Stringers
Jakeschincariol/instagram-agent-skill · Jakeschincariol

"Comment on Instagram and I'll send you the link": how a 140k-follower creator turned an open-source repo into the freebie for a Reel, 54 days to get 5,000 stars

What it is 13 free Claude skills plus 5 small local scripts that help you write Instagram Reel scripts, score openings, check captions, find viral content and remove AI-sounding phrasing. It writes only; it never posts.

Why it took off The author @opusjake has 140k Instagram followers; his Reel promoting it got 379k views and 15k comments; commenting a keyword triggered a DM with the link. Afterward, TikTok creators in Spanish, French, Russian and other languages used it as a free hook to spread it further, and the star count kept climbing without fading.

CHow you could follow If you have your own short-video account, build the ops skill pack for whatever platform you promote on (e.g., a RedNote pack for RedNote). Make it free, one-line install, with a scoring tool that people can screenshot, use "comment a keyword to get the link" to drive traffic, and hand it to creators in your niche to repost.

C grade: nearly all stars came from the author's 140k-follower funnel and reposting creators; late followers get nothing. Copycat packs on the same template and the author's own X version only have a few dozen stars, and the English-language slots are already taken. Someone with a short-video account can try it and it's worth a shot, but for most developers following it isn't worth the effort. Word of mouth in niche circles
Data (updated daily) · Repo created 2026-09-13 · 5,314 stars · 7 days +1,164
GitHub · stars
02k4k6k10-0910-1610-2411-0111-085.3k
Data 2026-10-09 → 2026-11-08 (31 days, 30 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

In Sep 13, an AI creator on Instagram pushed a repo to GitHub: 13 Claude skills for running an Instagram account, free, MIT. The repo has not had a single line updated since, 27 issues sit unanswered, yet the stars have kept rising every day, reaching nearly 5,000 stars and 1,020 forks by Nov 6. It wasn't discovered in developer circles; it was picked up one short video at a time on Instagram and TikTok. Below: what it is, where the stars came from, why the same author's other repos only got a few dozen stars, and how to think about building a project that can earn stars the same way.

A Reel with 379k views and 15k comments

The author is Jake Schincariol, Instagram account @opusjake, 11-06 followers, 139,714; his bio reads "I build Skills & Agents to 10x your life". The Reel promoting this repo had this caption:

Comment "Instagram" and I'll send you the link to turn Claude into your personal social media manager…

As of 11-06, this video has 379k views, 8,788 likes, and 14,753 comments. Comments outnumber likes because commenting the word "Instagram" triggers an automatic DM containing the GitHub link; each comment is roughly one delivery, and it also feeds the algorithm for this video.

His whole account follows this formula: a LinkedIn version (9-12.3M views, 9,370 comments), a YouTube version, a "clone" video (1.02M views), one at 2.6M views... nearly every Reel ends with "Comment 'XXX'". He has 19 repos on GitHub, 5 with over a thousand stars, 32k stars in total. The repos are all named "X-skill" and their READMEs all read "N Claude skills / Free, MIT", like a production line shipping one pack a week.

What it actually is

The README's first line: "Thirteen Claude skills that run an Instagram account. Free, MIT, no signup, no API key, nothing to connect."

13 slash commands cover the daily work of running an IG account: /ig-reel writes three openings from 26 hook formulas, scores them, and outputs a script with a second-by-second beat sheet; /ig-viral finds Reels trending in your niche, ranked by "views ÷ the account's own median"; /ig-caption shows you the 125 characters before "… more" in the feed; there's also profile scoring, a weekly plan, comments, DMs, and de-AI-ification. 5 additional zero-dependency Python scripts run locally, and their output looks like this table:

`` -> 85.6 STRONG Nobody tells you that your first 30 reels are supposed to f... 9.6 WEAK Hey guys, today I wanted to talk about content strategy dealbreaker: Greeting. Nobody came to the feed to be greeted. ``

It only writes, never posts, and doesn't connect to your account. Installation is "paste the repo link to Claude and say 'install this skill, then confirm /ig-reel works'", so people who don't know git can still install it.

