Caveman skill trims Claude output

Fun gimmick, real point: this is prompt compression dressed as comedy. If the token-savings claim holds in practice, it is a neat example of how output style alone can materially change cost and latency.

• –Targets Claude Code users with a simple install and trigger phrases like `/caveman` and `talk like caveman`
• –Preserves code blocks, exact technical terms, and error text, which is the only reason the style is usable for dev work
• –Sits closer to prompt-engineering and developer ergonomics than to a full product platform
• –The viral “why use many token when few token do trick” framing makes it easy to share, but the utility depends on whether teams actually want terse agent output
• –As open-source release, it is more of a reusable workflow hack than a standalone app