The 3 prompts
Bad answers usually mean Claude's guessing. These three fix the guessing in one chat, without spinning up a single extra agent. They work in normal Claude chat (and ChatGPT).
Interview Me First
For anything that depends on your business, your audience or your taste. Paste this at the end of your request. Claude asks what it's missing before it writes a word, so the first answer is built on your details, not its guesses.
The Mini Arena
For when you want the best version, not the first one. It's the same idea as the 100-agent arena (different approaches compete, get critiqued, one wins), but it happens in a single reply.
The Clean Retry
For when the answer's off and you're about to type “no, try again”. Don't. Click edit (the pencil) on your original message, paste this underneath it, fill in the brackets and resend. Claude starts again from your better message, and the chat doesn't keep growing.
Why edit beats “try again”
Every follow-up message makes the chat longer, and Claude counts conversation length against your usage limit. Five rounds of “not quite” costs you more and usually still misses, because the missing context never arrived.
Why Claude gives bad answers
Claude isn't dumb. It's guessing. When your request is missing context, it fills the gaps with the most average answer it can find. Asking again doesn't add context. Giving it the context does.
What you typed
“Write an Instagram caption for my new offer.”
What Claude needed
- Who the offer is for
- What it is and what it costs
- How you sound (paste a caption you've written)
- How long, and what to avoid
Hot tip
If you keep pasting the same background (your offer, your audience, your voice), put it in a Claude Project once. Every chat inside that Project starts with it, so you stop retyping it and Claude stops guessing.
The arena skill, in plain English
The arena skill is a free add-on for Claude Code. You give it one task and it runs up to 100 copies of Claude on it, each with a different strategy. They pair off, attack each other's answers and rewrite their own, and a judge scores every match until one answer is left.
How the bracket shrinks
Why it maxes out your plan
The skill's own instructions say it straight: “The skill is free. The tokens are yours, and 100 agents is a lot of them.” Anthropic's engineering team found multi-agent setups use about 15 times more tokens than a normal chat. On Pro and Max, Claude chat and Claude Code share the same limits, so a big run in Claude Code eats into your normal chat time too.
When it's actually worth it
Check your usage before it bites
Before you run anything big (an arena run, a long research job, a stack of file uploads), check how much you've got left.
Open Settings → Usage
You'll see two progress bars: your current 5-hour session and your weekly limit.
Low on the weekly bar?
Save the big jobs for after it resets, and lean on the Mini Arena instead. One reply, three competing answers.
Ready to go deeper?
Where to go from here.
Wright Mode Membership
Join a community of women entrepreneurs implementing AI and automation in their businesses. Live calls, templates, and ongoing support.
Claude Masterclass
Learn how to use Claude like a pro. From prompting fundamentals to building real workflows that save hours every week.
Claude Code Masterclass
Go beyond the chat interface. Build automations, process data, and create tools with Claude Code — no developer experience needed.