Trade-quant
Paper-trading day-trade scout: Claude Code skills that scan, filter, simulate, size and risk-check opening-range breakouts, with a KPI journal
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Check it deeper on your machine
It has changed since its scan on 21 Nov 2026.
As a Claude Code plugin, install it and then ask Claude Code:
/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/Soumava1/Trade-quant with scanskillsafety
Or from a terminal, clone it and run it:
git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/Soumava1/Trade-quant'
It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.
Badge for your README
Markdown
[](/listings/soumava1--trade-quant)HTML
<a href="/listings/soumava1--trade-quant"><img src="/api/listings/soumava1--trade-quant/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
mkdir -p trader cp config/config.example.json trader/config.json # then edit your limits python3 .claude/skills/trader-scanner/scan.py universe # stock list, refresh weekly
`trader/` holds your config and scan runs. It is gitignored and never committed. The paper journal lives in `journal/trades.jsonl` (set by `journal_dir` in the config) so its history is kept in git. Put your name and email in `sec_user_agent`, or in the `SEC_USER_AGENT` environment variable, to include SEC filings: the SEC asks every tool that reads EDGAR to identify itself.
The example config is set up for a $1,000 paper account:
| Setting | Value | | --- | --- | | Budget | $1,000, with 10% kept in cash | | Max loss | $15 per trade, $30 per day, $80 per month (then the scout stops until next month) | | Trades | 2 a day | | Reward:risk | at least 2:1 | | Prices | $5 to $150 | | Position cap by score | 85+ → 75%, 70+ → 60%, 55+ → 45%, otherwise (paper picks) 25% |
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 193 lines
- # Trade-quant
- A paper-trading day-trade scout built as Claude Code skills. Each run finds
- opening-range breakouts in US stocks that are "in play", checks the news,
- plans the stop and target, replays the trade 10,000 times on the stock's own
- recent bars, sizes it to your limits, and runs a risk check. You get a short
- ranked list of trades, or an honest "nothing today". A paper journal then
- measures how those picks actually did.
- **Paper trading only. Nothing here places an order.** The scout refuses to
- run unless `"paper": true` is in your config. Not financial advice: most
- people who day trade lose money, and the odds shown are a reality check from
- the last 60 days, not a forecast.
- ## How it works
- ```
- scanner -> news -> entry -> stocks-in-play filter -> exit plan -> Time Machine -> budget -> risk
- \-> paper journal -> KPIs
- ```
- | Step | Skill | What it does |
- | --- | --- | --- |
- | Scan | `trader-scanner` | Finds stocks whose 5- or 15-minute opening range has broken to the upside, across ~5,900 US stocks |
- | News | `trader-news` | Yahoo headlines and SEC 8-Ks. Halts, risk headlines and serious 8-Ks veto a trade; legal, regulatory, earnings and deal news are cautions |
- | Entry | `trader-entry` | A fresh breakout, or a late entry if the price is still holding within 1% above the range high |
- | In play | `trade-scout` | Opening-range relative volume at least normal, daily ATR at least $0.50, at least 1M shares a day, and above VWAP; ranked by relative volume |
- | Exit | `trader-exit` | Stop at the range low or 10% of the daily ATR; targets of 2R, 3R and 4R; always flat by 3:55 PM |
- | Time Machine | `trader-timemachine` | 10,000 replays of the rest of the day for each stop/target pair, plus every past session the rule fired; keeps the best average R |
- | Budget | `trader-budget` | Whole shares inside your max loss, position cap, cash buffer and order limits |
- | Risk | `trader-risk` | One batch check: daily loss, trades per day, reward:risk, entry window, duplicates |
- | Journal | `trade-scout/journal.py` | Records each pick, settles it on that day's real bars after the close, and reports the KPIs |
- Each trade gets a score out of 100, built only from the tools' numbers: Time
- Machine edge and average R, the rule's history, relative volume, VWAP and
- the market's direction, and freshness. Position size follows the score.
- - **HIGH:** passed every gate.
- - **PICK:** when nothing is HIGH, the best trade with a positive expected
- outcome, lower conviction, so the journal has something to measure most days.
- - **WATCH / REJECT:** each comes with the reason.
- The stocks-in-play rules follow Zarattini, Barbon & Aziz, ["A Profitable Day
- Trading Strategy For The U.S. Equity
- Market"](https://ideas.repec.org/p/chf/rpseri/rp2498.html) (Swiss Finance
- Institute RP 24-98).
- ## Requirements
- - Python 3.11+ (the tools use only the standard library)
- - Network access to `query1/query2.finance.yahoo.com`, `feeds.finance.yahoo.com`,
- `www.nasdaqtrader.com` and `www.sec.gov`
- - [Claude Code](https://code.claude.com) to use the skills and the agent (the
- scripts also run on their own)
- ## Setup
- ```bash
- mkdir -p trader
- cp config/config.example.json trader/config.json # then edit your limits
- python3 .claude/skills/trader-scanner/scan.py universe # stock list, refresh weekly
- ```
- `trader/` holds your config and scan runs. It is gitignored and never
- committed. The paper journal lives in `journal/trades.jsonl` (set by
- `journal_dir` in the config) so its history is kept in git. Put your name and email in `sec_user_agent`, or in the
- `SEC_USER_AGENT` environment variable, to include SEC filings: the SEC asks
- every tool that reads EDGAR to identify itself.
