ProseMirror/prosemirror

Is there a proper way to use prosemirror-commands within an InputRule, or are we supposed to compose transactions from scratch using prosemirror-transforms?

I'm trying to modify the original input rules for lists, so that I can not only toggle lists within paragraphs, but within other lists too.
For example, if we're in a bullet list, and 1. is typed, I would want to convert that first bullet list item into an ordered list.

My current approach is to use the commands given in schema list basic: use the liftListItem followed by a wrapInList.
However, I'm noticing that commands require the current state as input: updating the state results in a mismatched transaction, while not updating it between commands results in overwrites.

 // most of this is skeleton code of wrappingInputRule
      new InputRule(/^([-+*])\s$/, (state, match, start, end) => {
        let tr = state.tr.delete(start, end)
        let $start = tr.doc.resolve(start), range = $start.blockRange();
        let parentType = range.parent.type;
        if (parentType.name.includes("item")) {
          state = state.apply(tr); // mismatched transaction
          if (!sinkListItem(nodeType)(state, newTr => {tr = newTr})) return null
          $start = tr.doc.resolve(start), range = $start.blockRange();
        }
        let wrapping = range && findWrapping(range, nodeType, {})
        if (!wrapping) return null
        tr.wrap(range, wrapping)
        let before = tr.doc.resolve(start - 1).nodeBefore
        if (before && before.type == nodeType && canJoin(tr.doc, start - 1))
          tr.join(start - 1)
        return tr
      }),