"Listed and featured": a developer with 42 followers sent individual notifications to 185 projects and grew a self-media skill list to 200 stars
What it is A self-media operations skill list for AI agents. It collects 185 skills for Xiaohongshu, Douyin, WeChat Official Accounts, X, and more, each with a plain-language summary, script generation, and daily automatic refresh, and it teaches you how to install them into Qwen, Doubao, WorkBuddy, and similar agents.
Why it took off Chinese agents can now install skills, and self-media skills are too numerous for anyone to sort through by hand. The author opened an issue notification on every popular listed repo, handed out badges, and added an X roundup post that brought a single-day +54, so the stars accumulated bit by bit.
BHow you could follow it Pick a Chinese vertical you know well, use the same generator to open several lists at once, write plain-language summaries and safety checks, cold-start with polite listing-notification issues, then run a controlled benchmark on the one that responds best and pitch it to recommendation accounts like GitHubDaily.
- Data (updated daily) · Repo created 2026-10-19 · 219 stars · 7 days +154
GitHub · stars Data 2026-10-29 → 2026-11-08 (11 days, 11 observations). Points before 10-28 come from deeply / archive backfill, not every day has one
In October 25, an account called UnderleveledBuilder opened an issue under the baoyu-skills repo by Baoyu (27k stars) titled "Listing notice: baoyu-skills is listed in Awesome Social Media Skills." The body said it had been added to a curated self-media skill list, placed first in the "full-workflow operations" section, with the list's summary quoted, and closed with "No reply needed, you can close this issue directly. Thanks for the project!" In November 1, Baoyu replied "Thanks for listing!" and closed it. The same issue was opened under a long run of repos, including the Xiaohongshu MCP, WeWrite, and guizang's social-media card skill. The list was created 10-19, and by 11-06 it had 197 stars (GitHub's current count is 208). Not a breakout, but almost all of it came from a playbook anyone can copy.
What it is
A bilingual (Chinese/English) awesome list collecting "self-media operations skills" for AI agents: writing Xiaohongshu notes and WeChat articles, making covers and short videos, scheduled publishing, tracking trends, and reviewing data. Platforms covered include Xiaohongshu, Douyin, WeChat Official Accounts, WeChat Channels, Bilibili, X, YouTube, TikTok, and LinkedIn. The README says it lists "185 skill projects + 27 MCP tools and open-source platforms" (metadata 11-05).
A few ways it differs from a typical hand-written awesome list:
- Plain-language summaries. Instead of copying repo descriptions, each entry explains what it does and where login credentials are stored, for example: "Bilibili MCP: scan-to-login... send private messages; login credentials stored in a local JSON."
- Gatekeeping. Before listing, scripts check for binaries, install hooks, remote execution, and reads of sensitive files. The 21 rejected projects are logged separately in a file. The README opens with a warning: tools that post on your behalf usually need your account login, and bulk publishing can get accounts banned.
- Self-updating. Data lives in a config file, a script generates the README, and GitHub Actions refreshes star counts and update times daily. The author also manually adds 1 to 4 new repos each day.
- Written for non-coders. The FAQ specifically explains how to install into Qwen, Doubao desktop, OpenClaw (Lobster), WorkBuddy, Kimi Code, and Trae.
- The list is itself a skill. Once installed with one command, you can ask your agent directly: "I run a Xiaohongshu beauty account, which skill should I install to write notes?"
How the stars accumulated bit by bit
| Date | stars | Daily adds |
|---|---|---|
| 10-29 | 39 | |
| 10-31 | 58 | +9 |
| 11-01 | 65 | +7 |
| 11-02 | 119 | +54 |
| 11-03 | 145 | +26 |
| 11-05 | 190 | +22 |
| 11-06 | 197 | +7 |
(GitHub daily stars; forks are 32, about 16% of stars, which is high for lists, suggesting many people are forking it to customize.)
Three forces, none of them a big-account recommendation:
Send "listed" notification issues to the projects that get included. The issue-style outreach above is concentrated under a batch of repos with 1,000 to 100k stars, over 10-25 through 10-29. Each entry shows up once in other people's issue lists and in their followers' notifications, and the authors often give the repo a star in return, and some even put the "listed in" badge in their own README, which becomes a long-lasting backlink. The steady +7 to +10 a day through late October is largely attributed to this (I can't match it to individual stars).
