Recently, I wrote two more plug-ins with AI, self-media artifacts

In order to improve efficiency, the author uses AI technology to develop two practical browser plugins for topic selection and long-tail keyword mining. This article details the functions, development process, and usage methods of these two plugins for your reference.

Recently, I opened the “AI+ Self-Media” course, and there are two scenarios that need to be done a lot and repeatedly, so I wrote two browser plug-ins.

The two scenarios are:

1. Look at the topic selection, annotation, and organize it into a table for subsequent AI analysis;
2. Search for keywords to find more precise long-tail needs based on search recommendations

These two plugins use two AI programming methods respectively, and I will share the process with you:

1. Write in one step (corresponding to the plug-in for the first scene)
2. Transform on the basis of an existing plug-in (corresponding to the plug-in for the second scene)

Note-grabbing plugin

Popular note research + analysis should be the link that requires the most time to make an account, and there is no one.

My habit is to read all the topics that can be copied and referenced in a set of keywords at once, and then throw them to AI for analysis of available topics.

By the way, share a prompt:

It is too troublesome to organize notes into the table, and you have to copy links, titles, text, and comments, summarize and comment.

So I wrote a plugin to implement this function:

Click the plugin to directly synchronize the links, titles, text, tags and data of the opened notes to the Feishu Cube to add some annotations

Synchronize the content to the Feishu cube:

If you are interested in this plugin, you can go to Knowledge Planet to get the plugin compressed package and search for keyword plugins.

Share the development process of this plugin:

1. Take out the rules of the front-end page to collect notes and put them in the prompt words for writing plugins. If you can’t find it yourself, you can directly send the entire HTML snippet to the AI, tell it which fields you want to extract, and let it extract the rules for you.
2. The sample code of the development document for creating a record of Feishu Cube is copied and placed in the prompt for writing the plugin.
3. Clearly explain the interaction, operable items, and UI requirements of the plug-in

I’ve played it all over in Windsurf.

I started writing a popup interaction, because it was inconvenient to record comments while watching, so I changed it to a sidebar interaction, which was also changed in one round.

The complete prompt is detailed and reproducible in the last practice live broadcast of the AI programming class, and those who want to reproduce one by themselves can sign up to learn it.

Search for long-tail word collection

Another daily work of topic selection: dig deeper and more accurate demand words behind the keywords of topic selection.

After you enter a keyword in any search box, the system will recommend 5-10 keyword extension questions, which are more specific and in-depth requirements.

Only when the topic selection is deep can we reach a more accurate target group.

My habit is not to keep working hard in the search box, but to export them all at once, filter and dig further one by one.

A technical tycoon has developed a search recommendation collection plug-in before, but 1) it requires secret key authorization; 2) There are many more features that I don’t need very much, taking up browser space.

Although there is a key authorization mechanism, the boss also open-source the plug-in.

I built on his foundation and had Windsurf do a round of magic changes, and now it looks like this:

The above GIF demonstrates the collection level 1, that is, the list of keywords + recommended search terms,

It also supports level 2 and level 3, the so-called level 2 is the keyword + recommended search term + recommended search term search term

According to the GIF demonstration above, 10 search terms are recommended for 1 word, level 2 is 1×10×10=100 words, and level 3 is 1000 words.

In addition to Xiaohongshu, Bilibili and Douyin also support it.

If you are interested in this plugin, you can go to Knowledge Planet to get the plugin compressed package and search for keyword plugins.

Here’s a share of the development process of this plugin:

1. clone to the local plug-in
2. Let Windsurf read the project code, understand the plugin functionality and code implementation, find out what XX does, and introduce it
3. Tell Windsurf what you need and let it “start anew” and write a new plugin

I also went through the functions once, and then found two better-looking colors to let Windsurf optimize a round of UI together.

The git version only has two commits hahahaha

If you want to learn plug-in development, you can sign up for AI programming classes, no basic knowledge is required.

I recorded 13 lessons, and after completing the basic knowledge needed to read code and command AI programming, I could learn it in half a day.

There are also 13 live teaching playbacks, covering common programming scenarios.