Yes, I think we should have these. (Naturally, since I added a couple of them.)
Having categorisation tags for MTG would be helpful, whether of the format of [protection] et al, or [mtg-protection] et al (subtagging, as suggested and adopted on Arqade). People here seem concerned about picking up the former format, so we can go with the latter: Jeff Atwood suggests the latter would be legitimate under circumstances that are the case for MTG questions here (loads of questions, loads of potential tags).
I added those tags because we have virtually no categorisation among our MTG questions, and that makes them very burdensome to sort or search through, and that isn't a good situation. A similar tag system has worked exceptionally well on a site I'm a long-term user of. I would consider it a major loss in potential gain for the site if we weren't able to apply a similar system here, and it was enforced that MTG should have no mechanical tags at all, considering it also does no damage to the site for being present.
It adds value to the site
Like Rainbolt described. I can use tags like this to search for particular questions I'm after that are about Protection or other mechanics.
murgatroid99 suggests searching for MTG questions that mention protection is just as good, but it isn't: a lot of false positives turn up which are questions that are not about protection, but just happen to mention the word somehow. Heck, Are there any good intermediate deck building resources available in magic turns up as a result, because the person is talkng about protecting against various strategies. The tag itself adds distinct value not achievable without it.
This searchability is value for regular SE users, who can use it to dig through questions on a topic for reference/linkage/citation or to aid in identifying duplicates. It also has value for new players, who could do the same thing to check if their kind of question has been asked before. (Many will not do this, but for those users there is no net damage caused by the tag's presence.)
The non-standalone nature does not matter much
I say that from experience on RPG Stack Exchange. Take a look at our front page, and you'll see loads of tags that describe categories of mechanics that exist in particular games. Many of these cannot stand alone. stress, aspects, natural-weapons and so on are close to meaningless outside the context provided by another tag, for instance. However, this is fine: the tags still provide excellent categorisation value, and we find them very helpful, and the fact we might never answer a question with that tag alone doesn't really cause any trouble. (We have a very active and self-reflective meta community. We're keenly aware of problems with several of our tags. These aren't among them.)
This could set precedent for adding tags for various ability keywords. Since there's a lot of those (106!) and many other named mechanics (e.g. creatures, legendaries, and so on) this could open up us to an awful lot of tags.
However, that does not necessarily mean that all of these potential tags will ever be created. Like RPG.SE, we should still follow the policy that tags are only added when someone finds it useful to add them, and tags aren't generally useful until there's a lot of questions on a topic.
For many ability keywords, we simply do not have any questions at all (e.g.
amplify,
rampage,
horsemanship,
modular,
fear)
or will be unlikely to get very many (e.g.
banding,
phasing
— the more of these I searched for, the more keywords I found we just don't have questions about).
That leaves us with many ability keywords that will simply never have a tag anytime soon.