A Xiaohongshu skill roundup post. On the evening of 10-31, @qichen1019 on X posted a roundup titled "Full Xiaohongshu Operations Skill Set" listing 12 skills. The order and star counts closely match the Xiaohongshu section of this list, and the post summary carried a link to the list. The next day brought the only spike of the whole period, +54. The timing lines up, but there's no way to confirm that all 54 stars came from it.
Another list from the same author. Earlier, on 10-13, he had published a short-drama skill list, which GitHubDaily recommended, and it reached 1,845 stars. The lists link to each other at the bottom of their READMEs, so the self-media list got a boost along the way.
No trace of it on Hacker News, Reddit, Xiaohongshu, Douyin, or in media coverage. Its audience is a small circle of Chinese self-media creators and developers looking for "which skill should I install."
Same template, 6 lists, only one took off
This is the most interesting part of the project. From 10-13 to 10-26, the author used the same generator, the same README structure and badges, and launched 6 "Chinese vertical scenario × skill" lists (on GitHub, 11-06):
| List | Repo created | stars |
|---|---|---|
| Short-drama skills | 10-13 | 1,845 |
| Self-media skills (this one) | 10-19 | 197 |
| SEO / GEO skill | 10-19 | 49 |
| Web-novel skills | 10-15 | 48 |
| Story skills | 10-15 | 8 |
| Chinese GEO | 10-26 | 0 |
All done with the same care, yet the results differ by two orders of magnitude. The short-drama list got recommended because the author included a controlled, blind-tested comparison, which gave GitHubDaily something to talk about. The self-media list had no hands-on testing, just a well-organized directory. The pattern is clear: a template can't guarantee traction. Traction depends on whether a big account picks it up, and big accounts want to see that you've tested it.
Why it ranked second: the topic sits between "AI agent skills" and "making money in self-media," two of the highest-traffic subjects on Chinese web. The timing was right too. In the second half of 2026, Qwen, Doubao, and WorkBuddy all added skill support, and Xiaohongshu launched notes that attach skills in July. On GitHub alone there are hundreds of Xiaohongshu and WeChat Official Account related skills, so "which one to pick" became a real question.
Who's already doing this
| Project | Scale | How it differs |
|---|---|---|
| ComposioHQ/awesome-claude-skills, VoltAgent/awesome-openclaw-skills | 79k / 54k stars | Broad and comprehensive, social media is a small slice, with almost nothing on Xiaohongshu or WeChat Official Accounts |
| JuneYaooo/awesome-ai-media-cn | 359 stars, created 2026-04 | Collects AI video and self-media tools, not skills you can install into an agent |
| "Must-install skills for self-media" articles from outlets like NewRank and SMZDM | The SMZDM article reports 45k bookmarks | One-off articles, a dozen or so entries, never updated |
| Skill directory sites like skills.sh and SkillForge | Thousands of entries, SkillForge has 350+ in the self-media category | Lists everything without filtering, no plain-language Chinese descriptions |
| murphywuwu/awesome-social-media-skills | — | Copied wholesale, text word for word, data frozen at 10-26 |
Copies already exist, and the short-drama list was copied twice. Lists like this have a low barrier to entry, anyone can copy one under a new name. What keeps people is "still updated every day" and "I tested it."
If it were you
In one sentence: this is a star-gathering playbook anyone can follow, without a big following or a big company behind it, but a single list tops out at a few hundred stars. Getting past a thousand means betting on the right scenario and then publishing hands-on testing that makes big accounts willing to promote it.
Who it's for: developers who really know a particular vertical (e-commerce, cross-border, education, legal, design, games…) and are willing to spend half an hour a day finding new repos; people who want to build a reputation as "the person who knows this niche's skills best."
Who it's not for: people who want to compile it once and walk away. Once the list stops updating, copied versions and directory sites will drown it out. It's also not for people who only want a general English list, that slot is already taken by huge lists with tens of thousands of stars.
