The assistant wraps artifacts in nice frames that allow you to download, copy, or maximize the content. When the content is markdown, artifacts can be incorrectly split across frames. The user has to manually reassemble the artifacts.
This primarily seems to happen with nested code fences, but it can happen with headings and other elements as well.
Examples:
These are both from this assistant thread. I haven't shared it publicly, hopefully the link is sufficient to support troubleshooting.
I expect assistant to wrap frames around the entire content of each artifact so I can retrieve the artifact with the download / copy buttons without manual reassembly.
