Replaying one Tesla trade ten thousand times: how a Claude skill pack got 5500 stars in 17 days from other people's short videos
What it is 11 intraday trading skills for Claude: chart scanning, stop-loss setting, position sizing, plus a Time Machine that replays the market ten thousand times before you place an order. Free, no key, one-sentence install.
Why it took off "Replay it ten thousand times before you trade" fits into a 30-second video, and installation is simple enough for non-programmers to follow. So a crowd of AI creators on Instagram and TikTok used it as lead-gen material and funneled a wave of retail traders to GitHub.
BHow you could follow it Pick a mass-market topic (stock trading, job hunting, growing followers), build one filmable selling point, package it as a one-sentence-install skill pack, port it to another market (India's NSE, China A-shares, crypto), and proactively hand the material to mid-size creators who run DM funnels.
- Data (updated daily) · Repo created 2026-10-20 · 5,937 stars · 7 days +1,994
GitHub · stars Data 2026-10-22 → 2026-11-08 (18 days, 17 observations). Points before 10-28 come from deeply / archive backfill, not every day has one
In 2026 Oct 20, an AI consultant on Instagram pushed a repo to GitHub and never touched a line of code again. 17 days later it had 5530 stars and 1368 forks, and most people didn't come from GitHub. They came from short videos that said "comment TRADE and I'll DM you the link." The code is nothing special. The moat is packaging and distribution. Below is how it worked, which steps you can copy, and which one you can't.
A trade that's close to a coin flip
The example in the repo README is "the trade in the Reel." On Monday, Oct 19, 2026 year, at 10:15 in the morning, Tesla's 5-minute candle closed at 364.18, breaking out of the first 15-minute range. The scanner picked it out of 231 charts, and the entry, stop at 358.32, and target at 370.04 were all set. Before placing the order, a skill called Time Machine took Tesla's 59 prior trading days of candles for the same time window and replayed the rest of the day 10,000 times: 29.4% hit the target first, 25.5% hit the stop first, and 45.1% had no outcome by the close. The verdict: "close to a coin flip."
The conclusion itself isn't exciting, but the line "replay it ten thousand times before you trade today" is. Nearly every video that shared it used that exact sentence as its title.
What it is
day-trader-skill is 11 intraday trading skills for Claude: chart scanning, news monitoring, entry timing, stop and target setting, Time Machine, position sizing, risk controls, order placement, price alerts, and earnings reading, plus a "Pelosi Tracker" that follows US congressional trade disclosures. It's all Python standard library, reads free public data from Yahoo Finance, Nasdaq and SEC, needs no API key, and is MIT licensed. Paper trading is the default, and every order requires you to explicitly say yes before it goes out.
The install script is a single line: paste the repo link into Claude and say "install skill."
The author, Jake Schincariol (Instagram @opusjake), runs an AI consulting business and a free Skool community called Opus Lab. The repo itself doesn't earn money; it exists to bring people into the community and consulting funnel.
How the stars grew
| Date | star | fork |
|---|---|---|
| 10-22 | 122 | 25 |
| 10-24 | 812 | 172 |
| 10-27 | 1,948 | 407 |
| 10-31 | 3,549 | 816 |
| 11-06 | 5,530 | 1,368 |
Source: GitHub daily numbers. The single-day peak was +528 on 10-27. In the last 7 days it was +2406, down from +3002 over the previous 7 days, so it's cooling, but it still gains about 300 per day.
The turning point came on 10-23: a TikTok video about buildwithai.club drew 323K views, and the next day its stars jumped from 122 to 812. Before that, 10-21 Indian AI creator abhijayarora_ had already posted on Instagram: "Claude just killed day traders!… comment 'X'". Others picked it up after: baroobi (36K views, "Comment 'CLAUDE' and I'll send you the link"), the Thai account heenzaabra (63K), the Indian finance account its.breaking.money with an "honest breakdown" (conclusion: close to a coin flip), which still ended with "Comment TRADE, we'll DM you" to join its own community of 3000 people, and omarontape from 11-05. On YouTube, a tutorial titled "Claude Day Trader Skill: Full Setup" got about 53K views (per search snippet).
Almost nobody in the developer circles on X, Hacker News, or Reddit is discussing it.
