When should I trigger context compaction?

Compaction is recommended when context utilization exceeds 70-80% or when you notice a degradation in the AI's reasoning quality during long sessions.

What is the benefit of observation masking?

Observation masking replaces verbose tool outputs with short references after they've been used, preventing unnecessary tokens from cluttering future conversation turns.

How does KV-cache optimization work here?

It reorders context elements to place stable components like system prompts first, maximizing cache hits to reduce both latency and inference costs.

Does this skill increase the hard limit of the context window?

No, it extends the effective capacity by summarizing, masking, and partitioning data so that more meaningful information fits within the existing model limits.

Context Optimization

Name: Context Optimization
Author: sickn33

bysickn33

•

22,855

•

开发者工具

Extends effective AI context capacity through strategic compaction, observation masking, and KV-cache optimization to reduce token costs and improve performance.

This skill provides a comprehensive framework for managing AI context windows, enabling Claude to handle complex, long-running tasks that would otherwise exceed token limits. It implements advanced techniques such as summary-based compaction, observation masking to elide verbose tool outputs, and KV-cache-friendly data ordering to maximize inference efficiency. By focusing on signal preservation over raw data retention, this skill allows agents to maintain high performance across massive codebases and extended conversations while significantly reducing operational latency and API costs.

主要功能

01Intelligent context compaction and summarization

02Observation masking for verbose tool outputs

03KV-cache optimization for cost and latency reduction

04Sub-agent partitioning for isolated context workflows

05Trigger-based optimization based on token utilization

0622,855 GitHub stars

使用场景

01Managing long-running agent trajectories without performance degradation

02Analyzing massive documentation or codebases that exceed standard context limits

03Scaling production AI systems by minimizing token consumption and costs

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add sickn33/antigravity-awesome-skills context-optimization

For use in Claude.ai and ChatGPT

Download Skill