Back to blog

Day Trader Skill on GitHub: 11 Claude Skills for Day Trading

Day-trader-skill is an MIT pack of 11 Claude skills for day trading: chart scans, news, entries, exits and a Time Machine backtest. Install, safety and review.

Nov 5, 2026AINovaTools Team
Day Trader Skill on GitHub: 11 Claude Skills for Day Trading

Jakeschincariol/day-trader-skill is an open-source GitHub repo that turns Claude into a disciplined day trader: 11 skills that scan thousands of charts, track breaking news, time every entry and exit, and replay every outcome before you risk a dollar. It is MIT-licensed, free, needs no signup and no API key, and runs on plain Python 3.8+ with nothing to pip install.

The repo went up on October 20, 2026 and collected roughly 5,300 stars in its first two weeks. It is the sequel to the same author's trader-skill — the original "Trader skill" that copies 13F filings, screens stocks, tracks tax lots and rebalances a portfolio. Where that one works at the pace of position investing, this is the intraday version, and it feeds entirely on free data: Yahoo Finance charts, Nasdaq Trader symbol lists, SEC EDGAR and the House Clerk's congressional trading disclosures.

Below: what the 11 skills do, how to install them, a complete trade walkthrough from the README, and the risks stated plainly.

The 11 skills, one table

Every skill is a single slash command. Ten of them are free-standing, but eight chain into a pipeline that ends in at most one trade:

CommandJobWhat it does
/trader-scannerFind setupsScans thousands of charts for opening-range breakouts, bull flags and oversold bounces
/trader-newsWatch the tapeBreaking-news monitor for headlines that move your tickers
/trader-entryTime the entryBar-by-bar timer that fires when the entry actually triggers
/trader-exitPlan the exitSets stop and target before you are in, so the exit is never improvised
/trader-timemachineTest the oddsReplays the session 10,000 times, plus every past day the same setup fired
/trader-budgetSize the tradePosition sizing from your account size and max loss per trade
/trader-riskEnforce your rulesA PASS or BLOCK verdict against your own limits
/trader-brokerPlace the orderOrder executor over the Alpaca MCP server — paper trading by default
/trader-alertsMind the positionPrice, stop and closing-bell alerts while the trade is open
/trader-analystRead the filingsEarnings reader for the numbers behind a move
/trader-pelosiFollow CongressTracks congressional trade disclosures filed with the House Clerk

The trading-day pipeline runs scanner → news → entry → exit → timemachine → budget → risk → broker, producing one candidate trade with the math already done. While the position is open, /trader-alerts watches it; news, analyst and pelosi run on a schedule in the background. The skills coordinate through a trader/ folder in your project.

How to install it

There are four ways in, depending on your setup. If you have never installed a skill before, our guide on how to use Claude skills covers the mechanics.

1. Plain Claude. Paste the repo URL into a Claude chat and ask it to install the skill. Nothing else to configure.

2. Claude Code plugin. Inside Claude Code, register the repo as a marketplace and install it:

/plugin marketplace add Jakeschincariol/day-trader-skill
/plugin install day-trader-skill@day-trader-skill

The commands arrive namespaced, as /day-trader-skill:trader-scanner and so on.

3. Git clone. Copy the skill folders into your user-level skills directory and they appear under their plain names (/trader-scanner):

git clone https://github.com/Jakeschincariol/day-trader-skill
cp -r day-trader-skill/trader-* ~/.claude/skills/

For a single project, copy them into that project's .claude/skills/ instead.

4. No Claude Code at all. Open any SKILL.md in the repo, paste it into a regular Claude chat, and the workflow works — you just lose the Python tools that pull data and run simulations.

Requirements are light: Python 3.8+ on your machine and nothing from pip. One piece of etiquette: the SEC-based tools require you to identify yourself, so set sec_user_agent in trader/config.json or the SEC_USER_AGENT environment variable before using them. New to terminal agents entirely? Start with our Claude Code tutorial for beginners.

One trade, end to end: the README's TSLA morning

The README walks through a real session — Monday, October 19, 2026 — and it is the fastest way to understand the pack.

At 10:15 AM ET, /trader-scanner swept 231 charts and found 42 trading above their 15-minute opening range. TSLA's range was 354.47–362.18, and the 10:10 candle had closed at 364.18, 0.55% over the top — a live breakout. /trader-entry armed at 10:15 and the entry triggered. /trader-exit set the plan before anything else happened: stop at 358.32 (the range midpoint), target at 370.04 (1R), flat by 3:55 PM no matter what.

Then the honest part. /trader-timemachine replayed that session 10,000 times: target hit first in 29.4% of runs, stop hit first in 25.5%, and the trade was simply still open at the close in 45.1%. It also replayed every past day the same rule fired — 33 times across 59 sessions (13 targets, 15 stops, 5 neither) — and delivered its verdict: close to a coin flip.

