AI column

Data@donghyeok99Staff

Day Trader skill: 11 Claude skills for day trading

Before a Tesla trade, it replayed the rest of the day 10,000 times and called it close to a coin flip

From the TSLA opening-range breakout, the rest of the session is replayed 10,000 times until the odds settle and the verdict reads close to a coin flip.
From the TSLA opening-range breakout, the rest of the session is replayed 10,000 times until the odds settle and the verdict reads close to a coin flip.
Who
Made by Jake Schincariol, as the day-trading sequel (v2) to the same author's Trader skill.
When
Published on GitHub on October 20, 2026, with no tagged release yet.
Where
Installs as a Claude Code plugin or into ~/.claude/skills. The tools are standard-library Python 3.8+ running on your own machine.

What was hard

A breakout shows up mid-session and you have minutes to decide. Where the stop goes, how many shares to buy, and how this setup ended the last few times all need working out, so the click usually comes from a feeling rather than a number.

What it does about it

Eleven skills pass one trade along in order. The scanner finds opening-range breakouts, the entry timer picks the moment from a 5-minute close, and the exit planner writes the stop and target first. The centrepiece is the Time Machine: it rebuilds the rest of the session 10,000 times from the stock's own past 5-minute bars at the same time of day and counts which side gets hit first. It also shows what the same rule really did on each past session.

How it runs inside

  1. 1

    trader-scanner → trader-entry · trader-exit

    orb --at 10:15: stocks closed above the rangeKeeps only stocks whose 5-minute close cleared the first 15-minute opening range.trader-scanner/scan.py
  2. 2

    trader-entry · trader-exit → trader/plan.json

    entry 364.18 · stop 358.32 · target 370.04The first 5-minute close above the range is the entry; the stop is the range midpoint, the target 1R.trader-exit/exit.py
  3. 3

    trader/plan.json → trader-timemachine

    plan.json + one Yahoo request for 60 days of 5-min barsKeeps only full 78-bar sessions and turns each bar into ratios to the previous close, the pool it draws from.trader-timemachine/timemachine.py
  4. 4

    trader-timemachine

    10,000 replays: random bar from ±2 slots, seed 19Checks the stop before the target in every bar, so a bar touching both counts as the stop. Same seed, same numbers.trader-timemachine/timemachine.py
  5. 5

    trader-timemachine → trader-budget · trader-risk

    target 29.4% · stop 25.5% · open 45.1%, one verdictAdds the rule's real record on past sessions, repeats the verdict as printed, then hands off to /trader-budget.trader-timemachine/SKILL.md
  6. 6

    trader/plan.json → trader-budget · trader-risk

    $50 max loss → 8-share bracket order in orders.jsonShares are the floor of the smallest of max loss, position cap, order cap and cash.trader-budget/size.py
  7. 7

    trader-budget · trader-risk → trader-broker order ticket

    PASS or BLOCK per order, with reasonsChecks limits, orders per day, the entry window and the cash buffer, and records verdicts in trader/state/risk.json.trader-risk/check.py
  8. 8

    trader-broker order ticket

    READY ticket; records a yes for that one orderOnly the exact order that passed today is READY, and replies like 'yes to all' or 'ok' are refused.trader-broker/ticket.py

What goes in, what comes out

TSLA 10:10 bar closes at 364.18 (range top 362.18)Triggered; stop 358.32, target 370.04
Rest of the session, 10,000 replays (seed 19)Target first 29.4%, stop first 25.5%, still open at the close 45.1%
The same rule on the previous 59 sessionsFired 33 times: 13 target, 15 stop, 5 neither
$5,000 account, $50 max loss per trade8 shares ($2,913.44); a fill at the stop loses $46.88

Is it for you

Good for

  • Practising rule-based trades such as opening-range breakouts
  • Writing down the stop, target and share count as numbers before you order
  • Walking a whole trade through with Claude on a paper account
  • Keeping up with earnings reports and Congress members' stock-trade filings on a schedule

Not for

  • Not financial advice. The README itself says most people who day trade lose money.
  • The Time Machine is built from the last 60 days of 5-minute bars, so earnings, halts and headlines are outside it.
  • Yahoo Finance's chart endpoint is unofficial and can lag or rate-limit, and the tools read 1- or 5-minute bars, not ticks.
  • It plans long trades only. Real orders need a broker connector such as Alpaca's MCP server, which defaults to paper.

Reviewed by Devalley’s editor before publishing. How the column is made

Sign in to post, comment and like.Sign in

Report

Reason

Read the community rules