In my close vote queue, I sometimes get questions on high-level quantum mechanics or general relativity that I don't know enough to judge. This is annoying to me personally but on a larger scale, it may cause problems as people who are not very familiar with a subject may vote to close before going through it.
Hence, I propose that we get the ability to close vote questions with certain tags only after getting some amount of points from answering and asking questions related to that tag.
One does not need to be an expert in quantum-mechanics to see that some of the questions are clearly subpar and need to be closed. When in doubt, don’t be afraid to use the skip button! Alternatively, use the review queue tag filter to focus on questions you do feel comfortable in, e.g. newtonian-mechanics and thermodynamics.
Your suggestion leads to situations where a niche tag would not even have enough expert users with the ability to vote to close to actually close the question ...
This is already possible via a different system: the close-vote review queue already comes with the possibility to filter the questions it can show you.
This can be on the close reason and on the tags. This allows users the possibility to focus on the parts of the queue where they are most effective and most comfortable that they are making informed decisions.
(That said, of course, close-voters are still required to vote responsibly, which includes clicking skip if they don't know enough about the material to judge whether the question should be closed. This is an intrinsic feature of close-voting, and it can never be supplanted by automated systems. As such, setting tag-specific thresholds to restrict close-voting would be counter-productive and do little more than harm the system.)
That said, the fact that you missed this feature is indicative of one of the biggest problems with the queue: the filter functionality is extremely hard to discover, and does not offer the functionality that it could (or should) to help close-voters be at their most effective.
Fortunately, this aspect of the software is currently under active development: see the project overview, onboarding-improvements, review bans and filtering features and documentation threads over on Meta Stack Exchange. There are many improvements in the works and many more under consideration (so e.g. my own suggestions are here). If you feel that the current system isn't working well (such as e.g. the filtering function is too hard to discover), then please add your voice to the debate on MSE, either by voting on existing commentary or adding your own if your concerns have not yet been voiced.