Who gave these stars
The fork-to-star ratio is 24.7%, while new projects of similar size typically sit around 11%. The reason shows in the issues: "Nifty", "NSE port", "Adapt the trader skills to the Indian market", "Canada", "Hi", "trade", "test". Many people are new to GitHub, treating issues as a message board and forks as a "save" button. Looking at recent fork accounts, several were registered the same day and have only this one repo. These are retail traders pulled in by short videos who signed up just to get the code, not star-farming bots (bots don't write messy issues).
There are serious contributors too: someone submitted a PR fixing look-ahead bias in Time Machine, and someone fixed a bug where the price had already crossed the stop but risk controls still let the trade through. But none of the 11-06,19 PRs have been merged.
Why this one and not similar projects from a year earlier
Open-source "Claude trading skill" projects existed long before: tradermonty/claude-trading-skills launched in October 2025 year and now has 3663 stars; agiprolabs' version has 68 skills and 513 stars. Feature-wise, they are more complete. What this one added is the following:
- The topic has mass appeal. The author 10-14 first posted a medium-to-long-term version, trader-skill (copying 13F, following Congress trades, rebalancing), which got 1357 stars. A week later they switched to intraday trading plus Time Machine, and the star count was four times higher.
- It has a filmable visual. "Replay it ten thousand times and see your real win rate" can be explained in 30 seconds. Even "the result is a coin flip" can be turned into another viral clip.
- Non-programmers can install it. One-line install, no dependencies, no key. Viewers can follow along with the video, so creators feel safe promoting it.
- It feels safe. Paper trading by default, confirmation required for every order, and each skill ships with tests.
- It has a ready-made distribution network. This is the most important point. The author's earlier replica-skill (7853 stars), instagram-agent-skill (5075), and youtube-agent-skill (4768) already got a group of small AI creators into the habit of using his repos for their own "comment keyword → DM" funnels. One PDF guide made by someone says outright: "I didn't make this, I'm just teaching you how to use it." The day a new pack drops, someone shares it the next day.
He mass-produces like a factory: 19 repos total 32k stars, same "11 skills + the wildest one" template, but closer-skill has 32 stars, seed-skill 41, whale-skill 6. The hit rate isn't high, but one hit means thousands of stars.
If it were you
In one line: The path is worth learning and copying, but you lack his creator network and have to build your own. Worth a try, but don't expect five thousand stars on the first shot.
Who it suits: Developers who already make short videos or know a few AI creators; anyone who wants to collect their first stars and followers in the new "Claude skills / Agent skills" ecosystem.
Who it doesn't suit: People who just want to write good code and wait for the developer community to find it naturally. Those stars come from X and HN, a different game from this one. Also not for anyone building a serious quant tool, since this crowd won't file good issues.
How to do it:
- Pick a mass greed or anxiety theme (stock trading, job hunting, growing followers, saving money, losing weight), not a developer theme. His own data shows the theme matters more than the features.
- Build a "picture" for it that can be explained in 10 seconds, like Time Machine, one that works as a video title.
- Cut installation down to one sentence: paste the link into Claude and say "install skill." No keys, no pip, free data, read-only or paper trading by default.
- Go after a different audience and language. Issues explicitly ask for the Indian NSE market and a Canadian version, while the day-trading version only covers US stocks. Build an "Indian stock market version," "A-share/HK version," or "crypto version," or fix what it does badly (look-ahead bias, risk-control bugs) and make that the selling point.
- Turn it into a series: one or two of the same template a week, cross-linked, accepting that most only get a few dozen stars.
How to get users: Film your own Reel / TikTok demoing that picture, and bundle the repo, a one-page PDF guide, and a screen recording ready to cut. Proactively send it to a dozen small AI creators who run "comment keyword → DM" funnels. They lack material, you lack distribution, and the MIT license is the invitation. Finance and AI creators in India and Southeast Asia are the most receptive.
When to move: "Claude skills + stock trading" is a short-video hot topic right now (a parallel "Claude connects to a trading agent" video has 1.16M views). This wave probably won't last more than a few months, so ideally ship the first one before January 2027.
What would mean you got it wrong: You ship three or four packs, film videos yourself, and each one still stalls at a few hundred stars. Or creators share it and only forks rise, not stars, and it drops to zero within days. That means the distribution network isn't set up, or the novelty of this kind of "skill pack" has already worn off.