Where the stars came from: the author's funnel, plus reposting creators across a dozen countries

Star curve (GitHub, daily): 10-16 1,458 → 10-28 3,415 → 11-06 4,963. Over the last 7 days +1,293, the first 7 days +1,372, barely fading; the single-day peak was +381 on 10-28.

Two forces:

The author's own Instagram. That Reel above and his follower pool were the starting point. On 10-28 his Instagram pack gained +381, YouTube pack +330, replica pack +406, LinkedIn pack +90, all peaking the same day. A traffic event at the level of the author's account lifted the whole row of repos together; which specific video or mass message did it, I couldn't pinpoint.

Other people's reposts. Searching TikTok turns up a dozen or so introduction videos by other creators, in Spanish, French, Russian, Arabic and English (TikTok, 11-06):

CreatorDateViewsLanguage
samdespo (720k followers)10-19667kEnglish (the caption matches, no link posted, inferred to be it)
maks_reels10-29185kRussian
xabatzz (330k followers)10-1838kSpanish, names "Instagram Agent Skill" directly
antonspain.ia10-2231kSpanish, tells people to search the name on GitHub and download the zip
damien.mapt10-0130kFrench, "comment INSTA" to get the guide

These people also use "comment a keyword to get the link" to grow their own followings. A free, useful, open-source pack you can film right away is a ready-made hook for them. The daily star rate rising from 10-18 to 10-22 between 100 and 250 lines up with the timing of these Spanish and English videos.

Almost no one is discussing it on X, Hacker News or Reddit.

The people coming in aren't programmers

Forks/stars around 20%, versus a median of about 11% for new repos of similar size. Of the last 30 forkers, 13 are GitHub accounts registered after September, many created the same day, with only this one repo, and usernames like the auto-suggested xxx-cmyk or xxx-pixel. The issues are things like "Instagram" and "pasted the install command again" (GitHub, 11-06).

This is a group of Instagram users who signed up for GitHub specifically to get this thing. Star-farming accounts usually don't file messy issues like these, so they look more like real beginners than fakes (the stargazer list wasn't available, so this can't be fully ruled out).

Same template, why do some get 74 stars

The author's X version of the skill pack has 74 stars; hunter-skill has 80 stars. Template, copy, and author are all the same; the difference is whether the platform fits. Promote "help you run Instagram" on Instagram and the people watching are the users themselves; promote "help you run X" on Instagram and the audience doesn't care.

Another comparison is the copycat postzen-dev/instagram-agent-skill (10-06): 13 skills verbatim plus a skill that posts through its own SaaS, with a TikTok promo video at 65k views and 42 repo stars. The stuff is about the same, but it lacks the author's follower pool and the shape that makes repost accounts feel it's "free, clean, safe to share."

What the product actually does is make distribution possible: single platform, single role, explainable in one sentence (easy to film as "Claude just killed Instagram"); terminal output you can screenshot, so the video has visuals; free with no sign-up, so other creators dare to repost it and viewers can install it. These alone aren't enough, but without them the funnel won't turn.

The author is going after something other than this repo. The homepage of his site opusjake.ai uses "5,000+ GitHub stars" as social proof, below it reads "Custom builds start at $10 to 15k," and there's also a free Skool community and brand partnerships. Stars are a business card; followers are the traffic pool.

Who's already doing this

ProjectScaleHow it differs
coreyhaines31/marketingskills59K starsGeneral marketing skills, audience is developers and growth people
sergebulaev/linkedin-skills5.6k starsSame shape, LinkedIn
ScrapeCreators/social-media-research-skills5.8k starsOnly does hit-product research, needs their scraping API
op7418/guizang-social-card-skillAbout 8k starsChinese AI blogger Guizang's social media card skill, same creator-driven traffic
vivy-yi/xiaohongshu-skills530 starsXiaohongshu-focused, no big-account distribution
Same-name packages like postzen-dev / ishort-pro / AIProUKDozens of starsCopycats, some turned into an entry point for their own SaaS

GitHub star counts checked live before and after 11-06. For the "platform operations skill pack" slot, the English mainstream platforms (IG, LinkedIn, X, YouTube) are basically all taken; on Chinese platforms, bloggers have built traction, but I haven't seen a "full operations suite" for Douyin, Xiaohongshu, or WeChat Channels each take off like this (search not exhaustive).

