We propose a consistent test for discrimination between a simple hypothesis and a left-sided alternative about the tails of Weibull-type distributions. The test is constructed using just $k$ higher-order statistics of the sample.