It's common nowadays to annotate methods with the abstract Sequence[T] rather than list[T] when mutation is not required. However, I noticed that one sometimes encounters / is tempted to write the following pattern:

from collections.abc import Sequence

def test(args: Sequence) -> None:
    if not args:  # <-- potentially dangerous.
         ...

This is inherently unsafe at the moment, since it assumes that bool(...) behaves the same for any subclass of collections.abc.Sequence as it does with builtins.list or builtins.tuple, but nothing stops a custom subtype to implement __bool__ and override this behavior.¹

The documentation, both the data model docs and the collections.abc docs, appears silent about any expected behavior for bool(sequence). Given the prevalence of code that uses such patterns, I think one should either:

1. Make the boolean behavior on list/tuple officially a part of the specification of sequence types in one way or another.²
2. Clarify and document that sequences need not behave like bool(list) or bool(tuple), so that linter-libraries can justifiably trigger warnings in cases like the code above, and guide the user to use if len(seq) == instead.