Simply flip them upside down.
It breaks down if you hit 5 (difficult to differentiate from 2), 6 (9) or 8 (symmetrical) but, since senary just needs 1 2 3, using 3 2 1 is good enough.
This solution is not Unicode compatible: there's no upside-down 1 in Unicode at all, and the 2 and 3 that the dozenalists somehow sneaked there are incompatible with nearly every font. The example I used was a quick CSS trick, and (if this page loads correctly) the font I use on the entire site has dingbats reskinned to upside down numbers.
It is the best for handwriting though, especially if your 2 has a big loop.
The dingbats, ① ② ③. They are a bit ugly and clumsily sized, but they are also easy enough to tell apart from the positives, and most fonts actually have them. I got this idea from here, and have the dingbats on my copy-paste pins for ease of use.
Z = ①, Y = ② and X = ③.
It just made sense to use the end of the alphabet for negatives like how programmers use the beginning of the alphabet for the larger than ten numbers. Plus, hexasenary conversion gets smoother.
If you want to distance yourself from the indo-arabic numerals, however, I have made a simple nice looking system that I use sometimes:
Each digit has as many lines as it represents, with the positive numbers being vertical and the negative ones horizontal. This system breaks really quickly for larger bases but the small ones are nice. It also is probably a dyslexic hazard but there's nothing I can do without ruining the symmetry.
By the way, we use commas to separate the integer and fractional parts of our numbers (as per ISO recommendation, periods are allowed but discouraged) because we aren't madmen like the Dozenal Society and their semicolons. We separate every four digits with a space, since six to the fourth is quite close to ten to the third.