If it were you

In short: this repo's stars were pulled in by a funnel from an account with 140k followers; anyone can write the code, but the follower pool can't be copied. Without your own short-video account, copying it will most likely get you postzen-level dozens of stars.

Who it's for: developers who already have a short-video account (even a few thousand to tens of thousands of followers) and talk about AI or operations; or people who can write and are willing to film videos every day. Being a non-native English speaker is actually fine: Spanish, French, and Russian bloggers are all reposting the English pack, which shows the local-language native version is empty.

Who it's not for: people who just want to write code and expect to be discovered organically through GitHub Trending or Hacker News. This kind of thing barely makes a sound in developer circles; its users are on Instagram and TikTok.

How to do it: pick one platform and build the operations pack for that platform only: a Xiaohongshu operations skill for Xiaohongshu, a Douyin skill for Douyin. Follow its shape: free, MIT, zero dependencies, no account linking, installed by pasting the link to Claude with a one-line "install this." Include at least two or three working, screenshot-worthy tools (hook scoring, character-count window, beat sheet), and write the creator methodology for that platform into them, rather than a pile of prompts. The first sentence of the README should say what it does.

How to get users: post a vertical video demoing the scoring table in the terminal, ending with "comment XX and I'll DM you the link," paired with an auto-DM tool; then proactively hand the repo to AI bloggers in the same field, since they need free hooks. Also submit the repo to listing sites like skills.sh and skill leaderboards.

When to move: the Claude skill format is still new, and the slots on each platform are being taken one by one; the English ones are already full. Chinese platforms and non-English languages may still have a month or two (my inference).

Signs you're misreading it: ① your first promo video gets comments below 1% of its views, meaning that platform's audience isn't buying "AI runs your operations"; ② the repo has fewer than 300 stars after 30 days while similar projects on the same platform took off first; ③ the platform itself (Xiaohongshu, Douyin) launches a built-in creator assistant of the same kind, putting "hook scoring and title checks" into the publishing page.

GitHub repo ↗

Research notes: competitor table, acquisition evidence, full research report
Competitors
NameWhereLaunchedScaleHow it differs from this one
coreyhaines31/marketingskillsOpen source2026-0158.7k starsGeneral marketing skills for developers/growth people; it only does Instagram, is explainable in one sentence, and suits short-video spread
sergebulaev/linkedin-skillsOpen source2026-045.6k starsSame shape but the platform is LinkedIn
ScrapeCreators/social-media-research-skillsOpen source2026-065.8k starsOnly research and depends on a paid scraping API; it has zero dependencies and covers the full writing workflow
ZJU-REAL/EaselOpen source2026-084.4k starsA complete publishable agent aimed at technical users; it only writes, doesn't post, and is installed by beginners with one sentence
Ootto-AI/claude-content-skillsOpen source2026-06397 starsAlso a IG Reels skill, but tied to NotebookLM, no big-account distribution
postzen-dev/instagram-agent-skillOpen source2026-10-0642 starsCopycat version, adds a PostZen publishing skill as SaaS traffic capture
Metricool / Composio and other MCP connectors + ClaudeOther2026TikTok promo at the 1–100k view level (user count unverified)Actually connects accounts to pull data / schedule posts, requires paid signup; it has zero connections and zero accounts
Later / Buffer AI Assistant / Predis.ai / ManyChatWebYears agoLarge (unverified)Paid subscription, includes publishing and automation; it's free and only outputs drafts
Built-in AI from Meta AI / Instagram EditsBig-company features2025–2026All IG usersGeneral in-platform generation, without creator methodology like hook scoring, the 125-character window, or de-AI-ifying text
Where users come from
  • Verified The author's Instagram @opusjake (140k followers): comment-keyword → auto-DM link: Reel DdNW01LN8R9 "Comment "Instagram" and I'll send you the link…" 379k views, 14,753 comments (TikHub 2026-11-06); the whole account's Reels use the same template
  • Verified Third-party TikTok/IG repost accounts (Spanish/French/Russian/Arabic/English): samdespo 667k views (10-19, copy matches but no link posted), maks_reels 185k, xabatzz 38k naming the Instagram Agent Skill, antonspain.ia 31k, damien.mapt 30k, web.tactics posting the repo link directly, etc. (TikHub TikTok search)
  • Inferred A traffic event at the author's account level is driving the whole skill lineup: 10-28 instagram +381, youtube +330, replica +406, linkedin +90 all peaked the same day (our github repo_day); trigger point not identified
  • Verified Long-tail SEO / aggregator sites / press releases: AlphaSignal article the day the repo was created (utm_source=alphasignal), TALC AI install guide, skills.sh, mcpinstall, skillleaderboard, Substack reviews
  • Inferred X / HN / Reddit / ads: x-radar shows no related discussion in the last 60 days; no HN/Reddit posts or ads spotted

