What are the different optimization modes?

Light mode focuses on simple text cleanup; Medium provides balanced restructuring and standard transformations; Deep mode performs maximum compression through aggressive rephrasing.

Can it fix broken links or references in my files?

Yes, the skill includes 'Reference Integrity' rules that verify file paths, check URLs, and linearize circular references to ensure your documentation remains functional after optimization.

Does this skill change the meaning of my documentation?

No, the optimization rules are specifically designed to preserve facts and logic. Even 'Deep' mode, which rephrases aggressively for density, includes a quality checklist to prevent information loss.

How much can Text Optimizer reduce my token count?

Depending on the mode selected (Medium or Deep), you can typically expect a token reduction of 20% to 40% without losing the original meaning or logic.

Text & Prompt Optimizer

Name: Text & Prompt Optimizer
Author: kochetkov-ma

bykochetkov-ma

•

内容管理

Reduces token consumption and improves LLM instruction quality through automated text compression and structural optimization.

The Text Optimizer skill is a specialized tool for Claude Code designed to maximize token efficiency and model performance by applying 41 research-backed rules to prompts, documentation, and agent instructions. It offers three distinct modes—Light, Medium, and Deep—allowing users to choose between simple text cleanup and aggressive structural rephrasing that can reduce token counts by 20–40%. By optimizing Claude's behavior, improving reference integrity, and enhancing structural clarity using XML tags and imperative language, this skill helps developers create cheaper, faster, and more accurate AI-driven workflows while minimizing hallucinations.

主要功能

01Structural enhancements using XML tags, tables, and imperative formatting

02Three processing modes (Light, Medium, Deep) for varying levels of compression

03Validation of file paths, URLs, and cross-references to ensure integrity

04Automated token reduction of 20–40% without losing original meaning

0541 research-backed optimization rules across 6 specific categories

0616 GitHub stars

使用场景

01Optimizing Claude agent skills and instructions for faster model response times

02Refining complex system prompts to reduce API costs and improve instruction following

03Compressing long documentation files to fit within LLM context windows

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add kochetkov-ma/claude-brewcode text-optimizer

For use in Claude.ai and ChatGPT

Download Skill