I’m working with MCP server tools and have run into a question about the ordering of prompts and tool calls:

In my workflow I use a tool (for example split_task_raw) to decompose a task. After the tool runs, I then receive a prompt that instructs the next tool. But logically the parameters for the next tool are already known as the output of the previous tool. So it feels like the prompt for the next tool could be embedded in the output of the previous tool, rather than waiting until after the tool call.

My questions:

Why in the MCP pattern is the prompt for the next step placed after the tool invocation instead of being part of the previous tool’s output?

Would it be better (or more efficient) to include the next‑step prompt in the output of the previous tool, so the chain is more continuous and there isn’t an extra prompt step?

Are there recommended “best practices” for when to issue prompts versus when to call tools in MCP workflows?

Any insight, examples or references would be greatly appreciated.