caveman

Compress AI agent output tokens using telegraphic prose at three levels.

3|Updated Mar 25, 2026

Search Tags:#context management #AI agent #token compression #telegraphic prose #response speed

Ready to run

npx skills add https://github.com/adihebbalae/Attacca --skill caveman-adihebbalae

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: caveman
Source: https://github.com/adihebbalae/Attacca/tree/main/.github/skills/caveman
Command: npx skills add https://github.com/adihebbalae/Attacca --skill caveman-adihebbalae

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill compresses agent output, reducing token spend without losing meaning, to help manage context window size and speed up response times.

Core Features & Use Cases

Token Compression: Compresses output to 65-75% of original by using telegraphic prose.
Intensity Levels: Offers 'lite', 'full', and 'ultra' compression levels for different contexts.
Context Management: Helps preserve context budget for long-running tasks or when requested by the Manager.

Quick Start

Set compression level to 'full' with the command: /set-compression full

Frequently Asked Questions about caveman

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

FAQPage Schema

How do I reduce AI agent token usage for long-running tasks?▼

Token compression reduces AI agent output by 65-75% using telegraphic prose, which preserves meaning while managing context window size and speeding up response times for long-running tasks.

How do I set the compression level for agent outputs?▼

You can set the compression level by using the command `/set-compression` followed by the desired level. The Skill offers three intensity options: 'lite', 'full', and 'ultra' for different contexts.

What is telegraphic prose for AI context management?▼

Telegraphic prose is a token compression mechanism that condenses agent outputs to 65-75% of their original size. It helps preserve context budget during long-running tasks without losing meaning.

Does token compression affect the meaning of agent responses?▼

Token compression reduces output size without losing meaning. By applying telegraphic prose, it condenses the text to manage context window size while maintaining the original message's integrity.

When should I use ultra compression over lite compression?▼

Ultra compression is suited for contexts requiring maximum context budget preservation, while lite compression applies lighter condensation. Choose the intensity level based on your long-running task requirements.