Research notes: competitor table, acquisition evidence, full research report
- Competitors
Name Where Launched Scale How it differs from this one tradermonty/claude-trading-skills Open source 2025-10-19 3663 stars / 799 forks A year earlier, for knowledgeable investors, with no viral hook and no short-video distribution agiprolabs/claude-trading-skills Open source 2026-03-11 513 stars 68 skills, broad and multi-platform, and nobody is making Reels for it TauricResearch/TradingAgents Open source 2024-12 121K stars Multi-agent research framework, requires installing dependencies and configuring keys, aimed at developers virattt/ai-hedge-fund Open source 2024-11 65K stars AI hedge fund demo, aimed at developers Jakeschincariol/trader-skill (his own v1) Open source 2026-10-14 1357 stars Medium to long term (13F, congressional trades, rebalancing). The theme isn't as thrilling as day trading and has no picture like Time Machine. Trade Ideas / TradingView screeners Web Years Big Paid, mature commercial tools. It's free and runs in your own Claude, but its data is just the free Yahoo API. - Where users come from
- Verified Other creators' Instagram/TikTok "comment keyword → DM link": TikTok buildwithai.club 10-23 323k views, heenzaabra 63k, baroobi 36k; Instagram abhijayarora_ 10-21, its.breaking.money, omarontape 11-05 all use comment TRADE/CHART/X scripts (TikHub, Jina)
- Verified The author's own Instagram Reel + opusjake.ai landing page/PDF: README says "the trade in the reel"; opusjake.ai/r/day-trader 10-21 published; the about page lists Instagram @opusjake; monetization through consulting and the Skool community Opus Lab
- Verified YouTube/Facebook reposted tutorials: Bhupesh Thukral's "Claude Day Trader Skill: Full Setup" has about 53k views (search snippet); multiple Facebook pages reshared it
- Verified Blog/skill aggregator SEO: enterprisedna AI Pulse 10-22, orcarouter.ai blog post, indexed by skillsllm/qai.io/typingmind
- Verified Series cross-links (v1 trader-skill): cross-links at the top of the README; v1 same period +450/week
- Inferred X / HN / Reddit: 0 entries in the local x-radar corpus; no HN/Reddit discussion found
Jakeschincariol/day-trader-skill: why it worked
Research date 2026-11-06. Data: our github library (repo_profile / repo_day), live GitHub API lookups, Jina web search, TikHub TikTok search. The x-radar local corpus has 0 entries for all related terms (meaning X isn't its main battlefield, or the corpus doesn't cover it).
One line
Jake Schincariol (Instagram @opusjake), a creator who runs AI consulting and the Skool community Opus Lab, mass-produces "Claude skill packs" as short-video topics: each repo gets an eye-catching theme, a fixed pitch of "11 skills + the wildest one," and one Reel. day-trader-skill is the newest hit from this pipeline. It didn't take off through GitHub circles. Dozens of AI creators on Instagram/TikTok running "comment the keyword and I'll DM you the link" used it as a lead magnet, pushing large numbers of non-developer retail traders (India is a clear share) to GitHub.
Data
- 2026-10-20 created the repo, and since then there has been no code push (pushed_at 2026-10-20 14:18, same minute as creation, uploaded all at once). Source: GitHub API.
- Stars: 10-22 122 → 10-24 812 → 10-27 1948 (single-day peak +528) → 11-06 5530 (data card); live GitHub lookup 5701 stars / 1407 forks. Last 7 days +2406, the prior 7 days +3002. Cooling off, but still roughly +300 per day. Source: github library repo_day, data card.
- Forks/stars = 24.7%, far above the library's median of about 11% for new repos with ≥5000 stars (SCHEMA.md). Forks are high because non-developers treat "fork" as "get/save it"; see the issues below.
- The same author's other trading pack trader-skill (created 10-14, v1): 1357 stars, single-day +11-06 120, lifted by the v2 release (github library history).
- The author's 19–25 repos total 32k stars. Hits with the same pattern include replica-skill 7853, instagram-agent-skill 5075, youtube-agent-skill 4768, linkedin-agent-skill 3261, founder-skill 2553, arena-skill 1285. But there are also cold ones on the same template (closer-skill 32, seed-skill 41, hunter-skill 80, whale-skill 6). GitHub API 2026-11-06.
