The current approach of limiting CPU usage of comskip by restricting the number of threads is inefficient as it can result in resource underutilization and in some cases over-utilization when the user kicks in jobs manually. The mechanism to limit CPU usage by process has been available in unix kernels for quite some time:
tldr; instead of executing
$ comskip foo bar baz
the job should be started as
$ nice comskip foo bar baz
Alternatively the option asking comskip to play nice could be used
$ ./comskip -h
...
-n, --playnice Slows detection down