How to do it: First build the generator (config file → script → README → daily auto-refresh). Within a week, launch three or four Chinese vertical scenarios nobody has covered yet, then double down on whichever gets a response. Write a one-line plain-language description for each entry, note login requirements and risks, run a security check, explain how to install into domestic agents, and make the list itself a skill. Pick the one with the strongest response and run a controlled comparison of 5 to 10 popular skills, presented as a table at the very top of the README.
How to get users: Listing-notification issues are the cheapest cold-start distribution, but the tone should match this author's: say clearly which section the project landed in, how the description was written, and how to fix it if it's wrong. End with "no reply needed, feel free to close this" so people are willing to reply "thanks for listing" instead of deleting it as spam. Then hand the test results to recommendation accounts like GitHubDaily and to people doing skill roundups on X, and submit to directory sites like skills.sh.
Why now: Domestic agents have only just become widely able to install skills, and the "curated list" slots for each vertical are still being claimed one by one. Self-media and short-drama already have owners. The window is probably a month or two (my estimate).
Signs you've misjudged it: ① More than 50 listing notifications sent and still under 50 stars two weeks later, meaning creators in that niche aren't buying in, or the scenario is too cold; ② none of the lists you launched has averaged more than 20 new stars a day, and no recommendation account is willing to share; ③ directory sites like skills.sh start publishing scenario-based curated pages with Chinese descriptions and reviews, leaving no reason for a standalone list to exist.
Research notes: competitor table, acquisition evidence, full research report
- Competitors
Name Where Launched Scale How it differs from this one JuneYaooo/awesome-ai-media-cn Open source 2026-04 359 stars Collects AI video and self-media tools and apps, not skills that can be installed into an agent; no installation guides for domestic agents and no security checks ComposioHQ/awesome-claude-skills and VoltAgent/awesome-openclaw-skills Open source 2025-10 / 2026-01 78.8k / 53.6k stars Broad general-purpose lists, social media is a small category, and they barely cover Xiaohongshu, WeChat Official Accounts, or WeChat Channels social-media-skills/skills, blacktwist/social-media-skills, charlie947/social-media-skills Open source 2026-03~07 252 / 631 / 4.6k stars English social media skill collections written by their authors, not third-party curated lists, and they don't cover Chinese platforms "Must-install skill list" articles from NewRank, SMZDM, SegmentFault and similar outlets, plus roundup posts on X Web 2026-05~10 The SMZDM article reports 45k bookmarks One-off articles with a dozen or so entries, no updates, no security checks; this list has 185 entries that refresh automatically every day Skill directory sites like skills.sh, skillsdirectory, and SkillForge Web 2026 Thousands of skills, SkillForge has 350+ in the self-media category Full listing with no filtering, no plain-language Chinese descriptions, no grouping by platform Xiaohongshu RED Skill Big-company features 2026-07 Not verified One-click copy of skills attached to notes on the platform; covers Xiaohongshu only and offers no cross-platform comparison - Where users come from
- Confirmed Sending "listed" notification issues and badges to included projects: the author opened issues in a uniform format on baoyu-skills #223(10-25), dbskill #53, digital-human-koubo #2(10-26), gzh-design-skill #27, WeWrite #39, xiaohongshu-mcp #885(10-29), guizang-social-card-skill #21 and others, and Baoyu replied "thanks for listing."
- Inferred X roundup post: @qichen1019 2026-10-31 22:09 UTC, "Full Xiaohongshu Operations Skill Set," 12 entries that match the Xiaohongshu section of this list, with a search summary card linking to this repo. We observed 11-01→11-02 +54 as the only spike.
- Confirmed Cross-linking with sister lists (same-author matrix): the README's related resources cross-link short-drama (1,845 stars, recommended by GitHubDaily 10-24), seo-geo, and story
- Confirmed Automatic inclusion by skill directory and aggregator sites: the skills.sh social-media-skill-finder page, skillsdirectory.com Grade A, skillsllm, repolode, kopass, and others
Why UnderleveledBuilder/awesome-social-media-skills worked
Research date 2026-11-06. Data sources: our GitHub database (repo_day / repo_profile, daily since 2026-10-28), GitHub pages and API (checked 11-06), Brave / Exa search, and Jina for reading X posts. Stargazer-level detail isn't available (the API returns 404 / rate-limited for this repo), so which day and which people the stars came from can only be inferred from the timeline.
