Hackability?

Your documentation states:

* Hackable: build extensions on top of it however you like

However, after attempting to "hack" this library, I've found that this goal isn't being met satisfactorily. Many outstanding issues also seem to reflect this, a significant number caused (imo) by #53. The tight coupling between parsing, IO, and execution, in addition to a significant amount of essential API elements being non-constructable (Argument/Option/CLIEntity constructors are internal) without going through an ArgParser, make it nearly impossible to "hack" useful/unexpected behaviors on top of this library.

Can someone elaborate on the intention here? I freely admit I could be interpreting this goal incorrectly.

The issue names the documentation's “Hackable” claim, #53, and constructors for Argument, Option, and CLIEntity; start by reading those materials and the parser, IO, and execution boundaries. Done means the intended extension model and supported construction paths are clarified, with any resulting scope agreed.

Written by the indexing model from the issue text.