skills.rest
th-efool
th-efoolCOMMUNITY
Explore all skills by this author →
GitHub

caveman

Compresses model responses into terse caveman-style prose while preserving technical accuracy.

1|Updated Aug 8, 2026
Search Tags:#token efficiency#output style#token compression#classical chinese#caveman mode#concise responses
One-click install
npx skills add https://github.com/th-efool/SKILLS --skill caveman-th-efool
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: caveman
Source: https://github.com/th-efool/SKILLS/tree/main/caveman
Command: npx skills add https://github.com/th-efool/SKILLS --skill caveman-th-efool

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Long, verbose AI responses waste output tokens and slow down reading when you only need the technical substance. This Skill strips filler, articles, and pleasantries from every response while keeping code, error strings, and technical details exact. ## Core Features & Use Cases - Token Compression: Cuts output tokens by roughly 65% (measured) by dropping filler words, hedging, and pleasantries without losing technical content. - Six Intensity Levels: Choose from lite, full (default), ultra, and three Classical Chinese (文言文) modes for up to 80-90% character reduction. - Auto-Clarity Fallback: Automatically reverts to normal prose for security warnings, irreversible-action confirmations, and ambiguous multi-step sequences, then resumes compression. - Use Case: During a long debugging session, enable full mode so every explanation arrives as terse fragments like "Bug in auth middleware. Token expiry check use < not <=" instead of paragraph-length prose. ## Quick Start Tell the AI to enable caveman mode with /caveman to make all responses terse for the rest of the session.

Frequently Asked Questions about caveman

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reduce AI output tokens in chat responses?▼

Enable caveman mode with /caveman to strip articles, filler, and pleasantries from every response, cutting output tokens by about 65% while keeping code and technical terms exact. The mode persists for the whole session until you say stop caveman.

What compression levels does caveman mode support?▼

Six levels: lite (drop filler only), full (default, drop articles and use fragments), ultra (bare fragments), and wenyan-lite, wenyan-full, wenyan-ultra for Classical Chinese compression reaching 80-90% character reduction.

Does caveman mode change code blocks or error messages?▼

No. Code blocks, function names, API names, CLI commands, and exact error strings are always preserved verbatim. Only prose around them is compressed, and negation words like not, never, and only are never dropped.

When does caveman mode revert to normal prose automatically?▼

It drops to normal prose for security warnings, irreversible-action confirmations, multi-step sequences where fragments risk misreading, and when you repeat a question. Compression resumes after the clear part is done.

Does caveman mode affect written files like commits and docs?▼

No. Content persisted outside the chat, such as code, comments, commit messages, docs, and issue or PR text, is written in normal prose. Only conversational responses are compressed.

Related Skills

Explore More Software Engineering Skills→
junechen7414junechen7414

caveman

Compress assistant responses into terse, token-efficient prose while preserving technical accuracy.

Community
Intermediate
thesaifalitaithesaifalitai
8

token-optimizer

Optimize Claude token usage and model selection for sessions.

Community
Intermediate
kirillshilinkirillshilin

shai-create-tool

Scaffold SHAI tools with templates for instructions, skills, agents, hooks, and prompts.

Community
Advanced
JNRCHAYANJNRCHAYAN

caveman

Compresses assistant responses into terse, technically accurate language.

Community
Intermediate