context-compactor

Offload context to disk with checkpoints and recovery for long tasks.

Updated Mar 5, 2026

Search Tags:#planning #context #memory #checkpoint #recovery #compaction #offload

Ready to run

npx skills add https://github.com/bouwerp/agentic-framework --skill context-compactor-bouwerp

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: context-compactor
Source: https://github.com/bouwerp/agentic-framework/tree/main/skills/context-compactor
Command: npx skills add https://github.com/bouwerp/agentic-framework --skill context-compactor-bouwerp

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill helps AI agents manage and preserve context across long-running tasks by offloading data to disk, checkpointing progress, and enabling clean recovery after compaction.

Core Features & Use Cases

Offload large context to disk: Store context that would overflow RAM to the filesystem, keeping the in-memory window lean.
Checkpoint and progress tracking: Periodically write checkpoints and progress to files to enable resumption after interruptions.
Post-compaction recovery: Read checkpoint and git history to reconstruct state after context reduction.

Quick Start

Save your current session status to progress.md and offload large artifacts to scratch files before switching tasks.

Frequently Asked Questions about context-compactor

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

FAQPage Schema

How do I preserve AI agent context during long-running coding tasks?▼

To preserve agent context during long-running coding tasks, offload large data to disk-backed storage and write progress to checkpoint files, keeping the in-memory window lean and enabling recovery.

What is the best way to recover agent state after context compaction?▼

Recovering agent state after context compaction requires reading saved checkpoint files and git history to reconstruct the previous working state and resume the multi-step task seamlessly.

How do I checkpoint agent progress to prevent losing information?▼

To checkpoint agent progress, periodically write session status and progress tracking data to files on the filesystem, ensuring critical information survives memory limits or interruptions.

Does managing context overflow require explicit disk-backed storage patterns?▼

Yes, managing context overflow reliably requires explicit disk-backed storage patterns, such as plans and notes saved as scratch files, to offload data that exceeds in-memory limits.

When should I offload in-memory context to the filesystem?▼

You should offload in-memory context to the filesystem when long-running tasks exceed available RAM limits, proactively writing artifacts to disk before compaction or task switching occurs.