context-compression

Compress long agent conversations into anchored structured summaries preserving session context.

Updated Mar 13, 2026

Search Tags:#memory #compression #token-optimization #context-management #long-running-sessions #anchored-summarization

Ready to run

npx skills add https://github.com/marinvch/ai-os --skill context-compression-marinvch

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: context-compression
Source: https://github.com/marinvch/ai-os/tree/main/.agents/skills/context-engineering-collection/skills/context-compression
Command: npx skills add https://github.com/marinvch/ai-os --skill context-compression-marinvch

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Context management for AI agents with long-running sessions is challenging because token usage grows quickly and important details risk being forgotten. This skill provides a structured approach to compressing context while preserving essential information such as session intent, file changes, decisions, and next steps.

Core Features & Use Cases

Anchored iterative summarization that merges new spans with existing summaries to prevent drift.
Structured sections for session intent, files modified, files read, decisions, and current state.
Memory-friendly compression with guardrails for accuracy and completeness.
Use Case: Long-running debugging sessions where argumentation and file edits must survive context truncation.

Quick Start

Provide a structured, anchored summary of the current interaction span and merge it into the persistent context memory.

Frequently Asked Questions about context-compression

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

FAQPage Schema

How do I compress long agent conversations to save tokens?▼

Compress long agent conversations by applying anchored iterative summarization to merge new interaction spans with existing summaries, preventing drift while reducing token usage in long-running sessions.

What is anchored iterative summarization for context management?▼

Anchored iterative summarization is a context management method that merges new conversation spans with existing persistent summaries to prevent information drift and preserve essential session intent.

How do I preserve debugging context when hitting token limits in long-running sessions?▼

Preserve debugging context during long-running sessions by compressing conversations into structured sections that track session intent, files modified, files read, decisions, and current state.

Can I prevent information loss when truncating context windows during code exploration?▼

Prevent information loss during code exploration by using memory-friendly compression with explicit guardrails that enforce accuracy and completeness when reducing conversation context.

Does context compression work without external dependencies?▼

Context compression works without external dependencies, utilizing internal scripts and references to manage structured file tracking and anchored summarization directly within the agent environment.

What are the limitations of using anchored summarization for context management?▼

Anchored summarization requires structured input spans to merge effectively, meaning unstructured or highly fragmented conversational drift may reduce compression accuracy and limit context preservation guardrails.