Better AI prompts for customising an MQL5 Expert Advisor

A useful coding prompt describes a job, its boundaries and how you will judge the result. This is especially valuable when an EA already contains behaviour you want to preserve.

Use a five-part brief

1. Context: MQL5, MT5, the relevant function and account mode.
2. Change: one observable new behaviour.
3. Preserve: existing signals, sizing, exits and unrelated inputs.
4. Failure handling: what to do when data or a request is unavailable.
5. Acceptance: the cases that should pass after compilation.

An example brief for an entry filter

Ask the assistant to add an optional spread limit in points to the new-entry path. Require the disabled setting to preserve the original entry behaviour. State that exits and position protection must continue when the filter blocks entries. Request a small diff, a new input description and test cases for below, at and above the limit.

Ask for explanations before expanding scope

Have the assistant identify the intended integration point and explain the units it uses. If it proposes a broad rewrite, bring the request back to the stated change. Check any new API against the official reference. Keep prompt notes with the edited file so a later revision has the same boundaries.

Evaluate the output, not the confidence

Compile the actual file and inspect warnings. Compare the old and new runs under matching conditions. A confident explanation or an attractive code block is not the acceptance test. Save the result only after the requested behaviour appears and unrelated behaviour remains consistent in the checked cases.

Make it yours with MQLume

MQLume gives your customisation brief a practical starting point: editable EA source with product and parameter guidance. Explore Ultimate Breakout System MT5 and turn one well-defined idea into your next development step.

Sources and further reading

• MQL5 function reference
• MQL5 community AI coding introduction

Back to top