PEP 723 is provisionally accepted but it does not have a clear user-facing name that we can use to communicate about the core mechanism.
IMO we should have a specific suggestion for what to call this style of dependency specification, in conversation and documentation, ideally avoiding the need to state “PEP 723” or “pyproject.toml” when describing this style of dependency declaration.
Why is it bad to state “pyproject.toml”? I thought that the acceptance was provisional on a [run] table being added to pyproject.toml, and indeed that part of the rationale for the acceptance was that it would make things more coherent because the same information is in the script or in pyproject.toml.
So I think calling it “inline dependencies” or “in-script dependencies” is fine, and where convenient the documentation can also say something like “inline dependencies (in the same way they would be specified in pyproject.toml”. (And include a link to the documentation of pyproject.toml!)
The reasons for avoiding it are primarily-opinion-based so I’m not gonna spend time trying to convince anyone here. IMO it comes down to avoiding user confusion. The information that you can specify in the script is not equivalent to what you can with the pyproject.toml file. It is, definitionally, a strict subset. Using a reference to pyproject.toml can be misleading in that sense because it’s not “pyproject.toml in the script”, but rather a bit of (eventually-hopefully) shared support with the same schema being permitted across both TOML documents with hopefully similar semantics.