A LinkedIn skill pack with 3,200 stars, built from one Instagram comment section
What it is An open-source Claude skill pack that helps you write LinkedIn posts, comments and replies, rework your profile, and strip AI-sounding language. It only drafts; it never posts for you.
Why it took off The author is an AI creator with 140k Instagram followers. A Reel that said "comment LinkedIn and I'll send it to you" got 290k views and 9,369 comments, and people who claimed the link were sent to GitHub. His other viral videos have kept sending people there since.
CHow you could follow it Stop making English LinkedIn skill packs. If you have your own short-video audience, build the same structure for Chinese platforms like Xiaohongshu and Douyin, and pair each pack with a "comment to claim" video that leads to GitHub.
- Data (updated daily) · Repo created 2026-09-07 · 3,315 stars · 7 days +331
GitHub · stars Data 2026-10-09 → 2026-11-08 (31 days, 31 observations). Points before 10-28 come from deeply / archive backfill, not every day has one
A GitHub repo with 11 Claude skills for running LinkedIn has collected 3,241 stars in two months. Its stars didn't come from developers finding it on GitHub. They were "claimed" one by one in the comments of a 140k-follower Instagram creator's Reel. The code is easy for anyone to write, and bigger alternatives already exist. What's worth examining is the distribution route, and why it's hard to copy.
A Reel that asks you to comment "LinkedIn"
On Sep 12, 2026, Instagram creator @opusjake (Jake Schincariol, 140k followers) posted a Reel with a single line of copy:
Comment "LinkedIn" and I'll send you this Claude LinkedIn AI Agent skill to manage your entire profile for you.
By Nov 6, this Reel had about 290k views, 5,163 likes and 9,369 comments. Comments are 1.8x the likes, which is rare for organic content: most of them are just the word "LinkedIn", and each one triggers an auto-DM with a link.
The repo was created on Sep 7, but it had zero stars for the first 5 days. On Sep 13, the day after the Reel went out, the stars started climbing.
What it is
An MIT-licensed open-source package that adds 11 commands to Claude: /li-post writes posts using 21 hook formulas, /li-comment writes comments on other people's posts, /li-reply replies to comments on your own posts, /li-profile scores your profile out of 100 and rewrites it, /li-plan plans a week of posting and engagement, plus carousels, repurposing, DMs, inbox triage and post review.
The most carefully built piece is /li-human, the de-AI-ification tool. It ships two dependency-free Python scripts: one strips invisible characters, em dashes and 113 AI-cliché words (delve, leverage, "let that sink in", etc.), and one scores a draft on 5 local checks so you can compare before and after.
The README opens with a very candid line:
Nothing gets posted until you say yes. These skills write. You post.
It doesn't connect to LinkedIn, needs no API key and no sign-up. Each skill ends by handing you a block of text to copy. Installation is designed for non-coders too: paste the repo link into Claude and say something like "install this skill and confirm /li-post works".
No code updates since Sep 17.
How the stars climbed
The path: IG comment keyword → auto-DM → the author's site page for this skill (opusjake.ai/r/linkedin-agent) → GitHub. The page covers the 11 commands and install steps in detail, with no email gate, and you click through to the repo.
The curve shape says a lot. Typical open-source spikes come from Hacker News, GitHub Trending or a big account reposting: hundreds or thousands of stars in a day, then a fall-off. This repo never had that peak. Since Sep 18 it has added a steady 40 to 120 stars a day, with a single-day high of +127 (10-13). The last 7 days: +335; the 7 days before that: +388. It never made the top 25 on Trendshift, and X shows almost no discussion of it in the last ~60 days.
| Date | Cumulative stars | fork |
|---|---|---|
| 10-17 | 2,205 | 328 |
| 10-27 | 2,718 | 423 |
| 11-06 | 3,241 | 523 |
(GitHub, daily records)
The Reel is 8 weeks old, so why is it still growing? The author's site has a resource library with 43 free resources, and every repo is listed there. His other Reels hit big: one on Oct 14 with 2.6M views, one on Nov 1 with 1.02M views. People who came for those landed in the library and clicked through to other repos too. This part is inference, but it explains the steady trickle every day.
