Reducing Output token usage using caveman skill

As you’re already aware for AI inference we pay based on the input and output tokens usage.

Most of the time, the output uses filter words which you don’t want especially when you’re coding. You just need TLDR version of it which in turn reduces your output token usage.

That’s where caveman skill comes in.

Once you install it, you’ll start getting crisp responses like this

How to get started?

If you want to use it on Claude Code

claude plugin marketplace add JuliusBrussee/caveman
claude plugin install caveman@caveman

And for other agents like codex, cursor, etc

npx skills add JuliusBrussee/caveman

👉 https://github.com/JuliusBrussee/caveman

There is a good video by Prime where he dive deeps. You might find it interesting