MQL4/.github/instructions/mql5-experts.instructions.md

applyTo
**/MQL5/Experts/_Thivyam/**/*.mq5

Instructions for _Thivyam MQL5 Expert Advisors

CRITICAL: MQL5 Syntax Compliance

Before writing any MQL5 code, read and follow: mql5-syntax-critical.instructions.md

Key mandatory rules:

• Use dot operator (.) for pointer member access, NOT arrow (->)
• Declare loop variables BEFORE for-loop: int i; for(i=0; i<n; i++)
• Use dynamic arrays (Type[]) for ArraySetAsSeries() and CopyRates()
• Verify property constant names (e.g., OBJPROP_FONTSIZE not OBJPROP_FONT_SIZE)

EA-Specific Instructions

• Structure: All EAs must have a clear structure using the standard event handlers: OnInit(), OnDeinit(), and OnTick().
• Initialization (OnInit()):
  • All indicator handles (e.g., from iMA, iRSI) MUST be created only once in OnInit() and stored in global variables.
  • Set the EA's magic number using CTrade::SetExpertMagicNumber(). The magic number must be a unique input parameter.
  • Validate user inputs (risk %, symbol, timeframe) and return INIT_PARAMETERS_INCORRECT when invalid values are detected.
• Trading Logic (OnTick()):
  • Use the object-oriented CTrade class from the Standard Library for all trade operations. Avoid legacy functions like OrderSend().
  • All trading conditions must be encapsulated in clear, well-named boolean functions (e.g., bool CheckBuySignal()).
  • Enforce once-per-bar execution using a dedicated detector so automated agents do not accidentally trigger multiple trades on the same candle.
  • Before any trade, compute position size from equity and stop distance; never hardcode lot sizes.
  • Check margin requirements via AccountInfoDouble(ACCOUNT_MARGIN_FREE) and log informative warnings before aborting a trade.
• Deinitialization (OnDeinit()):
  • Properly release all resources, such as indicator handles and graphical objects, to prevent memory leaks.
• Error Handling: Wrap every trade execution call with success/failure checks, log GetLastError() details, and surface actionable messages for both human and AI reviewers.
• Telemetry: Emit concise status and risk metrics to the Experts log so downstream tools can parse behaviour during backtests.
• Testing: Provide default input parameter sets suitable for Strategy Tester, and document any required custom symbols or data in the project README.
• Agent Workflow: Leave breadcrumbs explaining complex state transitions (e.g., when escalating from H4 to M15 logic) so future automated refactors remain safe.