Another detail: the fork-to-star ratio is 16%, higher than his other repos (30% for the YouTube one). Developer repos usually sit well below that. Likely non-programmers from IG treating fork as "bookmark" (inference).
Same author, same template, 20+ repos
This repo was the first out of the author's "factory". He then used the same README template ("Eleven free Claude skills that…, Free, MIT") to make a series of versions for Instagram, YouTube, X, trading, founders and more:
| Repo | star(11-06) | Repo created |
|---|---|---|
| replica-skill (clone any app) | 7,853 | 10-03 |
| day-trader-skill | 5,701 | 10-20 |
| instagram-agent-skill | 5,075 | 09-13 |
| youtube-agent-skill | 4,768 | 09-16 |
| linkedin-agent-skill | 3,241 | 09-07 |
| founder-skill | 2,553 | 10-07 |
| arena-skill | 1,285 | 09-27 |
| x-agent-skill | 74 | 09-29 |
Same template, same author, but x-agent-skill has only 74 stars, because it never got a viral Reel behind it. Distribution decides stars, not the topic, and not the code.
How does the author make money? Not from these repos. The site lists custom projects from "$10 to 15k", plus a free Skool community, and brand deals with AI tools like Kimi and Higgsfield in his posts. For him, a repo with a few thousand stars works as social proof.
Who it's competing with
The LinkedIn-writing-skill space is already crowded, and this wasn't the first entry:
| Name | What it is | Compared with alternatives |
|---|---|---|
| sergebulaev/linkedin-skills | Open source, live 2026-04 | 5 months earlier, 5,636 stars, nearly the same features (write posts, comments, de-AI, you approve before posting) |
| Taplio's 26 free Claude skills | A lead magnet from a top LinkedIn SaaS | Tied to their own service; can read data and schedule posts |
| liftli.ai | 31 skills, open source plus a web page | More of them, with a posting skill, plus X and Substack versions |
| stickerdaniel/linkedin-mcp-server | Open source, 4,119 stars | Lets agents read and write LinkedIn, doesn't do writing |
| LinkedIn's built-in AI writing, AuthoredUp, Supergrow | Platform features / SaaS | Paid, used inside the platform |
Someone has even copied one wholesale (avi691/linkedin-agent-skill, identical description).
If it were you
One-line verdict: you can learn a good distribution playbook here, but the repo itself isn't worth copying. Its stars came from the author's 140k followers, which copycats don't get.
Who it's for: developers who already have tens of thousands of followers on a short-video platform and are willing to post weekly. What you lack is something to trade for comments, and a "free skill pack + comment for a link" fills exactly that gap.
Who it's not for: people without their own audience who expect GitHub to recommend the repo for them. The author's own x-agent-skill is the counterexample: same template, 74 stars. Another English LinkedIn skill pack is pointless, since sergebulaev and Taplio already own that spot.
How to do it:
- Switch platform and language. LinkedIn, X and YouTube in English are all covered; Chinese skill packs for Xiaohongshu, Douyin, WeChat Channels and WeChat Official Accounts are still rare. Copy the structure: around ten commands, a library of opening formulas, a working de-AI script, and a voice.md the user fills in first.
- Make installation as simple as "paste the link into Claude", and state in the first screen of the README that it doesn't auto-post or touch accounts.
- Pair each repo with a short video whose caption is "comment X and I'll send it to you". The DM lands on your own landing page, which leads to GitHub and your other resources. One template, several versions, bet on which video hits.
How to get users: through short-video comment sections, not GitHub. If you don't already have an account, build it on the target platform first; the repo comes second.
Why act early: Claude skill packs are still fresh, and hooks like "free AI skills" can still pull a lot of comments on short video. Once every platform has official or SaaS-given free skills (Taplio already is), the hook will feel stale. Probably a matter of months (est.).
Signs you're reading it wrong: if your videos get views and comments but few people who claim the link go on to star the repo, then the value of this route is in growing followers and consulting, not stars. And if this repo keeps adding dozens of stars a day after the author stops posting related videos, then GitHub and skill directories are driving traffic on their own, and this type of project can work without short video.
