Need to understand how to tell the LLM to generate content into a new buffer (not modifying a file)

[bzorn]

I have a prompt with the following header:

prompt({ title: "Prompt critic", 
         description: "Given a goal and a prompt, evaluate the prompt",
         replaces: "nothing",
         maxTokens: 4000,
         model: "gpt-4",
         categories: ["tutorial"]  })

def("SUMMARY", env.file)

I just want the output to go to a buffer (not modifying any file). When I run this I see this in the new buffer that appears. For some reason, only part of the AI output (the text withing triple backticks) is extracted into the buffer.

But the AI output in the trace is this: