One Bilibili comment earned 600 stars: a code-animation style pack of 341 styles brought over, how did it beat the original author

In Oct 28, a Bilibili creator freely released a set of 341 code-animation style prompts he had put together. The video took off, but the link was only in the very bottom of the description, invisible on mobile. The next day, a Shanghai developer with no connection to animation uploaded the whole pack to GitHub, wrapped it as a Claude Code / Codex Skill, and the link also appeared in the most-liked comments under the original video. 8 days later the repo had 622 stars, while the original author's own official repo had only 223. Below: why it grew, where the stars came from, and whether it's worth trying to get stars this way.

A video whose link you can't see

In 2026 Oct 28, Bilibili creator "Yangfu Guanzhao" posted a video titled "[Open Source Experience] Code animations made by Opus 5.5, DeepSeek can do them too! 341 style prompts, all free". The pitch in one line: what Opus 5.5 can do, DeepSeek can do for a few cents, which hit home for the many people in China who can't use Opus. The description also said: "All free, don't package it to sell courses, don't repackage it as paid skills".

As of 11-06, the video had 99,758 views, 15,748 favorites, 9,126 likes and 767 comments (Bilibili, 11-06). Favorites were 16% of views, showing many people came in just to take the material.

The problem was getting it. Bilibili comments don't allow external links, so the creator put the download address at the very bottom of the description. His pinned comment said "Can't post links, they can only go in the description, at the very bottom"; later, in the description of a second video, he added: "Some friends report they can't see it on mobile, and my pinned comment gets swallowed too."

In Oct 29, a user called "Hakim Mide" left a comment under the video: "github.com/XiaoyeCodes/code-animation-prompts I've already put it up on GitHub for everyone". The comment got 344 likes and was pinned as a top comment. A reply under it said: "The GitHub version is much easier, thanks for reposting" (23 likes). Whether this commenter and the GitHub author are the same person can't be confirmed.

What it actually is

The repo homepage's first screen shows a 24.5-second promo GIF, with a line underneath: "341 style prompts · 196 reference stills · 10 series · Claude Code & Codex Skill".

What's inside:

• 341 Markdown style prompts across 10 series: particles, neon, ink wash, paper cut, Dunhuang, American comics, plus the visual styles of 98 Oscar-winning films.
• 196 reference stills, plus 145 styles described in text only.
• An offline HTML style picker you can search, filter, tick, and export as a list.
• A small script to look up styles by number or keyword (e.g. styles.py search 水墨), so you don't have to load the whole library into context.
• A render script using Playwright and FFmpeg that exports HTML animations frame by frame to MP4.
• A "one-click import prompt": copy the whole block and paste it to Claude Code or Codex, and it downloads and installs everything itself, then replies with an example request for a "12-second particle-to-shape animation".

The content isn't the author's own. LICENSE.md is clear about it: the assets come from a free asset pack by 俯仰观照, and three of the series come from three other MIT-licensed open-source projects (Lemo-Opuscar, mg-styles-15, code-video). The author XiaoyeCodes's GitHub account was created 2026 years ago in January and has 8 followers. Their other repos are investment research and a tool for auto-posting to a Xiaohongshu finance account, nothing to do with animation. What they did is packaging: wrapping it as a Skill, adding scripts and docs, and spelling out the source and the "free" terms.

Where the stars came from

(GitHub, daily stars and forks)

The fork-to-star ratio is about 1:4, which is very high on GitHub. Most people coming in are there to grab the assets, not just bookmark it.

Wave one was the top comment on Bilibili. The repo was created by 10-29, less than a day after the video went up. The original creator's official repo, Akillll-DuoHe/fy-code-animation, was only created on 10-30 morning, a day and a half later, and it spread the link across cloud drives, Gitee and GitHub at once, so the entry points were scattered. His second video got only 23k views, far below the first. By 11-06, the official repo had about 223 stars.

In the same period, at least five other people reposted the same assets (wqb578, mengxing123456789, 2163882817, dream-cometure, and others), each with only 0 to 25 stars. The assets were identical; the difference came down to two things: this one grabbed the top-comment slot, and it did one more step of packaging than the rest. Most of the others just dropped a folder full of md files online. This one turned it into a Skill you can install, search and render.

Wave two isn't fully traced. After 11-02, the Bilibili source was already cooling, yet daily new stars went up from +85 to +136. Several things line up in time:

• Chinese Skill directory sites started listing it: kapyn.app (10-29), skillsllm, agentskillshub, hotskills.dev (396 stars when listed), and ByteNote (11-06).
• An English-language short-video account that reposts "a new GitHub repo every day" picked it up. On TikTok, @github.signals 11-03 posted "Code Animation Prompts: 341 Styles in One Click", which got 7,901 views and 540 saves; the same account also posted on Instagram, YouTube Shorts and Facebook. These are automated repost accounts, not big creators.
• The author 11-04 swapped the README header image for that promo GIF, and 11-05 added a Star History chart.

