Describe the strategy the way you would explain it to another trader. You do not need to name indicators precisely, and you do not need to write anything that looks like code. "Buy gold when it drops below its 50 day average" is a complete, buildable instruction.
Say four things
An instrument, a reason to enter, a reason to leave, and how much to risk. Those four are what a strategy needs to be testable at all.
If you leave one out, we fill it with a stated default rather than guessing silently: no stop described means the position needs some other defined end, and no risk described means 1% of the balance per trade. The review card names every default it applied, so you can see what was assumed before anything runs.
What to leave out
Do not tune numbers on the first pass. "A 14 period RSI below 32.5" is not more testable than "when RSI is low" — it is just a guess with more decimal places, and you can change it in one sentence after the first run.
Do not describe how you feel about a trade, and do not describe what you would do differently in different conditions unless those conditions are themselves rules. "I would probably skip it if the market felt heavy" cannot be tested, because nothing measures heavy.
If we cannot build part of it
We say so on the review card, name the part, and build the rest. A rule is never quietly dropped.
The homepage lists what the engine can work with. Anything outside it — equities, indices, options, order book depth, trailing stops, news blackouts — is named there as absent rather than discovered when your strategy comes back missing something.