Research notes: competitor table, acquisition evidence, full research report
- Competitors
Name Where Launched Scale How it differs from this one sergebulaev/linkedin-skills Open source 2026-04-14 5,636 star / 918 fork Same features, launched 5 months earlier and at larger scale. Supports Codex/OpenClaw, drives traffic through developer circles and Awesome lists; it relies on IG for traffic Taplio Claude skills (26) Web Not verified Taplio is a top LinkedIn SaaS Free skills connected to the Taplio MCP, can schedule posts; a lead magnet for the SaaS liftli.ai LinkedIn agent skills (31) Open source Not verified Not verified More skills, with a publish skill and X/Substack versions, installed via npx linkedapi.io skills Other Not verified Not verified Paid API that actually operates LinkedIn posting and outreach stickerdaniel/linkedin-mcp-server Open source 2025-04 4,119 star MCP that gives agents read/write access to LinkedIn data, doesn't handle writing LinkedIn's built-in AI writing / AuthoredUp / Supergrow Big-company features Already live Big Paid, inside the platform, doesn't run in Claude - Where users come from
- 查实 Instagram Reel 评论关键词自动私信:@opusjake 13.97 万粉;2026-09-12 Reel「Comment 'LinkedIn'…」29.26 万播放、9,369 条评论;仓库 9/13 开始有 star(TikHub IG 接口、Trendshift)
- Inference Cross-traffic from the opusjake.ai resource library: the library has 43 free resources, every repo is listed there; his other viral Reels (arena 2.6M, clone 1.02M views) keep sending people across the site, and this repo still adds +8~40 a day 60 weeks later
- Verified Skill directories / small-blog SEO: SkillsLLM, skills-hub.ai, ClaudePluginHub, Skill Leaderboard, and review posts on scalewiseva and digicuratoragency all list it
- Inference X / HN / TikTok: x-radar shows 60 mentions in 0 days; no TikTok account for him; never made Trendshift's top 25
Jakeschincariol/linkedin-agent-skill: why it runs
Research date 2026-11-06. Channel: GitHub. Data card: gh_Jakeschincariol__linkedin-agent-skill.json
Bottom line
It's not an open-source project that grew on its own. It's one of the landing pages in the comment-keyword lead funnel of Instagram creator @opusjake (Jake Schincariol, 140k followers). He posts a Reel, a commenter writes "LinkedIn", and an automated DM sends them opusjake.ai/r/linkedin-agent, which redirects to GitHub. The stars are a side effect of IG traffic, which is why the curve stays flat for 60 days, with only 40–120 a day and no single-day spike. The same author used the same template for 20 other "11 free Claude skills" repos, and 7 of them passed 1000 stars. He makes money from consulting (custom projects starting at $1–15k), the Skool community Opus Lab, and brand deals with AI tools.
What it is
- 11 Claude Skills (SKILL.md folders) for running a personal LinkedIn account: write posts (
/li-post, with 21 opening formulas), comment on others' posts, reply to comments on your own posts, score your profile, plan the week, make carousels, repurpose content, write DMs, triage your inbox and review published posts. There's also/li-humanfor removing AI-speak, with two dependency-free Python scripts:humanize.pystrips invisible characters, em dashes and 113 cliché words;detect.pyscores text on 5 local heuristics. (Source: README) - MIT license, no sign-up, no API key, doesn't connect to LinkedIn. It explicitly states it won't post for you, citing that automation violates LinkedIn's user agreement, so every skill ends by handing you text to copy.
- Repo created 2026-09-07, last code push 2026-09-17, no updates since (data card).
Data
| Metric | Value | Source |
|---|---|---|
| star / fork | 3,241 / 523(2026-11-06) | Data cards, github repo_day |
| last 7 days / previous 7 days | +335 / +388 | Data card |
| 60-day daily gain | 9/8–9/12 was 0; from 9/13 it averaged 7–31 per day; from 9/18 it has held steady at 40–120; peak +120(10/8), +117(10/12) | Trendshift |
| Trendshift top 25 | never once | Same as above |
| GitHub trending | Our DB has trending_days_any=0 (tracking only started at 10-28) | Data card |
Curve shape: no spike, just a long, steady trickle. A typical open-source project gets hit by Hacker News, GitHub Trending, or a big account reposting it, then surges and falls off. This curve doesn't look like that.
