This should use the existing pipeline-debugger skill and existing knowledge from the repo to figure out why a certain job failed by inspecting the log. It may require a sub-agent that is able to parse out the important details from a very long job log.