← sources

How much does a good prompt actually matter?

source: reddit (reddit:ClaudeCode)published 2026-11-03 11:43ingested 2026-11-04 02:41

retained text

I know this is probably a retarded question for most of you but I am a bit new to the whole agent coding thing - I believe it is the future and I really enjoy playing with it. However I am always unsure if I should be improving my prompts when delegating work to Opus / Sonnet or is it enough if I explain what I want in my own words?

One of the work flows I've been doing it recently is writing my original prompt, then having Argon / Gemini 3.1 pro or Astra (I have the pro subs, ChatGPT work provided) to improve that prompt. I don't know if it makes a huge difference - I don't do it always though. For example, If goal is to just bug fix, I explain as best as I can what I see on the screen or what needs to be fixed - if I'm starting a new project within that project, I try to use this workflow.

  
What's the best case here? And a bonus question - how do I know how much reasoning "effort" I need? I try to get away with most tasks with Sonnet on High, and I go to Opus for big implementations or project starters.

document fetched 2026-11-04 00:40 · canonical https://reddit.com/r/ClaudeCode/comments/1xmqs8p/how_much_does_a_good_prompt_actually_matter