Clarification on Combining Code Writer and Executor Agents

[Riya-33]

Hi, I was reading the AutoGen documentation, which suggests that the code writer and code executor agents should be separate. I wanted to ask — is it considered bad to combine both roles into a single agent?

For example, I have a processor that generates a lot of preprocessing code and also needs to execute that code. Combining both roles in one agent reduces the overhead of managing conversations, especially since I plan to use a group chat layer on top of this. I'm just not sure if this approach aligns with best practices.

Would appreciate your thoughts — thanks!