- The example config is set up for a $1,000 paper account:
- | Setting | Value |
- | --- | --- |
- | Budget | $1,000, with 10% kept in cash |
- | Max loss | $15 per trade, $30 per day, $80 per month (then the scout stops until next month) |
- | Trades | 2 a day |
- | Reward:risk | at least 2:1 |
- | Prices | $5 to $150 |
- | Position cap by score | 85+ → 75%, 70+ → 60%, 55+ → 45%, otherwise (paper picks) 25% |
- ## Use
- In Claude Code:
- - `/trade-scout`, or ask "find me high conviction trades", runs the full
- pipeline and summarises it.
- - The `trade-scout` agent does the same and reads the news behind each HIGH
- trade before reporting.
- From the shell:
- ```bash
- python3 .claude/skills/trade-scout/scout.py # full scan
- python3 .claude/skills/trade-scout/scout.py --symbols NVDA,AMD # your list only
- python3 .claude/skills/trade-scout/journal.py settle # after the close
- python3 .claude/skills/trade-scout/journal.py report # the KPIs
- ```
- **When to run:** after the opening range forms, 9:40 AM New York time for the
- 5-minute range and 9:50 AM for the 15-minute one. Most breakouts happen before
- about 10:45, and late entries can still be found until 3:30 PM. A full scan
- takes 5 to 10 minutes.
- **Scheduled runs:** Claude runs the scout on trading days at 9:47, 9:58,
- 10:12, 10:27 and 10:44 AM and 12:33 PM New York time, and settles the journal
- at 4:07 PM. It posts each summary and commits the journal to `main`. Entries
- before 9:45 are blocked by the config, so 9:47 is the first useful scan.
- **To take a trade on paper:** run `/trader-risk` on it (which records the
- PASS), then `/trader-broker`. It needs your explicit yes and a connected
- Alpaca paper account (below).
- ## Alpaca paper account (optional)
- The repo's `.mcp.json` adds Alpaca's official MCP server (`alpaca-mcp-server`,
- pinned to 2.4.0), locked to paper trading. It holds no keys: it reads them
- from your environment. On your own computer:
- 1. In Alpaca's dashboard, create **paper** trading API keys.
- 2. Install [`uv`](https://docs.astral.sh/uv/), which provides `uvx`.
- 3. Set the keys as environment variables in your shell profile, never in the
- repo or a chat:
- ```bash
- export ALPACA_API_KEY="your paper key" # macOS/Linux: ~/.zshrc or ~/.bashrc
- export ALPACA_SECRET_KEY="your paper secret"
- ```
- On Windows: `setx ALPACA_API_KEY "..."` and `setx ALPACA_SECRET_KEY "..."`,
- then open a new terminal.
- 4. Open the repo in Claude Code, approve the `alpaca` server when asked, and
- check `/mcp` shows it connected.
- 5. Ask "show my Alpaca account info" and confirm in the dashboard that it is
- the paper account.
- `.claude/settings.json` sets the permissions for this repo:
- - **Ask every time:** placing, changing, cancelling or closing an order.
- - **Blocked:** options, crypto, and "close or cancel everything".
- - **Allowed without asking:** read-only lookups (account, positions, orders,
- quotes).
- Never move an order tool to `allow`. Cloud sessions cannot reach Alpaca, so
- the scheduled scans never place orders. Place them from your own computer
- after your yes.
- ## KPIs
- `journal.py report` shows HIGH and PICK separately:
- - win rate
- - average R (the expected gain per trade, in multiples of the risk)
- - average win and average loss
- - profit factor (gross wins / gross losses)
- - P&L and return on the budget
- - max drawdown, in dollars and in R
- - longest losing streak
- - trades per day
- - how trades ended: target, stop or at the close
- Judge the results over 30 to 50 or more trades. An average R above +0.2 with a
- drawdown you can live with is the bar before risking real money, and real
- fills (slippage at the open) will be worse than the replays.
- ## Repository layout
- ```
- .claude/skills/trader-* the 11 Day Trader skills (vendored, MIT, see LICENSE-day-trader-skill)
- .claude/skills/trade-scout/ scout.py (pipeline), features.py (stocks in play), journal.py (KPIs), SKILL.md
- .claude/agents/trade-scout.md
- .claude/settings.json permissions for the Alpaca order tools (ask / blocked / allowed)
- .mcp.json Alpaca's MCP server, paper only, keys from your environment
- config/config.example.json starting limits for trader/config.json
- journal/trades.jsonl the paper journal (picks and their settled outcomes)
- tests/ offline tests on saved market data (TSLA, 2026-10-19)
- ```
- ## Development
- ```bash
- uv sync
- uv run pytest
- uv run ruff check .
- ```
- The tests run offline on fixtures from the upstream pack. Live runs need the
- network hosts above.
- ## Credits
- The `trader-*` skills come from
- [Jakeschincariol/day-trader-skill](https://github.com/Jakeschincariol/day-trader-skill)
- (MIT, see `.claude/skills/LICENSE-day-trader-skill`).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.