Growth driver: Instagram comment-keyword funnel (confirmed)
- The author is an AI tools creator on Instagram. @opusjake has 139,703 followers and 136 posts. His bio reads "I build Skills & Agents… ⏬ FREE AI Skills & resources", linking to opusjake.ai (TikHub instagram/v2 endpoint, pulled 2026-11-06).
- Nearly every Reel says "Comment 'X' and I'll send you…" (the kind of comment-triggered auto-DM you get from ManyChat). I pulled his 2026-05-24 through 11-06 posts, 158 total, and most follow this format.
- The LinkedIn Reel: posted 2026-09-12, caption "Comment 'LinkedIn' and I'll send you this Claude LinkedIn AI Agent skill…", 292,584 views, 5,163 likes, 9,369 comments (post code DdAAUCdtUXO). Comments are 1.8x likes, so most commenters are asking for the link. The repo started getting stars 9/13, which matches the Reel's timing. 2026-09-25 also posted a ChatGPT version of the LinkedIn Reel, which got only 36 views and barely did anything.
- DMs send people to opusjake.ai/r/linkedin-agent, not straight to GitHub. That page explains things in detail, then links to the repo. opusjake.ai/library has 43 free resources total, and every repo is listed there. So when another Reel takes off (his 10/14 arena post hit 2.6M views, the 11/01 clone post 1.02M, the 9/28 Instagram post 380k), people who come to the library also star the LinkedIn repo on their way out. That explains why the LinkedIn repo keeps gaining 40–60 per day steadily, 8 weeks after its Reel.
- Conversion path: Instagram comment → auto-DM → opusjake.ai resource page → GitHub star/fork. On the other side: the Skool community Opus Lab (free, 100+ members) → book a consultation ("Custom builds start at $10 to 15k", opusjake.ai). His posts also include brand deals with @kimiai.official, Higgsfield's "3-day free trial," and Skywork. That's another income stream.
- Long-tail distribution: auto-indexed by skill directories, including SkillsLLM, skills-hub.ai, ClaudePluginHub, Skill Leaderboard, and Claude Code Club. A few small blogs have written reviews (scalewiseva 9/14, digicuratoragency 9/18). All low traffic.
- Channels I couldn't find: searching the x-radar corpus for "opusjake," "linkedin-agent-skill," and "Jakeschincariol" returned 60 hits over 0 days. He has no TikTok account. Never on Hacker News, and never in the Trendshift top 25.
The author's "factory" (key background)
Same account, same README template ("Eleven free Claude skills that …, Free, MIT"), same Instagram keyword funnel:
| Repo | star | Repo created |
|---|---|---|
| replica-skill (clone any app) | 7,853 | 10-03 |
| day-trader-skill | 5,701 | 10-20 |
| instagram-agent-skill | 5,075 | 09-13 |
| youtube-agent-skill | 4,768 | 09-16 |
| linkedin-agent-skill | 3,261 | 09-07 |
| founder-skill | 2,553 | 10-07 |
| trader-skill | 1,412 | 10-14 |
| arena-skill | 1,285 | 09-27 |
| x-agent-skill | 74 | 09-29 |
| closer / seed / whale (newly created in November) | 32 / 41 / 6 | 11-01~11-06 |
(GitHub API,2026-11-06)
The pattern: whichever repo has a breakout Reel behind it reaches a thousand stars. x-agent-skill uses the same template but had no Instagram hit behind it, so it has only 74 stars. Distribution is what worked, not the LinkedIn topic or the code. linkedin-agent-skill was the first "11 skills" template in this factory, and the later repos are copies of it.
