I think that the suggested approach is probably formally correct, but operationally not very impactful
I think that, if we were building an SE-like site from the ground up, this might be a principle that would be pretty sensible to incorporate. For maximum searchability and flexibility of archived content it would be favorable to have tags be as generic as possible. In that sense, it wouldn't matter if D&D were totally unique in the TTRPG space in having a concept called (or described as) spell slots. If someone wants to look for content about spell slots and D&D they can search two tags rather than one very specific tag.
But in practice I don't see much difference arising from properly generic tags outside of changing how people search for questions and answers. In my observation, vanishingly few users come to RPG.SE looking for cross-game information in the first place. My imagination isn't anywhere the final word, but I don't have much idea of what a user looking for content tagged [attack] might ultimately be seeking, or how they would use search results for systemless [attack]-tagged questions and answers.
Even if it is better from an information-theoretical or database-architectural perspective to use more atomized tags, if virtually all searches apply a generic tag with a system tag then we haven't necessarily gained much in any dimension.
Tags are dynamic and (ideally) responsive to observed use-cases
We regularly add and remove tags, generally in response to a perceived need based on some observed event. If we have an [attack-dnd] tag (which I would not support, incidentally), it would be because some collection of users felt that such a tag would make it easier to navigate stack content.
The use-case I envision for more strictly generic tags would be more along the lines of "I want to find information on spell slots, but not for D&D (or at least not only for D&D)". In that situation I believe that a hypothetical compound tag, such as [dnd-spell-slots] would be divided. But it wouldn't matter, as a practical issue, until there was content that required that extra discrimination. I can't call to mind good examples which suggest that there is a practical benefit to be gained from atomizing tags right now (a single question about Legend of the Five Rings spell slots would be enough to shift that, for example, but it only becomes an issue once such a question is posted).
RPG.SE fits with specific questions more so than generic ones
A question about spell slots which is game-system agnostic strikes me as one likely to be pretty nebulous in a way that the SE format doesn't favor (at least on this stack). If you are interested in limiting use of magical skills while designing a game, then sure, a generic tag would be helpful. But would a generic spell slot tag be helpful, or would we need a broader tag like [magic-limitation] (or something that sounds less stupid than that)?
And even then, broad concepts might have system-specific applications that make questions spanning multiple games a bit hard to define. Games with combat tend to have some sort of health system and capacity to resist and survive damage from attacks, which we might call "defense" and tag accordingly. But what question would focus on [defense] in a way that would supersede the radical differences in combat mechanics between, say, D&D and the old World of Darkness games?
This may be a failure of my imagination again, as above, but I'm having trouble thinking of a question that would deal with something like spell slots generically but still be properly answerable (in the SE mode). If practical application of generic tags is that they must be accompanied by specific system tags then the distinction becomes pretty fine. Generic tags might still be better in that it makes tags easier to intuit (you don't need to know the abbreviation l5r to find the l5r-spell-slots tag, perhaps), but that doesn't seem like a huge advantage to me (as system tags appear to usually accompany the dense system-plus-concept tags). That said, if compound tags really are always accompanied by separate system tags then the compound tagging is superfluous anyhow.
Game system terminology seems to me to be pro-jargon, generally
RPG games seem, to me, to be developed and marketed in ways which emphasize the ways they are unique, especially in differentiation from powerhouse franchises like D&D. This tends to lead to diversity in game terms, but also in different mechanics defined by terms that overlap.
I'm not too familiar with Legend of the Five Rings, but a quick search on spell slot mechanics suggests to me that the application of the spell slot idea is really different from D&D. They do have a common function in limiting how much magic can be performed within an arbitrary timeframe, but don't appear to use the phrase spell slot in precisely similar ways. Even if I wanted to examine content about limiting magical abilities in a game, I'm not sure that a generic [spell-slot] tag would be the ideal way to do that; something still more generic might be better (to cover games with a similar idea but that don't use the exact phrase "spell slot").
Terms that are, themselves, more generic and therefore more likely to appear in multiple systems may still have strongly system-specific applications. As an example, D&D 5e's inspiration mechanic allows Advantage (another system-specific use of a very generic word), which is itself a specific mechanic around how D&D 5e uses a d20. If another game comes along, perhaps Bungalows and Bad Guys, which also uses a mechanic called "inspiration", is it reasonable to assume that there are usefully overlapping questions between the two just because they use the same common word? Are those two mechanics more similar than a game which copies the D&D 5e spell slot system exactly but calls them "spell ability tokens"?
Aside from being formally better (which I do believe), I don't see much in the way of practical advantages to enforcing generic tags stack-wide. The argument I really want to see in favor of this idea is a collection of questions (whether they currently exist on the site or not) about cross-system concepts where the use of generic tags would be useful.
