Caveman
Compresses AI responses by ~75% by dropping filler, articles, and hedging while preserving all technical accuracy and code blocks verbatim.
Install
npx skills add https://github.com/juliusbrussee/skills --skill cavemanWhat is this skill?
- 5 intensity levels from lite to wenyan-ultra
- Auto-clarity exceptions for security warnings and irreversible actions
- Code blocks and API names always verbatim — never compressed
- Persists across turns until explicitly stopped
- Matches user language (Portuguese input stays Portuguese)
Adoption & trust: 138 installs on skills.sh.
Who is it for?
Long technical sessions where verbose explanations slow you down
Skip if: Situations requiring careful security or legal communication
When should I use this skill?
User types /caveman or asks for terse, compressed responses
What do I get? / Deliverables
- compressed responses
- faster iteration
- reduced token spend
Recommended Skills
Grill Memattpocock/skills
Grill With Docsmattpocock/skills
Cavemanjuliusbrussee/blueprint
Lark Tasklarksuite/cli
Lark Workflow Standup Reportlarksuite/cli
To Prdmattpocock/skills
Journey fit
Primary fit
Caveman mode is a build-phase productivity tool that reduces token overhead during active development sessions with an AI coding assistant. It is agent-tooling — a meta-skill that modifies how the AI agent formats its own output, optimizing for signal density over readability polish.