Jakeschincariol/instagram-agent-skill: why it got traction

Channel: GitHub | Repo created 2026-09-13 | As of 2026-11-06: about 4,963–5,075 stars, 1,020–1,040 forks (our github repo / GitHub API checked live) | Research date 2026-11-06

Bottom line

This isn't an open-source project that grew out of a developer community; it's the lead magnet of an AI Instagram creator with 140k followers (@opusjake): the author used a Reels funnel of "comment a keyword → auto-DM the GitHub link" to send non-programmer viewers to GitHub, and then a dozen-plus multilingual TikTok/IG repost accounts (Spanish, French, Russian, Arabic, English) spread it for free, pushing the star count up steadily. The product itself (13 Claude skills + 5 local zero-dependency Python tools) is more detailed than comparable projects, but what drove the surge is distribution, not code.

1. What it is

  • 13 Claude Code skills (/ig-reel, /ig-viral, /ig-caption, /ig-carousel, /ig-story, /ig-profile, /ig-plan, /ig-human, /ig-comment, /ig-reply, /ig-dm, /ig-repurpose, /ig-audit), plus 5 zero-dependency Python scripts: hook scoring hookscore.py (26 hook formulas), spoken-narration beat sheet beats.py, caption check caption.py (simulates the 125 characters before "… more" in the feed), de-AI-ification humanize.py/detect.py, viral benchmarking swipe.py (ranks by "views ÷ that account's median"). MIT, free, no API key needed, doesn't connect to Instagram, doesn't auto-post. (README, GitHub)
  • Installation is very beginner-friendly: "paste the repo link to Claude and say 'install this skill'"; a plugin marketplace is also supported.
  • The repo has only 2 commits, pushed_at stopped on the day it was created (2026-09-13 14:10), and 27 open issues have no replies: a typical "publish and walk away" content asset.

2. Data

  • Our github repo_day: 10-16 1,458 stars → 11-06 4,963 stars; last 7 days +1,293, the first 7 days +1,372, barely decaying; single-day peak +381 (10-28).
  • Fork/star ≈ 20.5%, about 11% higher than the median for new repos of similar size (per the SCHEMA notes).
  • On 10-28 the author's multiple repos peaked simultaneously: instagram +381, youtube-agent-skill +330, replica-skill +406, linkedin +90 (repo_day), which indicates a traffic event at the author's account level driving the whole lineup, rather than this repo being discovered on its own.
  • Of the 30 recent forkers, 13 are GitHub accounts registered after 2026-09, many registered the same day with only 1 repos, and usernames with the auto-suggested xxx-cmyk / -pixel / -glitch / -sys suffixes (checked live via GitHub API 2026-11-06). The issues section is full of invalid entries like "Instagram", "Borah", and "pasted the install command again." → The people coming in are non-developers who registered on GitHub specifically to get this, which matches the social-media-funnel profile; it doesn't look like star-botting (bots usually don't file this kind of garbage issue).