- User profile (from issue/PR titles, GitHub API): "Nifty", "NSE port", "Adapt the trader skills to the Indian market (NSE)", "Canada", "Tred", "Hi", "trade", "test", "Add files via upload". Many are retail users who don't understand GitHub and treat issues as a message board. A few are serious PRs (timemachine removing look-ahead bias, a trader-risk bug fix where trades still went through after price broke the stop-loss).
- Sampling the most recent 25 forkers: 2014–2019 are old accounts, but 5 were registered the same or previous day and have only 1 repos. That fits people redirected from Instagram who signed up for GitHub just to get the code. It doesn't look like star-botting, which usually doesn't come with lots of messy issues and forks. The stargazers endpoint returned 404 on the live check, so I couldn't sample the starring accounts and didn't verify them.
Competitors
| Name | Where | Launched | Scale | How it differs |
|---|---|---|---|---|
| tradermonty/claude-trading-skills | Open source | 2025-10-19 | 3663 stars / 799 forks | A year earlier, aimed at knowledgeable investors (market analysis, economic calendar, screeners), with no "viral" hook and no short-video distribution |
| agiprolabs/claude-trading-skills | Open source | 2026-03-11 | 513 stars | 68 trading/DeFi/quant skills, broad and comprehensive, across Cursor/Codex/Gemini. Nobody is making Reels for it. |
| TauricResearch/TradingAgents | Open source | 2024-12 | 121K stars | Multi-agent trading research framework, requires installing dependencies and configuring API keys, aimed at developers/researchers |
| virattt/ai-hedge-fund | Open source | 2024-11 | 65K stars | An "AI hedge fund" demo, also aimed at developers |
| Jakeschincariol/trader-skill (his own v1) | Open source | 2026-10-14 | 1357 stars | Leans medium to long term (copying 13F, congressional trades, rebalancing). After v2 switched to the more thrilling theme of "day trading + Time Machine," traffic quadrupled. |
| Commercial tools such as Trade Ideas and TradingView screeners | Web/app | Years | Big | Paid and mature. day-trader-skill is free and runs in your own Claude, but its data is just the free Yahoo API. |
| Short videos about "Claude connected to a LiquidTrade/QuiverQuant trading agent" | Other | 2026-10 | One ai.with.andrew video with 1.16M views (TikHub) | Another "Claude stock trading" topic that the same AI creators are pushing shows "Claude + stock trading" itself is a current short-video hot topic. This specific product is unverified. |
Bottom line: open-source projects for "Claude trading skills" existed long before (tradermonty, a year earlier), and there's no technical barrier. What this one has over the others is a picture you can explain in 30 seconds (replaying one trade 10,000 times) + a ready-made short-video distribution network.
Growth tactics (evidence)
- The author's own Instagram/Reel + landing page: the README directly says "the trade in the reel" (the example is the 10-19 TSLA trade, the same one from the Reel). opusjake.ai/r/day-trader (10-21 published) and opusjake.ai/resources/day-trader.pdf act as resource pages; opusjake.ai/about lists Instagram @opusjake. The author's business is consulting plus the free Skool community Opus Lab ("5.0 from 100+ member reviews"). The repo is the lead magnet, and the money comes back from consulting and the community. (Confirmed)
- Other AI creators use it as "comment keyword → DM" lead material (the largest source, confirmed):
- TikTok: buildwithai.club 10-23, one video with 323k views; heenzaabra (Thai) 10-31, one video with 63k; baroobi 10-22, one video with 36k ("Comment \"CLAUDE\" and I'll send you the link"); aiwithabhijay 10-21; joshfpocock, sambit.ai.tech, build.with.alan, girmetheus.ai, agilesingh and others all use "Comment TRADE/TRADER". (TikHub search 2026-11-06)
- Instagram: abhijayarora_ 10-21 "Claude just killed day traders!… comment 'X'"; its.breaking.money (an Indian finance account) ran an "honest breakdown" version, still "Comment TRADE, we'll DM you" and funneling to its own 3000+ member community; omarontape 11-05 "comment CHART". (Jina read the Instagram page)
- Facebook videos/posts, YouTube tutorials (Bhupesh Thukral's "Claude Day Trader Skill: Full Setup" at roughly 53k views, per search snippet; plus re-uploads carrying Amazon affiliate links).
