Have a few issues I'm experiencing with A*, looking for ideas to fix them

I did not understand much of what was written, Google translator did a poor job.

I have already come up with two methods.
1.How to avoid big agent to overlap obstacles
2.Is there a function like BFS that would collect all the nodes in a list with a hurdle?

Games are different with direct and indirect control, 3d, 2d, etc. Each of these games has its own ways to make jumps. They are not in my project and I can not tell you how to do them. But on the Internet, I think there are many videos on this topic.