The ADEPT-design is to utilize a separate VM (the 'kommandir') for actually running the meat of "jobs". The master/entry-point VM (the 'exekutir') is only intended for executive functions (initiating, triggering, or scheduling). This arrangement is utilized heavily by other CI/CD systems, and generally considered best-practice for scaling reasons.
Unfortunately, the design of ADEPT is a bit wonky in this regard: The prescription for the kommandir VM is under the 'exekutir' ansible-directory. This was done because at the time, there was a desire to support different Ansible runtime environments/versions between the two. Also, since literally any job can/will create a common kommandir VM by default, it's source needed to be common.
In hind-sight, a better design, would be to allow jobs to specify a central Ansible directory / playbooks for this task. Utilizing some default if unspecified. In other words, nuke the separate 'exekutir' directory, and just treat the exekutir operations and kommandir operations as kind of meta-jobs. For example, have separate Ansible directories for them, and use the adept.py transition file exekutir.xn to run against them in a similar way to how the current job/job_path variables are handled.