What type of projects does the OP favour and what is the representative budget? The optimum mix and form of the test gear will change a bit depending on what he wants to do, and which operations are one-off, occasional or regular. Very good PC-based systems are not especially cheap, although there is a fair range to choose from these days. I've used a few PC-based instruments (including Picoscope and NI devices) and been quite impressed. However, for a lot of bench work, I much prefer separate instruments with well-designed ergonomics.
Assuming the basics such as a decent soldering station and bench power supply, my list would be similar to Dave's (above) although, despite having and liking some TTi gear, I'd spring for one of the small-ish Rigol DSA800 series spectrum analysers with a tracking generator for about the same cost as the TTi hand-held unit. A scope that you're comfortable with is also a must, and a low-cost VNA - such as the nanoVNA - covers a lot of ground. I've "ruggedized" mine, making it more convenient for both indoor and outdoor use.
As a further comment on the RF spectrum analysis, think also about what your transceiver and bandscope can do for you. For example, in some recent work looking at the phase noise of class D MF amplifiers, I found the FFT scope in my TS-890S gave the bench real-time spectrum analyser a run for its money, noting the absolute calibration advantages of the latter. As others have suggested, a low-ish cost SDR can also be useful and I sometimes use the RSPdx unit for quick, wideband off-air scans.
Good, small RF signal generators are not that easy to come by, although for many applications some of the eBay Chinese offerings based on various flavours of DDSs will get the job done. I covet a better unit than my Rigol DSG generator but, not having the space for a boat anchor, and noting the costs of units with significantly better phase noise, I'm making do.
73, Peter.