that is why i was going to use chi-square, since the number of bars of a certain color in a bag is a count doesn't that make it attribute data? or am i wrong in thinking that this would still be considered attribute instead of continuous?
You need to consider the hypothesis that you want to test in addition to the type of data.
With count data, you have two options. The first is to treat it as count data and use a test for discrete data. But there is a second option. You can use the Freman-Tukey variance stabilizing transform for count data, and analyze the transformed data as variable data.
In your situation, there is also a third option. You are dealing with a mixture, so you can also treat the data as a proportion.
While you are correct that Chi-square is suited for count data, the hypothesis tested may not suit your needs. For the Chi-square Goodness of Fit test, the null hypothesis is an expected count of 10:10:10. If Chi-square is significant, you reject the null hypothesis and conclude that the counts do not equal 10:10:10. This will confirm that you have a problem, but no more.
You can perform a 1-sample t-test with the transformed data to test whether the count is significantly different from 10, or do a proportions test on the proportion data. This can provide directionality to the problem, but little more.
I recommend performing a Poisson capability study. This is typically used for defects (which are count data), but can be interpreted as you need. It will provide the percentage of a given color with the expected variation about that percentage.