Can I judge skewness from a scatterplot with bivariate data in R? 
plot(filterdacsom5$Median_Income,filterdacsom5$Total_Population,
     xlab="Income", ylab ="Population",
     main="Demographics plotted for all zip codes in 2017 ",col="red" )

I am new to R and understanding skewness.
This is a scatterplot of Median_Income on the horizontal axis and Total_Population on the vertical axis. From the scatterplot, Is it safe to say that the data is left or negatively skewed?
 A: 
Is it safe to say that the data is left or negatively skewed?

No, it is not safe: Firstly, the appearance of the plot is of positive (right) skew, not negative (left) skew.  Regardless, you need to be careful here because there is over-plotting, which means that you can't actually see what is going on in that big red mass in the middle.  Although it is unlikely, it is possible that this red mass of points is hiding concentrations of points that would detract from positive skewness of one or both variables (or it might even induce negative skewness).  To get a better assessment of the skewness of the two variables I would recommend constructing kernel density plots of the variables of interest and calculating the sample skewness of these variables (R code for this below).
library(moments);    #Make sure you have installed this package first

DATA <- filterdacsom5;

#Check skewness of median income
skewness(DATA$Median_Income);
plot(density(DATA$Median_Income));    

#Check skewness of population
skewness(DATA$Total_Population);
plot(density(DATA$Total_Population)); 

Note that the scatterplot gives you information about the joint distribution of the variables, which you won't get from individual density plots.  If you would like to see a better representation of the variables in the scatterplot, I would recommend you adjust it to deal with over-plotting --- e.g., use alpha-transparency or a contour plot.
A: This approad may be missleading and this is why.
The scatterplot can tell you something about the distribution of each variable. But the scatterplot also tells you something about the relationsship between two variables, which can lead to problems if one is making an interpretation about one of the variables alone, e.g. interpreting the skewness.
Let's assume some data with heteroscedasticity where y doesn't have negative values (like in your example). The resulting plot could be like this:

The resulting plot looks relatively close to the provided plot
and the plot suggests that x is skewed although this is actually not the case since x has an uniform distribution (see code for data generation below) as the histogram for x shows:

Thus, the relationship between the variables can result in an missleading scatter plot in terms of interpreting the distribution of one variable.
The code I used for the plot:
set.seed(568)
x      = rep(1:10000,2)
a <- 20000
b      = -2
sigma2 = x^2
eps    = rnorm(x,mean=0,sd= rev(sqrt(sigma2))) # heteroscedasticity
y      = a + b*x + eps
y[y<0] <- -y[y<0] # no negative values in y
plot(x, y)

EDIT:
I agree with Ben that the transparency and overplotting are important in this case and this is why I choose such a big sample size for my example. Using transparency fot the same data is less missleading.
plot(x, y, col = alpha("black", 0.05))


