Can't we just use an existing DAG tool as the job engine? We do not want to use an existing DAG tool as the job-engine, because that makes it very hard to get step status, and therefore how to present it nicely and log it to persistence.
- p1: In a monorepo, if you init, it uses the git root. Instead, we want it to uses the current directory.
- p1: Remove all the Shell() calls, because they presuppose that we want to start in the calling directory