i would be a bit leery of the Rank 3 you listed as it seems better than the existing masteries for other skills. All of the numeric bonuses to skill checks are either +5 or +1k0 (~ +3), with the +5 going to narrower or less useful applications. And most of those are at Rank 5. What you propose seems significantly better. In fact, in an opposed Courtier check with a member of a Clan, you would get a bigger bonus from your Lore: Clan at Rank 5 than from Courtier Rank 5. That doesn't seem quite right.

Really? It seems fine to me: Courtier 5 gives you a +2-3 bonus with *all* your Courtier checks, whereas Lore: Clan 5 only gives you a bonus with that single clan. I think any player who invests 15 XP in knowing lots of stuff about one particular faction in Rokugan has earned a leg up in that specific context, while the player who invests the same amount of XP in being socially effective in general will still be a better manipulator everywhere else.

Regarding the "hidden message" thing, I forgot to say that the issue with encoding things like that in all kinds of artwork is that it's hugely vulnerable to misinterpretation. That same Crab might also think "the chrysanthem is an imperial symbol. Something about the Emperor? But purple is a Unicorn color . . . the Emperor isn't married yet. Are we supposed to be supporting a Unicorn candidate for Empress?" I like the idea of hiding messages in art, but to my eye, it still isn't anywhere near on par with ciphered text. Which is mostly my argument for why I'd keep Calligraphy separate; it doesn't mean a Cadence-type thing can't work as a special quality for artwork.

I'm still chewing on the topic of crafting, so I'll skip that part for now. You've definitely prodded me to go look at cooperative/cumulative mechanics, though -- I'd actually kind of forgotten about those.

I agree in many ways about Lore and Raises, but I feel we need to work with the system we have currently, unless we want to completely re-write it.

I don't think it's a complete re-write to say it's just scalar success. Some mechanics in the system already operate that way, and to treat knowing more as a matter of Raises really runs counter to the IC sense of what a Raise represents. Personally, I prefer to keep the application of the mechanics internally coherent.