Hi all, I'm building my own MT5 Expert Advisor (MQL5) with Claude as part of Zero One. The infrastructure runs on a demo account: risk module, trade log, news filter, daily and weekly reports. What I don't have yet is a strategy that survives my own tests. I've tested 9 ideas and all of them failed. Before I start round 10, I'd like your input. My fixed rules (not negotiable): - Account €250, risk per trade 0.5% (≈ €1.25 to the stop), minimum lot 0.01 - One market, one open position at a time - Hard stop-out (1/3 of peak equity is never touched), daily stop 2% - Everything closed on Friday 20:00 CET, nothing held over the weekend - No new trades around high-impact news (±30 min) - No martingale, grid or averaging down; every trade has a stop-loss - Broker: Fusion Markets (MT5): forex, index CFDs, metals My bar for a strategy: - Profit factor ≥ 1.2 in the test period and in a separate out-of-sample period - Still profitable with 2× costs (spread, commission, swap) - At least 50 trades - Edge per trade at least 5× the spread - Public rules and some independent evidence (paper or replication), so no black-box EAs or signal services What I tested (all rejected): - Donchian breakout, EURUSD M15 - Asian range breakout at the London open, EURUSD - Time-of-day effect, EURUSD (Breedon & Ranaldo) - On the AEX (NETH25): turn of the month, RSI(2) pullback, intraday momentum, long above the 200-day average - Opening range breakout on index CFDs The pattern: with €1.25 risk the stop has to be small, so trades are short, and the spread eats the edge. The strategies with the strongest evidence (trend, momentum, carry) hold for weeks, which clashes with my Friday close. My questions: 1. Do you know a strategy with public rules and evidence that fits these rules? 2. Or is this combination simply not realistic? If so, which constraint would you look at first? (Risk per trade and the loss limit stay as they are.) 3. Which MT5 markets have the best daily range compared to the spread for a small account?

The Day 43 correction mentioned that explicit prompt-file loading was only tested on macOS, and that Windows was still untested. I can confirm: it works on Windows 11. Claude summarised my agent's role, its three roles and all nine non-negotiable rules from the file. It read nothing and executed nothing. Tested with: Windows 11 Pro (Hyper-V VM), PowerShell 7.6.9, Claude Code 2.1.311. How I did it, step by step (Windows): 1. Check whether the claude command exists. I had Claude working in VS Code, but claude in PowerShell gave "not recognized". The VS Code extension doesn't install the terminal command, so you need the CLI. 2. Use PowerShell 7, not as administrator. Windows PowerShell 5.1 can drop the empty "" argument in --tools "", which breaks the test. pwsh -v # if missing: winget install --id Microsoft.PowerShell -e 3. Install the Claude Code CLI (official installer, in PowerShell 7): irm https://claude.ai/install.ps1 | iex It installs to %USERPROFILE%\.local\bin\claude.exe. 4. Add that folder to your user PATH. The installer warns that it isn't in PATH. This changes your own user PATH only, no admin needed: [Environment]::SetEnvironmentVariable("Path", [Environment]::GetEnvironmentVariable("Path","User") + ";$env:USERPROFILE\.local\bin", "User") Close the terminal completely, open a new PowerShell 7 window, and check: claude --version 5. Sign in once. Type claude, log in with your subscription in the browser, then /exit. 6. Run the loading check with the absolute Windows path to your prompt file: claude -p --strict-mcp-config --tools "" --append-system-prompt-file "D:\path\to\zero-one\system-prompt.md" "Loading check only. Summarise the agent role and rules from the supplied instructions. Do not perform the agent task." ✅ Pass: Claude summarises your role and rules from the file. With --tools "" it has no tools at all, so it can't read or run anything. Tips: Run real sessions from your project folder (cd to it first), not from C:\Windows\System32. Otherwise Claude doesn't pick up your CLAUDE.md.

Before writing a single line of EA code, I spent a long session making sure the foundation is solid. Here's what got done: 1. Risk rules first, in hard numbers I wrote a dedicated risicoregels.md (risk rules) file. Every number in it becomes a fixed constant in the code, not a setting anyone can tweak: - Demo account only, €250 start, EUR/USD only, max 1 open position - Risk per trade: 0.5% of balance below €500, 1% from €500 - If even the minimum lot (0.01) risks too much → no trade, no "just this once" - Daily stop: −2% in a day → no new trades until tomorrow - Max spread 1.5 pips, no new trades 30 min before/after major news, everything closed Friday 16:00 New York time - Loss limit with two safety nets, whichever is stricter: one third of the highest equity is never touched, and half of the highest profit is always kept. At the start that means the account can never drop below €83.33. 2. Thinking in pips, not euros My first instinct was "keep the profit target tiny so I don't give it back." The maths showed the opposite: a €0.50 target against a €1.25 risk needs a 76% win rate just to break even. So instead: - Take profit = 1.5× the stop distance (in pips) - Once a trade is 1× the stop distance in profit, the stop moves to entry +1 pip → that trade can no longer lose - With that setup, a ~44% win rate is enough to break even Small fish swimming with the whales, but letting the current carry you a bit. 🐟🐋 3. Clear roles between me and AI I briefly wanted to give Claude "carte blanche" to build and run everything. We talked it through and settled on this split: Claude researches, proposes (with reasoning) and builds; I decide and approve every step. Hard limits can only be changed by me, in the code. No AI, and no person, gets uncontrolled access to my money. 4. Python ↔ MT5 bridge working - Took a Hyper-V checkpoint of my trading VM first, so there's always a way back - Found out the MetaTrader5 Python package doesn't support Python 3.15 yet (max 3.14) → installed 3.14 next to it and created a dedicated virtual environment just for Kairo - Wrote a read-only connection test script: it connects to MT5, checks the account, balance and symbol data, and never places an order

What it does: every morning it checks the connection, the account and the economic calendar → from 08:00 to 21:00 (Belgian time) it looks for a Donchian breakout on EUR/USD at every 15-minute candle → every signal goes through a hard risk module (loss limit, 0.5% per trade, news filter, spread check) → only on a green light does it open a trade with a stop-loss → everything gets logged, including rejected trades → all positions closed on Friday evening → an evening report lands in my inbox. It proposes improvements; I decide, and every change goes through a backtest and demo first. My soul.md in one line: "Small steps, hard limits, think it through, and keep going." #Architecture-Complete

To keep my first version focused on one strategy and an MT5 demo account, I’ve identified three repeatable tasks: /demo-check — Check the MT5 connection, confirm the selected account is a demo account, and report whether price data and configured risk limits are available before a demo session. /review-trade-log — Read the trading log and produce a consistent summary of executed trades, rejected trades, rejection reasons, and technical errors. /review-test-results — Compare available backtest and demo results with the previous version and suggest improvements for my review without applying changes. These skills would support my workflow while order execution and risk enforcement remain in the trading code, with changes subject to my approval.