I fixed Claude to MASSIVELY boost my productivity

I don't know about you guys, but I was getting so frustrated with how much text I had to read just to work out what Claude actually needed from me. So I added a few lines to my CLAUDE.md that make it end every turn with a short "What I need from you" list, or "Nothing needed from you right now" if it doesn't need anything. In this one I go through the rule and a few before and afters from my own chats, with some numbers to back it up so it's not just my vibes. Here's the exact text I use. Just copy it, paste it into your agent and say "add this to my global CLAUDE.md" (or AGENTS.md if you're using Codex): # What I need from you - End every turn where you're blocked on me, or where the next step needs me, with a short "**What I need from you**" section. Numbered, one concrete action or decision per item: exactly what to do, where (which site, app, file or person), and what to send back. e.g. "Open the Convex dashboard, add `OPENAI_API_KEY` under Settings → Environment Variables, then reply 'done'", not "I need the API key set up". - If there are several, put the quickest or most blocking one first, and say what you'll get on with in the meantime. - If you don't need anything from me, say "Nothing needed from you right now", and say whether the work is finished or still going. - Never bury a request for me in the middle of a long update. Oh and one thing I found when testing it, swap that example for a specific one from your own stack, as when the example was vague the asks got vague too. So what's the one line in your CLAUDE.md or AGENTS.md that you couldn't live without? Let me know down in the comments. Chapters 0:00 What do you need from me? 1:40 The fix: a few lines in my CLAUDE.md 2:39 Before and after, from my own chats 4:34 Bonus: it remembers my to-do list 5:05 Does it actually work? 39% vs 100% 5:53 What changed in my real chats 6:35 Setup: one copy and paste 6:51 What about Codex? 7:28 You're the bottleneck now 8:30 What do you think?