Don't understand why everyone want to have specific agents to write software
retained text
Hello I am a software engineer. I'm wondering about something with all the over engineerered around specific agents workflow. Some of my colleagues made their own workflow by designing an architecture of multiple agents. Typically, an orchestrator, an implementer, a reviewer, a tester. But for me, it's unnecessary and I don't understand the need to do that. LLM has to take context into input. Now we have agent skills which allow agents to be specialized in one specific workflow (for a library, a method of testing). And the specification said that we have description in metadata, which describes when the skill should be invoked. So for me just use the standard agent mode in GitHub copilot with the skills you want you apply in your project is self sufficient. Engineers who write GitHub Copilot and Vs code have already thought about all the workflow from the prompt to the result of the generation. My question is: is it me who is dumb ? Or is that my colleagues who over engineerered something ?
document fetched 2026-09-18 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1wiyhpz/dont_understand_why_everyone_want_to_have