Bottom line
This isn't a "hit"; it's the second product from someone who mass-produces awesome lists. The author UnderleveledBuilder (X @UnderleveledDev) launched 5 "Chinese vertical scenario × agent skill" curated lists from 10-13 through 10-26. The short-drama list reached 1,845 stars after GitHubDaily recommended it. The self-media list never got a big-account recommendation. Its 197 stars came mainly from sending "listed" notification issues and badges to each of the 185 projects that were included, cross-linking with sister lists, and a wave from 10-31 a Xiaohongshu skill roundup post (11-02 with +54 in a single day). It has some exposure within a small circle but is far from widely discussed, and someone has already copied the whole thing.
1. What it is
- A bilingual (Chinese/English) awesome list of agent skills for self-media operations (Xiaohongshu, Douyin, WeChat Official Accounts, WeChat Channels, Bilibili, X, YouTube, TikTok, LinkedIn), covering 185 skill projects + 27 MCP / open-source platforms (README, metadata 2026-11-05).
- Not hand-written Markdown:
data/skills.toml+scripts/build.pygenerate the README, and GitHub Actions automatically "refreshes metadata and README" 16:2x daily at UTC. The author manually adds 1–4 new repos per day (commits record 10-19 to 11-05). - Each entry has a plain-language Chinese sentence explaining "what this skill actually does," plus platform, last-updated date, skill count, license, and 💤/📦 markers. Entries are grouped by "full workflow / writing / visuals / publishing / topic data / growth / MCP / complete apps," with a separate lookup by platform.
- Self-described as "manual review + scripted security checks" (checking binaries, install hooks, remote execution, reads of sensitive files). Rejected projects are recorded in
data/ignore.toml(21 entries). - The list itself is also a skill:
npx skills add UnderleveledBuilder/awesome-social-media-skillsinstalls social-media-skill-finder, and you can ask your agent "which one should I install to run a Xiaohongshu beauty account?" It has already been picked up by skills.sh, skillsdirectory.com, and other directories. - The FAQ specifically explains how to install it into domestic agents such as Qwen Office, Doubao desktop, Longxia OpenClaw, WorkBuddy, Kimi Code, Trae, and Qoder. That's its target audience: people in China doing self-media who aren't necessarily programmers.
2. Growth data
| Date | stars | fork | Daily adds |
|---|---|---|---|
| Repo created 10-19 | 0 | ||
| 10-29 | 39 | 8 | (first observation) |
| 10-30 | 49 | 9 | +10 |
| 10-31 | 58 | 11 | +9 |
| 11-01 | 65 | 12 | +7 |
| 11-02 | 119 | 18 | +54 |
| 11-03 | 145 | 24 | +26 |
| 11-04 | 168 | 29 | +23 |
| 11-05 | 190 | 32 | +22 |
| 11-06 | 197 (checked live on GitHub: 208) | 32 (checked live 34) | +7 |
Source: github repo repo_day. fork/star ratio 16%, skews high for list-type repos (people fork them to customize, and some copy the whole thing, see below).
Same author's matrix (github repos, 11-06):
| Library | Repo created | stars | Trend |
|---|---|---|---|
| awesome-short-drama-skills | 10-13 | 1,845 | 10-20→10-21 +349,10-26/27 re-accelerated, 7 days +590 |
| awesome-social-media-skills | 10-19 | 197 | 7 days +148 |
| awesome-novel-skills | 10-15 | 48 | slow climb |
| awesome-seo-geo-skills | 10-19 (created in the same minute as this repo) | 49 | 10-28 jumped +22 then stalled |
| awesome-story-skills | 10-15 | 8 | barely moved |
| awesome-geo-cn | 10-26 | 0 | — |
The same template (README structure, badges, FAQ, cross-links to related resources) was reused 5 times. Only the short-drama one really took off, with the self-media one in second place. So the template alone doesn't guarantee traction. Traction depends on whether a big account picks it up.
