Claude Day Trader Skill: What the 11 Skills Do and Don't Do
By Matt Cullerton, founder of Mavric and Apso
The Claude "trader skill" is a free, MIT-licensed pack of 11 Claude skills for day trading, published on GitHub as day-trader-skill by Jake Schincariol. It scans charts, tracks news and times entries, but it does not make winning trades automatically. Every order needs a risk check and your approval in chat, and paper trading is the default.
A TikTok from @techesha claims the pack lets you "make winning trades automatically" and that its time machine skill "replays every possible outcome of the trade". The repo's README says something narrower on both points.
What is the Claude trader skill?
The repository Jakeschincariol/day-trader-skill contains 11 skills, each invoked as a slash command:
/trader-scanner,/trader-newsand/trader-entryfor scanning charts, tracking breaking news and timing an entry/trader-exit,/trader-riskand/trader-budgetfor exits, risk checks and position sizing/trader-timemachinefor replaying a trade before you take it/trader-brokerfor sending orders through Alpaca/trader-alerts,/trader-analystand/trader-pelosifor alerts, earnings analysis and congressional trade tracking
The video names only the scanner, news, entry and time machine skills. An OrcaRouter blog post dated 2026-10-23 describes the same structure: eleven instruction files, eleven small Python tools and a trader/ folder where each step leaves notes for the next. It says the repository appeared on GitHub on 2026-10-20.
The video never shows a repo name. The match rests on the 11-skill count, the skill jobs and the time machine graphics, which is strong evidence but not proof that this is the exact repo.
Is the Claude trader skill free?
The repository is free. The README lists an MIT license, no signup, and free public data sources: Yahoo Finance chart data, Nasdaq Trader symbol lists, SEC EDGAR and the House Clerk.
That covers the code only. I did not check what a Claude plan costs or whether an Alpaca account carries fees. The README also notes that the Yahoo endpoint is unofficial and can lag or rate-limit, so the scanner's view of the market can be behind.
Does it make trades automatically?
No. The README says nothing in the pack places an order on its own. Orders go only through /trader-broker, after /trader-risk passes the order and the user approves it in chat. The broker skill connects to Alpaca's MCP server with paper trading as the default. The OrcaRouter post quotes the README line "No trade goes through until you say yes."
Nothing in the README or the blog promises winning trades. The README's disclaimers say it is not financial advice and that most day traders lose money. They add that stops are not guaranteed fills and that the tool does not account for taxes, wash sales or pattern-day-trader rules. The OrcaRouter post states that no audited track record exists.
The video's opening line, that Claude "just killed day traders", has no support in either source.
Does the time machine replay every possible outcome?
No. Before a trade, /trader-timemachine simulates the rest of the session 10,000 times. Each simulated 5-minute candle is drawn from the stock's own candles at the same time of day over the previous 59 sessions, using a fixed seed of 19 so the results can be reproduced. It also reruns the same rule on those 59 real sessions.
In the README's TSLA example, about 29% of simulations hit the target first, 26% hit the stop first and 45% were still open at the close. The historical rerun fired 33 times, with 13 targets, 15 stops and 5 neither. The author calls the result close to a coin flip. OrcaRouter gives the same split as 29.4%, 25.5% and 45.1%.
The README warns that the simulation uses only about 60 days of data and cannot anticipate events such as earnings or halts. Resampling 10,000 paths from 59 sessions produces an estimate. It does not enumerate every outcome, and the README does not claim the "most accurate statistics" that the video promises.
The video also shows a chart titled "How the 10,000 replays end" with "Target first 10.9%" and "Stop first 0.0%". Those numbers do not match the README example, and the video does not name the ticker or date, so treat the chart as an illustration of the output format.
Is it comparable to Bloomberg software?
The video says the time machine is "the same type of software Bloomberg charges thousands for". I found no Bloomberg product or price for a trade-replay tool to compare against. The README describes a resampling script that reads free Yahoo data, which is a much smaller tool than a Bloomberg Terminal.
What about the presenter's own results?
The presenter says she trades with a similar setup and shows a "Claude Trading Bot" card with equity growing from $100.00 to $157.33, a gain of 57.33%. The card has no broker name, statement, trade list or link, and it does not say whether the money is real or paper. The video also shows animated explainer graphics rather than the skills running in Claude.
Bottom line
The day-trader-skill repo is real, free and has 11 skills. It requires your approval for every order and defaults to paper trading. Its time machine is a 10,000-run simulation built from 59 recent sessions, and in the author's own TSLA example it came out close to a coin flip. Treat the video's "winning trades" and "every possible outcome" claims as marketing.
Sources
- day-trader-skill on GitHub (Jakeschincariol): skill names, license, data sources, approval flow, replay method, TSLA example and disclaimers.
- Claude day trader skill: what it actually does, OrcaRouter blog, 2026-10-23: independent description of the eleven skills, replay figures and the lack of an audited track record.
- TikTok post by @techesha: the video that made the claims.
Published October 26, 2026. Based on a TikTok by @techesha. Full video notes (markdown)