Particularly in final year subjects, you may need to consult the literature to put your project/assignment into some context, or to give credit to what inspired your approach.
Your lit review should contain high quality and reputable sources. The number of citations can sometimes be a good indicator of this, sometimes not so much. Get a feel for what the most reputable venues in your field are using a resource like: https://scholar.google.com/citations?view_op=top_venues
Even top venues can publish rubbish articles, though they tend not to. You should get a feel very quickly for if the paper or book you're reading is any good.
When does the citation rule break? If the paper is new (published in the last year or two) - it may not have found its audience yet and so may not have many citations. If it is more than 1 year old, you would expect it to have several, from authors who did not write the original paper.
Can I use arXiv papers? Yes - in some fields like machine learning, the field moves quite quickly and so research is often disseminated via pre-print services like arXiv. You should check to see if the paper has been published however and find a non-arXiv link if you can.
Single author papers from people with fairly few papers, or citations are often not great - be especially cautious of these.
Papers published by low ranking university journals or department journals are almost always inappropriate sources (and sometimes picked up by ghostwriting services to pass as undergraduate papers!).
Remember: it's not true unless it's in LaTeX... this is a joke of course, but sometimes it sets a very good bar - if someone is a serious researcher in the engineering field or the sciences they will use LaTeX.
However you like in the context of your project, with one key qualifier: tell us why this resource is relevant or helpful. You should consistently relate key works to your own rather than just recapping what they did.
Works that provide background or give evidence for a point you are making without having direct impact on your own work can be mentioned in passing. For example:
"Thermal insulation for spacecraft has traditionally been manufactured from layered mylar-fibreglass blankets [1, 2, 4, 13], however more recent approaches have sought to use lightweight composite fabrics [2, 6]. These passive thermal regulation techniques are being used in conjuction with active technologies..."
An example of some seminal work you wish to introduce alone could be as follows:
"Oppenheimer et al. [12] proposed that a stable and quasi-Newtonian unstable state could exist in a low mass neutron core. Our simulation, however, assumes stability to prevent collapse of Thorne-Żytkow objects under perturbations introduced during their formation.
Important works should be dealt with in a little more detail, fleshing key points about their method and what makes it important, but all works should be related back in the context of what you are writing about.
It is not enough to give a poorly written recap of what a particular paper - it gives no bearing or credibility to the work you have performed. For this, quite simply, you'd do better to not have written anything at all. Do not expect to receive high marks if it is clear you are regurgitating abstracts.
It is often said in science that researchers stand on shoulders of giants. In actuality, we stand on a giant pile of papers which have enabled our work. As an undergrad you are always working somewhere halfway up that pile - where things have been done before and since. In an honours thesis you might work nearer the top of the pile, while in first or second year you will be working further down. You have the largest selection of papers to choose from, and can even reference subsequent works which have improved (to inspire your future work).
If you are writing a lab report, something like 5-10 sources is probably expected. If you are writing something up in an academic article format somewhere more than 20 or 30 is probably appropriate. A thesis, 60 plus perhaps. It really all depends on your field, and how many works you rely on. You should in any case always be honest - inflating your citation count is blatantly obvious as your marker will often known the literature landscape better than you do.