Replay a Day Trade 10,000 Times Before Risking a Dollar with the Day Trader Skill

Most day-trading tooling asks for an account, an API key, and trust. The Day Trader skill — 11 Claude skills by Jake Schincariol, 2k stars, MIT — asks for none: the tools are plain Python that read free public data, paper trading is the default, and nothing in the pack can place an order on its own.

Why This Skill Matters

The README positions this repo as the day-trading sequel to the author's original Trader skill (13F copying, screener, tax lots, rebalancing live in Jakeschincariol/trader-skill). The data comes from Yahoo Finance charts, Nasdaq Trader's symbol lists, SEC EDGAR, and the House Clerk — free, with no signup. One skill scans thousands of charts for setups, another tracks breaking news on every stock you watch, one watches a setup bar by bar, another sets your stop and target before you enter. The README calls the Time Machine the craziest one: before you risk a dollar, it replays the rest of the session 10,000 times from the stock's own recent bars, plus every past day the same setup fired, and reports the odds.

Installation

Paste this into Claude:

https://github.com/Jakeschincariol/day-trader-skill

install skill

Or as a plugin in Claude Code, where the skills are namespaced as /day-trader-skill:trader-scanner and so on:

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

For plain /trader-scanner names, copy the folders instead — globally into ~/.claude/skills/ or project-locally into your repo's .claude/skills/:

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

The tools need Python 3.8 or newer and nothing to pip install. The SEC tools (news filings, the analyst) ask you to identify yourself: put your name and email in trader/config.json as sec_user_agent, or in the SEC_USER_AGENT environment variable. Without Claude Code at all, the README says you can paste any single SKILL.md at the top of a chat and it runs as a mode — you lose the Python tools, but the method works.

The Eleven

Everything lands in a trader/ folder in your project, and each skill reads what the last one wrote. The README's pipeline for one trade is scanner, news, entry, exit, timemachine, budget, risk, broker; alerts run while the trade is open, and news, analyst, and pelosi run on a schedule.

Real Workflow: The README's TSLA Trade

The README walks one trade on real data — Tesla, Monday October 19, 2026 — and its offline test fixtures reproduce every number:

1. /trader-scanner at 10:15 AM ET: of the 231 charts scanned, 42 had closed above their 15-minute opening range. TSLA's range was 354.47 to 362.18, and the 10:10 candle closed at 364.18, 0.55 % over the top.
2. /trader-news pulled the morning's TSLA headlines. Tesla was due to report earnings that week, so a day trade is out before the close and not exposed to the report.
3. /trader-entry watched bar by bar: waiting at 10:12, triggered at 10:15 on the first 5-minute close above the range.
4. /trader-exit set the plan before the entry: stop at the range's midpoint, 358.32 (5.86 a share at risk), target one times that risk at 370.04, flat by 3:55 PM either way.
5. /trader-timemachine replayed the rest of the day 10,000 times, each candle drawn from TSLA's own candles at that time of day over the previous 59 sessions (seed 19): target first 29.4 %, stop first 25.5 %, still open at the close 45.1 %. On those 59 real sessions the rule fired 33 times — 13 hit the target, 15 the stop, 5 neither. Its verdict: close to a coin flip.
6. /trader-budget with a $5,000 account and a $50 max loss: 8 shares ($2,913.44), a $46.88 loss if the stop fills, written as one bracket order.
7. /trader-risk checked it against the max loss, daily loss limit, trades per day, reward to risk, entry window, position size, and cash — PASS or BLOCK, with the reason.
8. /trader-broker showed the ticket in plain words. You type yes for that exact order; Claude calls your brokerage connector (Alpaca's MCP server, paper account by default), and Claude's own "Allow once / Deny" card asks once more.
9. /trader-alerts rings on the stop, the target, and at 3:55 PM.

Tips

• Every tool runs from the shell too — the README lists each one, for example python3 trader-scanner/scan.py orb --at 10:15 for opening-range breakouts so far, or python3 trader-timemachine/timemachine.py run --write for the replay.
• Set your brokerage tool to "Needs approval" in Claude, never "Always allow", so Claude's own card stays a second check; no options, margin, short selling or leverage unless you turn them on.
• The README notes FINRA removed the $25,000 pattern-day-trader minimum on June 4, 2026, with brokers allowed to phase the change in until October 20, 2027 — if yours still applies the old rule, set "pdt_guard": true and /trader-risk counts your day trades.
• The tests run offline on fixtures cut from real data: python3 -m unittest discover -s tests -v.

When Not to Use This

The README's fine print is the boundary, verbatim in spirit: this is not financial advice, most people who day trade lose money, and you can lose money. The Time Machine is a simulation built from the last 60 days of 5-minute bars, not a prediction; a stop is a plan, not a guarantee, and can fill well past its price in a fast market; and the data is free and unofficial — Yahoo's chart endpoint can lag, change, or rate-limit you, and the scanner and the entry timer read 1-minute or 5-minute bars, not every tick. Check anything that matters against your broker before you act on it.

See the leaderboard for more skills.