Competitors
| Name | Form | Launched | Scale | How it differs |
|---|---|---|---|---|
| sergebulaev/linkedin-skills | Open source | 2026-04-14 | 5,636 stars / 918 forks, +281 in the last 7 days (GitHub repo) | Launched 5 months earlier and bigger: 12 skills, the same "write posts, comment, de-AI the text, wait for your approval" flow, supports Codex/OpenClaw, and the author runs a full X/IG/YouTube setup. Nearly identical feature set, driven by Awesome Claude Code listings and developer-community referrals. |
| Taplio "26 free Claude skills" | Free lead magnet for a web + commercial SaaS | Not verified | Taplio is a top LinkedIn SaaS | Skills connect to the Taplio MCP, so they can read your data and schedule posts. The free skills are the SaaS's acquisition hook. |
| liftli.ai "31 free LinkedIn agent skills" | Open source + web | Not verified | Not verified | More of them, installed with npx skills add, including a publish skill. Also has X/Substack versions. |
| linkedapi.io/skills | Commercial API + skills | Not verified | Not verified | Can actually operate LinkedIn (posting, outreach). Paid. |
| stickerdaniel/linkedin-mcp-server | Open source MCP servers | 2025-04 | 4,119 star | An interface that lets agents read and write LinkedIn. Doesn't handle writing. |
| aaitor/skill-publish-to-linkedin | Open source | Not verified | Small | Posts directly through the official Posts API |
| AgriciDaniel/claude-ads, borghei/Claude-Skills | Large open-source collections | 2026-01/02 | 10k / 1k stars | LinkedIn is only one part |
| LinkedIn's built-in AI writing (Premium), AuthoredUp, Supergrow, and others | Big-company features / SaaS | Already live | Big | Paid, inside the platform, doesn't run in Claude |
| Ruben Hassid's "Claude + LinkedIn" tutorial | Item | first half of 2026 | LinkedIn power user | Teaches you to build your own skills instead of giving you a finished package |
Bottom line: there are many competitors, and they came earlier. On features it has no moat. sergebulaev's version is nearly identical and has more stars.
Differentiation
- On the product side, two things stand out:
/li-humanincludes two working Python scripts with before-and-after scoring, and the README is candid ("fine print, which is the honest part"), not claiming it can "beat detection." People who know the space find that credible, but neither is why it took off. - The real difference is distribution. Most competitors rely on GitHub, developer communities, and SEO. This one relies on a 140k-follower Instagram account aimed at non-programmers, plus keyword DMs. Many of these people don't use Claude Code, so the README's first step is "paste this link into Claude and let it install itself." Fork-to-star ratio is 16%, and the author's other repos are just as high (youtube 30%, day-trader 25%). My guess, unverified, is that non-programmers from Instagram treat forks as bookmarks.
Visibility
It's spreading in small circles. A 290k-view Instagram Reel put it in front of ordinary users, and skill directories and a few small blogs picked it up. But there's no discussion on X or Hacker News, it never reached the Trendshift top 25, and there's no press coverage. The author hasn't disclosed any revenue. The LinkedIn-writing-skill space is already crowded, though: Taplio, sergebulaev, and liftli are all in it.
Copycats
- Followers have appeared: Taplio uses 26 free skills as a lead magnet, and liftli.ai shipped 31 skills with X/Substack versions. I can't confirm they copied this project, and I haven't verified launch dates.
- Some people have directly reposted it: avi691/linkedin-agent-skill has an identical description.
- The biggest "copier" is the author: the same template was duplicated into 20+ repos, including Instagram, YouTube, X, ChatGPT-YouTube, founder, and trader.
Takeaways
- Instagram/short video + comment-keyword auto-DM, with a landing page pointing to an open-source GitHub package: free open source makes people willing to comment, the comments boost engagement and help the Reel get recommended, and GitHub stars then serve as trust signals ("5000+ stars") for the consulting business.
- One template, shipped in batches: the "11 skills + 21 hooks + humanizer + voice.md" structure works for any platform or profession. Each new repo gets a Reel, and you bet on which one takes off.
- Zero-friction install: "paste the repo link into Claude and let it install itself," which is very friendly to non-programmers.
- Be upfront about the limits: no auto-posting, and de-AI-ing is just a heuristic score. That avoids account-ban risk and overpromising, which helps reputation.
Unknowns
- Before 9/12, the star count was 0 (Trendshift), and it only started climbing from 9/13. I couldn't verify whether the DM sends people to opusjake.ai or directly to GitHub, or whether it asks for an email (I can't access the DM content).
- Star manipulation: the curve is smooth and the fork ratio is somewhat high, but there are no abnormal spikes. The GitHub stargazers endpoint was rate-limited this time, so I couldn't sample account quality.
- The author's income is not public. Consulting volume, whether Skool has a paid tier, and brand-deal rates are all unknown.
- I haven't verified launch dates for the Taplio and liftli pages.