@aikangarooking: https://x.com/aikangarooking/status/2066919229153300990
Summary
袋鼠帝 shared two open-source AI skills he developed (viral-topic and viral-title) to help self-media creators quickly find low-follower viral topics across platforms and generate suitable titles. He also introduced the experience of using them in 办公小浣熊 桌面端2.0.
View Cached Full Text
Cached at: 06/17/26, 01:44 AM
Lately, Traffic Anxiety Solved by These Two Open‑Source Skills
Hi everyone, I’m Kangaroo King.
A couple of days ago a friend asked me how to keep grinding on content creation while holding down a day job, and still do it well.
In my view, there are two common mistakes:
- When you’re launching a new account, don’t go too heavy — don’t sink too much energy into polishing content, and definitely don’t hire someone.
- If your main job is super busy and you’re constantly working overtime, it’s not a good fit to split your energy into content creation — unless you’re willing to make a trade‑off.
I have plenty of friends who do great on social media while working full‑time, but as far as I know, they’re generally not that busy at work — either they’re in management making decisions, or they’re in a relatively relaxed department.
Let’s be honest: if you’re working intense hours every day, working late into the night, even on weekends, do you really have the time and energy to run a media account? Even if you do, you won’t be focused. Without focus, it’s hard to do well.
Most people have heard that when starting an account, you should copy low‑follower viral content. That’s true — I was doing exactly that two years ago.
PS: “Copy” here means copy the topic, not copy the content word‑for‑word.
Also, the title is critical for a WeChat Official Account. The quality of your title determines the lower bound of your post’s readership.
So in the early stages, you should spend the bulk of your time on topic selection and title crafting, and keep the content as lightweight as possible.
Many people do exactly the opposite. They spend a ton of time refining content early on, doing all sorts of heavy lifting — only to get single‑digit reads after publishing… and eventually they give up.
I fell into that trap myself. In 2022 I wrote some tech articles. I kept trying to write deep technical pieces, digging into source code and underlying mechanisms. The result? I worked like crazy and nobody read them. Most people don’t like reading — and can’t understand — super deep content (pure self‑indulgence 😄).
Later I realized one thing: topic accounts for 60%, title for 30%, and content maybe just 10%.
Of course, later you still need to raise content quality to grow sustainably.
Topic and title determine the lower bound of traffic; content determines the upper bound.
In the past we had to spend a lot of time hunting for topics and brainstorming titles.
But in the AI era, we can leverage AI to save a ton of time.
To help myself launch accounts on other platforms, and also to find better topics and titles, I developed two skills that I’m sharing today:
One is called viral-topic — it can pull low‑follower viral content from various platforms for a specified domain in recent days.
The other is called viral-title — it gives you reference viral titles adapted to different platforms.
I’ve put them into my kangarooking-skills collection.
Currently supported platforms: WeChat Official Account, X, YouTube, Bilibili (others are being developed).
And these two skills can also be used directly in Office Raccoon · Desktop 2.0 without downloading or installing (more on that later).
Happy 😄 — Raccoon picked them up again (specific usage will be posted in the sticky comment below, since the listing process is still ongoing…).
I’m still optimizing them, but the current results are already pretty good.
viral-topic in Action (Topic Selection)
For example, get low‑follower viral articles about AI from WeChat Official Accounts in the last 7 days.
For WeChat, I collected 22 from the last 7 days. I didn’t strictly follow the “low‑follower” rule (because it’s hard to get exact follower counts on WeChat), but instead defined “viral” as a post getting more than 2x the account’s average read count. I then filtered out some media accounts. The collection looks decent.
PS: Looks like I haven’t had a hit recently — gotta step it up 💪
When fetching from X, it’s even more interesting — it spawns 5 subtasks running in parallel~
In the end, X filtered down to 10, and these are genuinely low‑follower viral posts.
YouTube selections are also great — it directly filtered 58 items. These are high‑traffic content from various countries, with very comprehensive info~
This topic selection skill tries to be as general as possible.
As you can see, whether it’s WeChat, X, or YouTube, the content topics it finds are excellent — all high‑traffic content, and most accounts have low follower counts. Not only useful for launching a new account, but also great for established accounts to reference. After all, if someone can go viral with a small following, the topic itself is a good one.
viral-title in Action (Title Generation)
Honestly, one of the most mentally draining things about writing on WeChat is coming up with titles.
If I already have a good title in mind before I start writing, I have no traffic anxiety, and I can focus all my energy on content quality.
But if the title isn’t decided after finishing the content, it usually means that particular post is hard to title — and then I get anxious.
After all, nobody wants their hard‑earned content to go unnoticed. Everyone hopes more people will see it.
So I made this auxiliary title‑generation skill.
The title of this very article was inspired by viral-title.
The logic of viral-title has three layers:
The first layer is general title methodology, or formulas.
The second layer is platform‑specific viral title methods and formulas.
The third layer is gathering as many viral titles as possible from the same platform, and having the Agent adapt based on how well the current content matches those viral titles. Content that has gone viral can go viral again — same goes for titles.
But be careful: the biggest taboo here is clickbait. You shouldn’t craft a title just for the sake of a title; it must match the content as closely as possible.
That’s why writing a title is hard — you need to find a middle ground: not too bland, not too provocative.
Even with AI assistance, I usually spend 30 minutes to an hour on a title.
Because AI only gives suggestions — most of its titles are unusable. From my testing, you should never completely trust AI; always lead with your own judgment, not AI’s.
By the way, both of these skills were built using Office Raccoon · Desktop 2.0.
I recently saw Jiamu, Cong Ge, and Cang He using Raccoon, so I gave it a try. It’s really handy — extremely feature‑rich, with a great user experience.
A few things that stood out to me:
First, local memory.