3. Where users came from (growth tactics)

3.1 The author's own Instagram "comment keyword → DM link" funnel (confirmed)

  • Author Jake Schincariol, Instagram @opusjake, 139,714 followers (TikHub 2026-11-06), bio: "I build Skills & Agents to 10x your life … FREE AI Skills & resources".
  • The Reel promoting this repo: instagram.com/reel/DdNW01LN8R9, caption "Comment "Instagram" and I'll send you the link to turn Claude into your personal social media manager…", 379k views, 8,788 likes, 14,753 comments (TikHub, 2026-11-06). The abnormally high comment count comes precisely from "you need to comment the keyword to get the link"; each comment ≈ one automatic DM delivered.
  • His whole account runs on this playbook: almost every Reel is "Comment "XXX" and I'll send you…". The same template was pushed to LinkedIn (9/12.3M views, 9,370 comments), YouTube (9/25), a replica "clone" (1.02M views, 14,806 comments), founder (813k), arena (2.6M views), day-trader and more. On GitHub there are 19 repos, 5 of them past 1k stars, 32k stars combined.
  • Every skill pack follows the same naming: "X-skill / 11 (or 13) Claude skills / Free, MIT". It's a repeatable production line.

3.2 Third-party reposter creators (confirmed, and the main engine of ongoing growth)

Searching TikTok (TikHub, 2026-11-06) turned up plenty of explainer videos by others, most naming the repo or pasting the link:

CreatorDateViewsLanguageNotes
samdespo (720k followers)10-19666,571EN"Claude can now manage your Instagram account…", no link posted, content matches this repo's features (probably it, 100% unverified)
maks_reels10-29184,517RussianSMM skill pack test, comments ask for the link
xabatzz (330k followers)10-1837,826SpanishNames the "Instagram Agent Skill" directly, 13 skills
antonspain.ia10-2231,030SpanishTells people to search GitHub for "instagram agent skill claude" and download the zip
damien.mapt10-0129,828FrenchNames Instagram Agent Skill, 13 skills, comment INSTA to get the guide
nicolemccain_ai / joshfpocock / aiwithavi / metaengines10-09~10-223k–7kENAll describe 13 skills
web.tactics / johann_andre / bizi.beast / ai.with.ni / adamamira.ia / aiman.salah910-10~11-030.1k–4kEN/FR/ARSome paste github.com/Jakeschincariol/instagram-agent-skill directly

These creators use "comment a keyword to get the link" to grow their own followings too. A free, useful, hook-friendly open-source skill pack is ready-made material for other AI creators, which set off multilingual secondary spread. The acceleration in stars for 10-18~10-22 (daily gain 100→250) lines up in time with the videos from xabatzz, samdespo and antonspain.

3.3 Long-tail SEO / aggregator sites (confirmed to exist, small in scale)

AlphaSignal ran a press-release-style article the day the repo launched (with utm_source=alphasignal). TALC AI install guides, hysenlabs, mcpinstall, skills.sh, skillleaderboard and abhijeetdesigns all index it. A Substack post "I Let Claude Run My Instagram Account" (10-14) calls it "the most complete one after research".

3.4 what I didn't find

X (our x-radar corpus) has no relevant discussion in the last 60 days; none on Hacker News or Reddit; no ad evidence.

4. Where the money comes from

The repo is free, so monetization happens on the author's side: the opusjake.ai homepage says "Custom builds start at $10 to 15k" for consulting, plus a free Skool community, Opus Lab (100+ members), a newsletter, and brand deals (his Reels @-tag partners like Luma, Manus, Higgsfield, Kimi, Alibaba Accio). The homepage uses "5,000+ GitHub stars" as credibility. GitHub stars = social proof, Instagram followers = traffic pool, consulting/sponsorship = revenue.

