context-window-management

This skill provides strategies for effectively managing LLM context windows. It addresses challenges like token limits, context rot, and information loss. It focuses on curating the right information, understanding the serial position effect, and choosing between summarization and retrieval techniques.

It offers capabilities like context engineering, summarization, trimming, and routing. The skill also provides patterns like tiered context strategy and intelligent summarization, along with anti-patterns to avoid, such as naive truncation and ignoring token costs.

By leveraging this skill, developers can build more robust and efficient LLM applications that effectively handle long conversations and complex information.

What It Does

Provides strategies and techniques for managing the context window of large language models (LLMs), including summarization, trimming, routing, and prioritization to optimize performance and avoid common issues.

When To Use

When building LLM applications that handle long conversations, complex information, or are prone to exceeding token limits or suffering from context rot.

Installation

Copy SKILL.md to your skills directory