caveman — Ultra-Compact Mode

• Trigger: /caveman or keywords "caveman", "less tokens", "be brief", "talk like caveman"
• Best for: Saving context in long sessions, quick Q&A, tight token budgets
• Not for: Tasks needing detailed explanation, complex code generation, teaching

Overview

caveman switches Claude Code to ultra-compact communication. It drops filler, articles, and pleasantries while keeping full technical accuracy. Token usage drops ~75%.

Trigger Conditions

caveman auto-triggers when you:

• Mention "caveman mode", "less tokens", "be brief"
• Use the /caveman command

Examples

Example 1: Quick Q&A

/caveman What's new in React 19?

Normal → "React 19 brings several important new features including..."
Caveman → "React 19: Server Components stable, Actions API, use() hook, ref as prop, Document Metadata."

Example 2: Mid-Session Switch

/caveman Continue the task

All subsequent replies are compact.

Example 3: Compact Code Review

/caveman /review src/utils/

Comparison

	Normal	Caveman
Token usage	100%	~25%
Technical accuracy	Unchanged	Unchanged
Code quality	Unchanged	Unchanged
Explanation depth	High	Low
Best for	Learning, docs	Fast execution

Notes

• Caveman mode doesn't affect code generation — only conversation
• Exit: say "exit caveman mode" or trigger another Skill
• Not suitable for tasks needing detailed reasoning (e.g., teaching)
• When combined with loop, ensure critical info isn't missed

Related Skills

• loop — Run compact background tasks
• handoff — Exit caveman before creating handoff docs