There's no data to say which of these contributed most. On X, no big account reposted it; over the last 30 days, searches for the repo name, the author's name, or "341" each returned 0 posts.

Where it ranks in this niche

In 2026 October, "let Claude Code write code to make animations" was a wave, and dozens of related Skills and repos appeared within a month:

(GitHub,11-06)

Looking at this table, 622 stars puts it in the lower-middle. Its only edge is quantity: others have 7 to 35 styles, it has 341. "341 Styles in One Click" is a good title, and directory sites and repost accounts love it. But anyone can get these assets, there are over a dozen copies already and 170 forks, so it has nothing others can't take.

What it wins on, and what it lacks

It didn't win on product, it won on a gap: the content had gone viral, but the original creator had no easy way to get the link to the people who wanted it. Whoever first put the files somewhere easy to grab, and then put the link in the top comment everyone sees, caught this wave. The quality of the packaging is what amplified the traffic, making directory sites and English repost accounts willing to share it.

The weaknesses are also clear:

• Little staying power. The assets aren't its own, the original creator's official repo keeps updating, and it has added storyboards and voiceover. Once this Bilibili wave passes, there's no reason for it to keep growing.
• Copyright is grey. Content like the visual styles of 98 Oscar-winning films has never had its rights boundaries checked.
• The relationship with the original creator is borrowed. It avoided conflict only by stating the source and honoring the "free" terms. That approach might not be accepted if the original creator were different.

If it were you

Verdict: C. You can learn from it, but copying it step by step isn't worth it. This 341-style pack has already been reposted a dozen times, and another repost won't earn stars. What's worth learning is the playbook behind it: free material that goes viral on Chinese platforms often gets stuck at the step where the link can't be posted, or the cloud drive requires a transfer. Whoever first turns it into an easy-to-install GitHub Skill and puts the link in the top comment of the original post can ride a wave. But the scale of this wave is a few hundred stars, the window is a day or two, and it depends on whether the original creator is willing.

Who it's for: people who scroll Bilibili, Xiaohongshu and Douyin daily for AI tool content and can act within 24 hours of a hit; people who can turn scattered assets into a Skill that's easy to install, search and run with scripts; people who want to build a few repos with a few hundred stars each to get their GitHub going.

Who it's not for: people who want an open-source project that keeps growing and makes them memorable. Reposted stars don't make the reposter memorable, only the content. It's also not for people with no original content who aren't willing to reach out to the original creator: reposting without a word easily gets called out by the creator and their followers.

How to do it, ranked by how worthwhile:

1. Build the layer nobody does well in this category. Right now everyone sells either asset quantity (this one), an engine (Huashu), or finished-video showcases (awesome lists). Still open: a style library organized by use case (intros, explainers, product launches, each with templates that render directly), or an end-to-end pipeline of style + voiceover + subtitles into a finished video. Doing this takes your own work, and the star ceiling is higher.
2. Do "repost plus packaging" on other platforms and with other audiences. Watch for free material (prompt packs, templates, workflows) that is blowing up on Xiaohongshu, Douyin or Jike with broken links. Reach out to the original creator first, or offer to build their official repo for them and ship the same day. That's safer than reposting in secret, and the original creator's audience will accept you.
3. Bring Chinese content to the English world. This repo got reposted by English accounts on TikTok and Instagram, which shows English users want it too. A Skill that goes viral in the Chinese scene usually has no English version. Build an English README, English style names and English examples, and directory sites and repost accounts will come to you.

How to get users: the top comment on the original post on Bilibili or Xiaohongshu is the most valuable spot. Go comment as soon as you launch, and say clearly "free, and here's the source". Put a GIF showing the result at the top of the README, and a big number in the title. Then submit to the Skill directory sites, and English repost accounts will find you.

When the window closes: a single opportunity's window is only 24 to 48 hours after the original video goes up. For the "AI writes code to make animations" theme, this wave will probably stay hot until 2026 year-end (inference); once stronger video models arrive, the novelty of code animation will drop noticeably.

What would show you misjudged it: ① if the repo still gains dozens of new stars a day after mid-November, that means its moat isn't just getting there first, and the packaging itself has value, so the second and third paths are worth more; ② if the original creator's official repo overtakes it within two to three weeks, that means the "repost to grab the entry point" approach doesn't last a week, so don't treat it as a playbook; ③ if the original creator or the platforms start publicly opposing reposts, drop this path.

Data and sources

GitHub: the XiaoyeCodes/code-animation-prompts repo page, README, LICENSE.md and daily stars (10-30 to 11-06); star counts of comparable repos (11-06). Bilibili: 俯仰观照 BV1bqx76wEUM (10-28) and BV1vsx66BE4p, with views, saves and descriptions, plus the original video's comment section (11-06). TikTok: @github.signals 11-03 video. Directory sites: kapyn.app, skillsllm, agentskillshub, hotskills.dev, bytenote. Places marked "inference" or "couldn't confirm" are inferences or unverified.