5. Competitors

NameWhereLaunchedScaleHow it differs
coreyhaines31/marketingskillsOpen source2026-0158.7k stars, 8.6k forksGeneral marketing skills (CRO/SEO/copy), not Instagram-specific; audience is developers and growth people
sergebulaev/linkedin-skillsOpen source2026-045.6k starsSame shape but the platform is LinkedIn
ScrapeCreators/social-media-research-skillsOpen source2026-065.8k starsResearch only (viral content, comment mining), needs their scraping API
ZJU-REAL/EaselOpen source2026-084.4k starsFull agent, can publish; heavy, aimed at technical users
Ootto-AI/claude-content-skillsOpen source2026-06397 starsAlso an IG Reels skill, but tied to NotebookLM, with no large creator distribution
postzen-dev/instagram-agent-skillOpen source (SaaS lead magnet)2026-10-0642 stars; promo videos on TikTok with 65k viewsCopycat: 14 skills = 13 plus a publishing skill that posts, comments, DMs and analyzes through PostZen, turning the free pack into an entry point for their own SaaS
moboutrig/instagram-claude-skillOpen source2026-0214 starsManages accounts via the Meta Graph API, requires developer credentials
MCP + Claude from Metricool / Composio and othersOther (SaaS connectors)2026On TikTok, creators like itsmariahbrunner and jayanthi.ai promo it at 1–100k viewsConnects real accounts to pull data and schedule, requires signup and payment; this repo has zero connections and zero accounts
Later, Buffer AI Assistant, Predis.ai, ManyChatWeb SaaSYearslarge (specific numbers unverified)Paid subscription, with posting and automation; this repo only writes, doesn't post, and is free
Built-in AI from Meta AI / Instagram EditsBig-company features2025–2026Full user baseGenerated inside the platform, without the creator methodology of "hook scoring, 125-character windows, de-AI-ing"

6. Differentiation, and whether it was the reason for the growth

  • Product differentiation: it turns the "Instagram creator methodology" into executable checks: scoring against 26 hook formulas, timing voiceover beats, a 125-character visible window, a cap of 5 hashtags, viral multiples computed against the account's median, and a humanizer that strips em dashes and watermark characters. "Writes only" and "no account connection" lower both the psychological barrier for beginners and fear of account bans. Compared with general packs like marketingskills, it is single-platform, single-role (your IG operator), which can be explained in one sentence. That fits a short-video title like "Claude just killed Instagram".
  • Is it the reason for the traction? Partly. The real driver is distribution (the author's 140k-follower funnel plus multilingual reposting). But the product form is the precondition for distribution to work: free, MIT, no signup, "paste the link to Claude and it installs", which lets other creators repost safely and lets viewers install it. "Instagram operations" is also a pain point for the IG audience itself (promoting an Instagram tool on Instagram, with a fully matched audience). Counterexamples from the same author, x-agent-skill (for X, 74 stars) and hunter-skill (80 stars), show that the same template doesn't take off when the platform audience doesn't match.

7. Visibility and copycats

  • Visibility: spreading within small circles. It has spread widely in AI creator circles and the IG/TikTok multilingual community (dozens of reposting videos, the top one at 667k views), but barely on X, HN, Reddit or mainstream tech media. No revenue is publicly visible (it's free by design), and the author hasn't shared any income figures.
  • Copycats: PostZen forked the idea into a "14 skills" SaaS lead-magnet version and runs promo content on TikTok. ishort-pro/instagram-claude-skills (11-02), AIProUK, androskent-cell (SocialMaker connector) and other same-named packs keep appearing. The author is also cloning it themselves (LinkedIn, YouTube, X, a ChatGPT-YouTube version).

8. Takeaways

  1. "Comment keyword → auto-DM the link" funnel: the GitHub repo becomes the prize for an IG Reel. Comment volume equals reach, and it also feeds the algorithm.
  2. Designed to be reposted: free + MIT + no signup + one-line install + terminal output that screenshots well (a hook-score table, a 125-character box), so other creators can film it immediately.
  3. Platform match: promote on a platform with a tool for that platform (IG tools promoted on IG, LinkedIn tools on LinkedIn).
  4. Productized: uniform naming and uniform "11 skills / Free, MIT" copy, one pack a week. Star counts build personal-brand credibility, which pays back in consulting ($10–15k) and brand deals.

9. Uncertain

  • Whether samdespo's 667k-view video is this repo (the copy matches, but no link was posted).
  • The specific trigger for 10-28 multiple databases peaking at once (possibly a big creator's Reel or a Skool/newsletter blast) is not pinned down.
  • Stargazer details could not be retrieved (the API returned 404), so star inflation can't be fully ruled out, but the fork profile looks more like real beginners.
  • The author's consulting and sponsorship income is not public, so the return can't be quantified.

Sources

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.