The rest of the pipeline still ran. /trader-budget, given a $5,000 account and a $50 max loss, sized the trade at 8 shares ($2,913.44 committed, worst case −$46.88). /trader-risk returned PASS. /trader-broker printed the ticket in plain words and waited for a human to type yes.

That last detail is the point: the pack's best skill is not finding trades, it is refusing to romanticize them.

The Time Machine: 10,000 replays before real money

Most trading tools show you a chart and a hope. The Time Machine does two things instead. First, it takes the current session and replays it 10,000 times, sampling forward moves from the stock's own recent bars — a Monte Carlo built from the instrument's actual behavior, not a generic bell curve. Second, it finds every past day where the same setup fired on the same stock and replays those too, so you see how the rule has actually behaved.

The output is a plain distribution: how often the target came first, how often the stop did, how often the trade just drifted. In the TSLA case the answer was unflattering, and the skill said so. That honesty is the feature. A marketing page would have printed a win rate; this prints the odds and lets you decide. Treat it as a discipline engine rather than an edge engine — it cannot make a setup profitable, it can only stop you from sizing up a coin flip.

The guardrails: what it can and cannot do

The README's safety framing is unusually direct: no trade goes through until you say yes. Nothing in this pack can place an order on its own. Paper trading is the default.

Concretely:

  • Paper first. The broker skill connects through the Alpaca MCP server and starts in paper-trading mode. Live orders require explicit configuration on top of that.
  • Double confirmation. Even with everything wired up, the broker prints the ticket in plain words and waits for you to type yes — and Claude's own permission card asks again with Allow once / Deny.
  • One trade per pipeline. The chain produces a single candidate with budget, stop and target already attached; it does not loop out more orders while you sleep.
  • All local, all readable. Plain Python, MIT license, no telemetry surface beyond the public data sources it fetches. If you want to know what a skill does, you can read it.
  • SEC politeness is mandatory. The EDGAR and House Clerk tools refuse to run anonymously — set sec_user_agent in trader/config.json or the SEC_USER_AGENT env var so SEC can identify your traffic.

The risks, honestly

  • This is not financial advice, and the pack is not an edge. Day trading is high-risk and most retail day traders lose money. What this pack automates is discipline — sizing, stops, checklists — not profitability.
  • The Time Machine's own verdict was "close to a coin flip." Past sessions do not predict future ones, and a 33-sample history is a small one. If the demo trade is a coin flip, assume yours will be too.
  • Free data is not a pro feed. Charts come from Yahoo Finance, which can be delayed and revised. A setup that looked clean on a free 15-minute feed can look different on a broker's live tape.
  • It is US-centric. Scanner, news, entry and exit work on US-listed tickers, but SEC EDGAR and House Clerk tooling have no value outside US disclosure regimes, and the broker path assumes a US-accessible brokerage.
  • A narrow setup family. Opening-range breakouts, flags and oversold bounces are specific regimes; when the market's character changes, the playbook changes with it.

Who should try it, and who shouldn't

Good fit: anyone who already watches US markets and wants a paper-trading copilot that forces the same checklist every time; developers curious about multi-skill packs; and traders who know their weakness is sizing and stop discipline rather than ideas.

Skip it if you have no interest in markets — this pack does nothing else, and there are better ways to explore skills. Skip it if you trade outside US hours and instruments, where much of the data plumbing loses value. And skip it if you are looking for a money printer: the pack's own worked example concludes with a shrug.

The bigger takeaway is the same one from Clean Your TV: a well-written skill pack turns an agent into a specialist with rails. The pattern keeps spreading — Ponytail changes how agents write code, Reverse Engineer Anything points them at binaries — and this repo points one at the tape. Picking an agent to run it on? Start with our best AI agent tools roundup.

FAQ

Is day-trader-skill free?

Yes. MIT license, no signup, no API key. All data comes from free sources — Yahoo Finance, Nasdaq Trader, SEC EDGAR and the House Clerk. The only optional cost is a brokerage account if you ever move past paper trading.

Can it place trades on its own?

No. Paper trading is the default, live orders need explicit setup, and even then the broker shows the ticket in plain words and waits for you to type yes — after which Claude's permission card asks again.

Is the market data real-time?

Not guaranteed. The pack runs on free Yahoo Finance data, which is not a professional feed and can be delayed or revised. Expect live behavior to differ from the README's walkthrough.

Is this financial advice?

No — it is open-source software, and this article is a review. Day trading carries a real risk of loss, most retail day traders lose money, and the pack's own demo verdict was "close to a coin flip."

How is it different from trader-skill?

The original trader-skill copies institutional 13F filings, screens stocks and manages tax lots and rebalancing — slower, portfolio-level work. day-trader-skill is its intraday sequel: entries, exits and session replays within a single trading day.

Tools mentioned in this article