Right away it can detect your local Agent memory and read the info you set previously. You can see it immediately knew my name, “Kangaroo King” (it auto‑filled my name). That’s pretty nice — it gets to know you deeply from the start, has your context, which saves a lot of time and makes the work more aligned with your preferences.
Second, strong browser control.
Its browser manipulation ability surprised me. After enabling the browser tool, it could actually export posts from my WeChat Official Account history. The first time I tried, since I hadn’t enabled the browser plugin, it told me I needed to enable it, and also kindly provided a link. Clicking the link took me directly to the plugin location — nice little detail. Then it actually exported my posts from the last six months, including data 🤔 There were 84 items in total. I could even have it sync that data directly to a Feishu multi‑dimensional table — super convenient.
Oh, and the viral topics fetched by viral-topic can also be directly imported into a multi‑dimensional table via Raccoon. It also scored and graded them — so smooth.
Even later, when I was generating titles, it somehow referenced the 84 posts I had it fetch earlier! That was a pleasant surprise.
Third, the Expert Team feature: one person becomes a team — perfect for OPC vibe working. It even includes a “Writing Editor” — after selecting a topic with the topic skill, you can gather relevant materials and hand them to the Writing Editor for help. Then hand it to the title skill, and you have a small content creation loop.
How to Use
These two skills are very simple to use. In Office Raccoon · Desktop 2.0, click “Add Skills” and paste my skill links, and you’re ready.
When you lack topic inspiration, use the viral-topic skill to gather a batch of low‑follower viral content from various platforms.
For example, go to the Agent input and type: “Use viral‑topic to collect low‑follower viral posts about AI from [platform] in the last few days.”
When you can’t think of a title, throw your content at it and run viral-title for three rounds.
By the way, Raccoon recently launched a competition: SenseTime Raccoon OPC Capability Challenge. I saw the prize pool — 550,000 RMB, and the highest individual prize is 100,000 RMB! I immediately signed up for the OPC Master Creation Challenge. Let me dream — hope to win ten grand~
Finally
An important point for a one‑person company (OPC) is to use AI to solidify your own workflow, and continuously optimize and evolve it.
As AI advances and your workflow improves, your OPC will evolve too.
Naval Ravikant once said: “Arm yourself with specific knowledge, take responsibility for leverage, take responsibility for outcomes, and society will reward you in the form of equity.”
In the AI era, I think specific knowledge is exactly these workflows or skills you refine. They are your core assets.
Looking back at account launching — before, you had to manually browse various platforms for topics, spending a ton of time. Now it’s done in one sentence — topics and titles are sorted. The time and energy saved are what you should really invest in content. That’s how I ease my traffic anxiety.
Launching an account is never difficult — what’s difficult is persistence. And the biggest enemy of persistence is not seeing feedback.
If your first few posts can get decent data through good topics and titles, you’re unlikely to give up.
Finally, I want to remind everyone: don’t rely too heavily on AI. You still need to cultivate your own sense for content. AI is just a productivity booster — the final decisions (which topic to choose, what title to use) still rest with you.
I’m Kangaroo King, a blogger dedicated to turning AI into productivity for you. See you next time~
Only the rare few have read this far!
If you found it helpful, feel free to give a like, watch, and share~
If you want to receive updates as soon as they’re published, you can also star me ⭐
Thank you for reading my article patiently~
Similar Articles
@axichuhai: This AI search skill has topped GitHub Trending with 40k stars. It scrapes hot topics from Reddit, X, YouTube, HN in the last 30 days, sorts by real interaction data weighted, and clarifies what you're looking for before searching, resolving names and company names...
An AI search skill tops GitHub Trending with 40k stars. It scrapes hot topics from Reddit, X, YouTube, HN in the last 30 days, sorts by real interaction data, understands search intent, merges duplicate content, and generates HTML briefs, suitable for competitive tracking and topic research.
@QingQ77: A collection of AI skills I use daily to auto-transcribe multi-platform content, manage knowledge bases, and monitor industry intelligence. https://github.com/chubbyguan/chubbyskills… This open-source skill pack includes 11 skills, all of which are tools the author uses daily.
An open-source toolkit containing 11 AI skills that supports automatic transcription of multi-platform content, knowledge base management, and industry intelligence monitoring, ready to be loaded and used in AI agents like Claude Code.
@XAMTO_AI: Brothers and sisters, don't say it's hard to find information gaps anymore! This magical tool scans 10 platforms at once, and directly condenses what the whole internet has been talking about and betting on in the past 30 days into one report! last30days-skill has topped GitHub Trending #1! 7000+ stars in a week, 13k total stars! Not...
An AI agent-driven search engine that aggregates the latest content from platforms like Reddit, X, YouTube, etc., and automatically generates cross-validated reports. It has topped GitHub Trending #1.
@IndieDevHailey: This is literally a godsend for content creators! The viral open-source project AiToEarn: helps you publish content across all platforms and monetize automatically — already at 9.3k Stars and trending on GitHub. No more staying up late editing videos, grinding platforms, replying to comments, or stressing about monetization. One open-source tool to handle your entire workflow: content creation → cross-platform publishing → engagement…
AiToEarn is a wildly popular open-source tool that has garnered 9.3k stars on GitHub and topped the trending charts. It supports one-click publishing to 10+ platforms (Douyin, Xiaohongshu, TikTok, and more), automated engagement management, AI-powered content creation, and a built-in monetization marketplace — helping content creators complete the full loop from content creation to earning money.
@yunxi0623: https://x.com/yunxi0623/status/2066861480579125719
This article introduces five core skills for AI-powered content creation (topic radar, persona profiling, content generation, human-like polishing, distribution and review), helping creators build a reusable content production pipeline and move beyond relying on inspiration.