This is what's missing from fasmg.

A detailed reference manual of its commands with their inputs/outputs.

Not to be mistaken with the current manual or internals. It would show explicitely the difference between the "tokenized" world and "binary" world with their bridges.

If there's anyone willing to design and edit some nice tables, I can assist with information and clarifications. But I'm unlikely to find time for yet another sub-project. Currently in the front of the queue I have some CALM improvements that require a careful design.

That person would need a deep understanding of fasmg internals.

The language of fasmg was designed before a single line of implementation was written, and the implementation followed the intent of the language, not the other way around. I consider it proven by the fact that large portions of fasm and fasmg languages overlap, while they have substantially different implementations (there was nearly no code reuse between fasm and fasmg projects). Some additions to the language were implemented into both, and required very different approach in fasm 1 engine compared to fasm g - yet still they follow the same rules imposed by language design. And even fasmg-specific features like recognition context were designed as language perks before I knew how I'm going to implement them, because I first needed to know how I want them to behave, before I could start working on making an implementation that would provide the intended behavior.

Therefore I believe that what is required is deep understanding of fasmg's language design, not internals.

I agree: strong mastery of the "model"/language design.