I think that a manual is a good idea as long as it is understood as a set of rough guidelines. It could consist of two parts. One on general use of the EA, and another on the meaning of the inputs, perhaps in tabular form, as before. For example, I have no clue what the Grace inputs are about, or even whether to use them. We might also consider a section on optimization technique and backtesting, as I am sure that many may be confused about this (for example, me). E.g., does 99% backtesting on tick data represent a significant advantage over 90% backtesting on M15 timeframe data? While we're at it, how about a preamble about the general philosophy behind the EA, i.e., how its approach differs from others?

It does not have to be a finished product, since the EA isn't finished either. But some working guidelines would be most helpful.