caveman
Reduce output tokens by simplifying language while maintaining technical accuracy.
- Type
- Agent Skill
- Open source
- Yes
- GitHub Stars
- ★ 76.3k
- Source
- skill-github
- Repository
- github.com/JuliusBrussee/caveman
Overview
caveman is a Claude Code skill/plugin that enables AI to communicate in simplified language (e.g., caveman-style), reducing output tokens by approximately 75% while preserving full technical accuracy. This skill works with various agents, including Codex, Gemini, Cursor, and more. Users can switch between different compression levels using simple commands, significantly lowering token costs. Ideal for developers seeking efficient communication and cost reduction.
Capabilities
- ▪Reduces output tokens by 75%
- ▪Maintains technical accuracy
- ▪Supports multiple compression levels
- ▪Multi-language support
Use cases
Setup
curl -fsSL https://raw.githubusercontent.com/JuliusBrussee/caveman/main/install.sh | bash (macOS / Linux / WSL / Git Bash) or irm https://raw.githubusercontent.com/JuliusBrussee/caveman/main/install.ps1 | iex (Windows PowerShell 5.1+)
This information was compiled by AI from public sources and may contain inaccuracies — please refer to the source.
FAQ
How do I trigger caveman mode?
Enter /caveman or say 'talk like caveman'.
What compression levels are supported?
Supports lite, full, ultra, and wenyan levels.
Related skills
mantis
Provides security review capabilities for AI coding agents, automatically discovering, reproducing, and fixing vulnerabilities.
sast-skills
Transform your AI coding assistant into a SAST scanner to automatically detect vulnerabilities in code.
Lightswind-UI-Library
AI-native CLI-first React component library with MCP Server support.
scientific-agent-skills
Transform any AI agent into a scientific assistant with 147 ready-to-use research skills.
susi_alexa_skill
A skill that enables question-and-answer interactions between Alexa and Susi AI.
claude-context
Provides code search capabilities for Claude Code, turning the entire codebase into context.