3. Competitors
Similar 'self-media / social media skill curated lists'
- JuneYaooo/awesome-ai-media-cn (359 stars, created 2026-04): a collection of AI video and self-media tools, 150+ open-source projects, leaning toward 'tools' rather than 'skills'; also awesome-ai-media-skills-cn. It came out half a year earlier but never took off.
- IRONICBo / replynodes' awesome-social-media-skills (English, 26 skills, 7 stars on the replynodes version): actually a lead-gen repo for the social media SaaS ReplyNodes. Same name, different thing.
- Leo-Drq/self-media-skills, ttfake92-lab/skills (326 stars) and others: skill collections written by their own authors, not third-party lists.
- Chinese article-style 'must-install skill lists': Xinbang's 'The 15 skills truly suited to new-media people' (2026-07), SMZDM's 'Must-install Skill list bookmarked by 45k people' (2026-09), SegmentFault, Wuju AI, etc.; plus roundup threads on X from @qichen1019 (10-31), @MissCat_AI and @Smartpig. These are the real head-to-head competition: the same crowd looking for the same answer, but they are one-off articles and never updated.
- Skill directory sites: skills.sh, skillsdirectory.com, SkillForge (skills.yangsir.net has a 'self-media operations 350+ skills' category), everythingskill.net. Bigger volume, but no curation and no plain-language descriptions.
- The platform itself: RedNote (Xiaohongshu) launched RED Skill on 2026-07 (notes can attach a skill, one-click copy of the prompt, @Gorden_Sun post), which is effectively a distribution entry inside the platform.
Large general skill lists (all with social media categories, but thin coverage of Chinese platforms): ComposioHQ/awesome-claude-skills 78.8k, VoltAgent/awesome-openclaw-skills 53.6k, sickn33/agentic-awesome-skills 49.2k, VoltAgent/awesome-agent-skills 36.9k (github repos).
The top listed projects themselves (users might go straight to them without needing a list): Agent-Reach ~109k, baoyu-skills 27.3k, social-auto-upload 16.2k, xiaohongshu-mcp ~16.7k, guizang-social-card-skill 8.0k, ScrapeCreators/social-media-research-skills 5.8k (7 days +700, the fastest recently), sergebulaev/linkedin-skills 5.6k, wewrite 3.6k (github repos).
4. Where the users come from (evidence)
- Every listed project gets a "listed" notice issue with a badge snippet (confirmed). The authors themselves opened issues on baoyu-skills #223(10-25), dbskill #53(10-26), digital-human-koubo #2(10-26), gzh-design-skill #27(10-27), WeWrite #39(10-28), xiaohongshu-mcp #885(10-29), guizang-social-card-skill #21, cy-carousel-skill #24 and others, all with the same wording: "Listed and featured in a section, ranked #X... add the badge if you like." These are high-traffic repos with 1k–100k stars, and the issues show up in their issue lists and follower notifications. Listed authors often star the repo too (baoyu replied "Thanks for listing!" and closed the issue 11-01). This is the base for the +7~10 a day from 10-19 to 11-01 (inferred, not reconciled one by one).
- 10-31 @qichen's 1019 X roundup thread on 'Xiaohongshu full-suite operations Skills' (likely, timing matches but not fully verified). The post went out 2026-10-31 22:09 UTC and listed 12 Xiaohongshu skills; the order and star counts closely match this list's Xiaohongshu section. The post summary Brave retrieved includes this repo's link card. We observed 11-01→11-02 +54, the only jump across the whole period.
- Cross-links from sister lists (confirmed): the README 'related resources' section links to seo-geo, short-drama and story. The short-drama list has 1.8k stars, was recommended by GitHubDaily (10-24) and @wlzh, and incidentally sends traffic to this self-media list. We found no separate recommendation of this self-media list from GitHubDaily or any other big account.
- Auto-indexing by skill directories and aggregator sites (confirmed, small volume): skills.sh's social-media-skill-finder page, skillsdirectory.com (Grade A), skillsllm, repolode, kopass, srclog and others.
