I have two metrics for measuring source code cohesion. To verify that both have some agreement can I use Cohen kappa for that? Both the metrics have values in range [0 to 1]. It's not necessary if both have the same values. If both the metrics are similar than I consider the common case and in case of conflicting as conflict. If one metric is decreasing and other is decreasing than both the metrics are common otherwise conflicting vice versa.
If I understood correctly, it seems to me that you have a continuous variable: the values can be any number in the interval $[0; 1]$. In this case Cohen's k is not appropriate.