Imagine stacking one coin above a number line for each data value. The outline of the stacks is the distribution's shape. Heights of 12-year-olds make a single mound in the middle whose sides fall away evenly: symmetric. The number of pets per household makes tall stacks at 0 and 1 with a thin trail stretching out to 6 or 7: skewed right, because the long tail points toward the high values.
Name a skew by its tail, not its peak. Scores on a very easy test pile up near 100 with a few low scores trailing off to the left: that is skewed left, even though the peak is on the right. The tail pulls the mean toward it, so the mean is larger than the median in right-skewed data and smaller in left-skewed data. A flat outline, like the results of rolling a fair die many times, is uniform; two separate mounds, like the heights of a group of first-graders and their parents, are bimodal.
Shape belongs to numerical data on a histogram or dot plot, where the left-to-right order of the bars means something. A bar graph of favourite fruits has no shape in this sense: the fruits could be listed in any order, and the "tail" would move with them. Before calling a graph skewed, check that its horizontal axis is a number line.
Think of it like this. Reading a distribution's shape is like looking at sand poured onto a ruler from above. A single cone is symmetric, a cone with sand trickling far to one side is skewed toward that side, a flat layer is uniform, and two cones mean two pourers. Unlike sand, small data sets can look bumpy just by chance.
Core idea. The outline of a histogram, whether symmetric, skewed, uniform or bimodal, shows how values are spread and which summaries describe them honestly; a skew is named after the direction of its long tail.