The Claude Quant Desk

Install The Claude Quant Desk on my computer. Walk me through it.

Repo: https://github.com/tabani-png/quant-desk
What it does: A real hedge-fund research desk that runs inside your own Claude Code, on the subscription you already pay for — no Anthropic API key, no Bloomberg terminal, no $300K analyst seat. Seven specialist agents study a stock and argue both sides, a trader proposes ONE concrete strategy, then a real, walk-forward, cost-aware backtest refuses to trade unless that strategy actually beat the market on data it never saw. The refusal is the product: most AI trading bots are built to say buy, this one is built to say no, and to tell you in plain English when the honest answer is just buy the index.

I am comfortable copy-pasting and following instructions, but I am not a developer.

Rules:
- Plain English. Define jargon the first time it appears (repo, env var, port, dependency).
- One step at a time. Exact command in a code block. Tell me which app to paste it into (Terminal on Mac, PowerShell on Windows).
- One sentence per command explaining what it does and what success looks like.
- After each command, wait. I will tell you the output before you move on.
- If a tool is missing (git, node, docker, python), give me the one-line install for my OS first.
- If something errors, diagnose before the next step. Do not skip.

First message: ask only "What is your operating system — macOS, Windows, or Linux?" Then start step 1.

Reference steps from the public guide (adapt to my OS, do not just paste them at me):
1. Clone the open-source desk (public, free): git clone https://github.com/tabani-png/quant-desk, then run ./install.sh. It copies the skill and the 7 analyst agents into your Claude Code and sets up a small Python env (yfinance + pandas). There is NO Anthropic API key anywhere — it runs on the Claude Code subscription you already have.
2. In Claude Code, type /quant-desk run AAPL. The desk pulls a year of real prices, runs the market, fundamentals, and news analysts, stages a bull-vs-bear debate, has the trader propose one concrete rule, and the risk manager size it — every step narrated in plain English.
3. The rule then hits the backtest gate: walk-forward (the verdict is decided only on a held-out slice the strategy never saw) and cost-aware. It trades ONLY if it beats holding the S&P 500, has a positive Sharpe, and draws down no worse than holding the stock. Fail any check and the trade is BLOCKED — the AI's opinion does not override it.
4. Optional: place PAPER (pretend-money) trades with free Alpaca paper keys; without them, trades run in DRY RUN and no real money is ever at risk. Prefer no terminal? Copy the 9 prompts below to run the desk by hand in any Claude chat — but only the installed engine runs the REAL backtest on live data.

Stop when the app opens and I confirm it works.