- This is the same path as the author's earlier instagram-agent-skill: someone made a PDF guide saying "built by Jake Schincariol, I did not build it, I am showing you how to use it", and an Hindi-language creator wrote "Guide ke liye comment karo SKILLS". In short: the author released MIT-licensed material for free, and a crowd of small creators took it and built their own DM funnels, growing their followings and distributing it for him.
- Blogs / SEO / aggregator sites: enterprisedna.co covered it in its AI Pulse 10-22; orcarouter.ai wrote "What the 11-Command Pack Actually Does"; skillsllm, qai.io, skillsdirectory and typingmind auto-indexed it in their skill marketplaces. (Confirmed, small scale)
- Series cross-promotion: README at the top links to v1 trader-skill; both repos rose together. (Confirmed)
- X / HN / Reddit: 0 local x-radar corpus posts; Reddit search only turned up a small subreddit reposting the author's other repo; no developer-circle discussion found. (No evidence found)
Why it's working
- Chose the most mass-appeal topic: greed: day trading + "Claude trades stocks for you" reaches a far bigger audience than "managing LinkedIn"; after the same author switched to trading (trader-skill 1357 → day-trader-skill 5700), traffic clearly grew.
- A single shot that works as a short video: "replay today's market 10,000 times before you trade, and it tells you your real win rate"; nearly every creator who reposted used this line as their title (TikHub/Instagram titles). Even the "honest version" holds up: the result is "close to a coin flip", which became a different hook in itself (Breaking Money).
- Zero-friction install pitch: "paste this link into Claude and say install skill"; pure standard-library Python, no API key, free data sources. Non-programmers can follow along from the video, and that is the precondition for pulling in a large number of retail traders.
- Safety framing: paper trading by default, you must say yes to each order, every skill has tests; this lets creators feel safe promoting it and viewers feel safe installing it.
- An existing creator distribution network: the author's earlier hits (instagram/youtube/replica-skill) had already gotten a set of AI creators into the habit of using his repos as lead-gen material; when a new pack came out (10-20), the next day (10-21) someone posted about it, 10-23 a video with 300k views appeared, and 10-24 its stars jumped from 122 to 812.
- Factory-style mass production: the same template, one or two new packs a week; the hit rate is not high (more than half get a few dozen stars), but one hit yields thousands of stars. day-trader-skill is currently the series' second biggest.
Differentiation
Functionally it is not unique (similar open-source projects came earlier and were more complete). The real differentiation is packaging and distribution: a fixed narrative of "11 skills, each does one thing, the craziest one", a visual gimmick for each pack (Time Machine), a polished one-click install prompt, the author's own Reels and landing page, then MIT-licensed remixers promoting it for free. Yes, this is why it took off: not the code, but a design that others can use as a lead-gen hook.
Visibility
Word of mouth in niche circles: it spread widely in AI-creator circles on Instagram/TikTok and among retail traders in India and Southeast Asia (top single video at 320k views), but essentially not on X, HN, or mainstream tech media; no public revenue (it's free; monetization sits in the author's consulting/community). The "copying" pattern: many creators repost the same repo to build their own DM funnels; the author himself also copies the same template for his own work.
What others can borrow
- Designing an open-source repo as short-video material: a visual selling point you can explain in 10 seconds + a line "paste into Claude and it installs";
- MIT + no key + free data, lowering the bar for reposting creators and viewers, letting others' "comment-a-keyword-for-a-DM" funnels distribute it for you;
- serialized, templated mass production, trading hit rate for hits; cross-linking to the previous entry;
- monetization outside the repo (consulting, community), the repo only handles acquisition.
Uncertain
- Star account quality was not spot-checked (stargazers endpoint 404, then GitHub API rate-limited); a sample of fork accounts looked like real newcomers.
- The play count of the author's original Reel was not obtained (Instagram pages don't show the number).
- No evidence of paid promotion to creators (paying creators); judging by each creator steering to their own community, it looks more like organic reposting.
- Whether stars convert to actual usage is unknown: issues show many people can't even use GitHub; the share who actually run it is unknown; code had zero updates for 17 days with 19 PRs backlogged.