- The author submitted to other awesome lists (confirmed for the short-drama list: hesreallyhim/awesome-claude-code #3360); none seen for this self-media list.
- Not found: evidence on HN, Reddit, Xiaohongshu, Douyin or ad buying (the local X corpus has 0 hits for both the repo name and the author name; Brave finds no Chinese media coverage).
5. Why it took off (versus similar lists)
- Timing: in the second half of 2026, domestic agents (Qianwen Office, Doubao desktop, WorkBuddy, the OpenClaw 'lobster') all opened skill installation, Xiaohongshu launched RED Skill in July, and 'self-media people installing skills' spread from developer circles into operations circles. There are already hundreds of Xiaohongshu / WeChat Official Account skills on GitHub (a check of our library finds 50+ with ≥300 stars each), so 'which one to pick' became a real need.
- The topic sits at the intersection of two large circles: 'AI agent skills' and 'making money through self-media' are both the topics that get the most traffic on Chinese X and WeChat, and one list serves both audiences.
- More careful than comparable lists: plain-language descriptions (each entry says clearly what it does and where login state is stored), security checks and rejection records, automatic daily refresh, cross-referencing by platform, instructions on installing into domestic agents, and the list is itself a skill. These are the prerequisites for people to share it and for listed authors to claim it.
- Distribution was done deliberately: notification issues + badges turn the 185 listed projects into backlinks and sources of stars. This is the standard awesome-list playbook, but it was executed very systematically (the same template sent out in batches).
- Riding the matrix's momentum: after the short-drama list took off, the author's other lists got a boost too.
6. Differentiation, and whether it was the cause
The differentiation is the single angle of 'Chinese self-media platforms × standard SKILL.md × installable in domestic agents', plus editorial quality through 'verification + security checks + plain-language descriptions'. Large general lists (Composio, VoltAgent) don't cover the niches of Xiaohongshu / WeChat, and the Chinese article-style lists are not updated and don't form a system. This difference is why it could be cited in roundup threads and claimed by listed authors. But the stars mostly came from distribution actions and the momentum of sister repos, not from organic spread driven by the differentiation itself: the seo-geo, novel and story lists built on the same template are just as careful, yet have only 8–49 stars.
7. Visibility
Within small circles the word is: one mid-sized X roundup post may have cited it, it got auto-indexed by a pile of skill directory sites, and the listed authors know about it through their issues. No big account specifically promoted it, no media coverage, and the local X corpus has 0 hits.
8. Copying / following trends
- murphywuwu/awesome-social-media-skills: a full copy (README text word for word, data frozen at 10-26 with 173 projects), not a GitHub fork.
- The short-drama list already has two copies, cd1517 and zhangwenhao66; the author also built a separate shortdramaskills.com site.
- A same-named English repo, IRONICBo / replynodes awesome-social-media-skills, is unrelated to this one (it's SaaS lead-gen).
9. What others can learn
- The 'vertical scenario × curated skills' matrix: one generator (data→build.py→README) lets you open several scenarios a week and bet on whichever one a big account picks up.
- 'You've been listed' notification issues + badges are the cheapest distribution for a cold-start list. The tone has to be like Baoyu's, one that gets a 'Thanks for listing!' reply, not like spam.
- Spell out how to install into domestic agents (the directory paths for Qianwen / Doubao / WorkBuddy), extending the audience from developers to operations people.
- Make the list itself a skill, so once installed in an agent it can be queried. This is a new entry point for list-type projects in the agent era.
- Big accounts only push things with proof: the short-drama list was recommended by GitHubDaily because 'the author ran a control-group blind test'. This self-media list had no such test, which is likely one reason no big account picked it up.
10. Still unclear
- Whether the +54 for 11-02 came from @qichen's 1019 post, some WeChat group / Official Account, or overflow from the short-drama list: we can't get stargazer timestamps, so unverified.
- How many 'you've been listed' issues were opened in total and how many stars they brought in: not counted.
- The author's identity (account registered 2016 years ago, 42 followers, X @UnderleveledDev not found in our corpus) and whether there's a commercial goal (no monetization entry seen so far).
- Growth has slowed back to +7/day at 11-06; whether it can be pushed again later is uncertain.