller questions. For each one, run a search, open the most-liked notes, and review both the text and any screenshots or images. Compare the notes for repeated hooks, structure, format, and tone, then merge the shared patterns into a short skill guide called [skill name]. Stop before any action that would post, like, comment, submit, or make changes, and ask me to approve first.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on mcp.csdn.net โ†—

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.