What is the idea?

Currently, there are two systems we use in conda for parsing arguments: the traditional argparse setup where the majority of commands live and the conda shell.* commands which mostly live in the conda.activate module.

To make this easier to maintain, we would like to move the conda shell.* commands over to the main argparse system where the other commands live.

What should happen?

This will involve extensive updates to the way that the conda shell.* commands are invoked. We will first need to come up with a prototype of these changes and then make sure they are backwards compatible.

The existing tests for these commands should be reevaluated and a decision should be made whether more should be written.

Additional Context

Some of the inspiration for undertaking such work comes from the following issue:

• Core CLI argument parsing is overly fragile and breaks under common usage scenarios #10748