Verified:GitHub

compaction-recovery

Persist mission state and checkpoints across LLM context compaction events.

44|9|Updated May 7, 2026

Search Tags:#task management #agent orchestration #checkpointing #session recovery #state persistence #context compaction #TODO survival

Ready to run

npx skills add https://github.com/Omar-Obando/qwen-orchestrator --skill compaction-recovery

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: compaction-recovery
Source: https://github.com/Omar-Obando/qwen-orchestrator/tree/main/skills/compaction-recovery
Command: npx skills add https://github.com/Omar-Obando/qwen-orchestrator --skill compaction-recovery

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill prevents critical task progress from being lost when the AI context window is compacted, so your ongoing implementation and TODO structure survive across sessions.

Core Features & Use Cases

Session state persistence: Writes mission, TODO, and progress details into the active session directory instead of relying on conversation memory.
Checkpoint-based recovery: Creates scheduled checkpoints and provides a clear recovery checklist to resume from the last verified state.
Self-contained TODO survival: Enforces TODO entries that remain understandable after summarization/compression.
Use case: You are working on a multi-step feature over a long session, and periodic context compaction would otherwise erase sub-task details and implementation status.

Quick Start

Use compaction-recovery to keep your active mission state, TODO progress, and checkpoints intact during long-running Qwen Orchestrator sessions.

Frequently Asked Questions about compaction-recovery

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

FAQPage Schema

How do I preserve TODO task state across LLM context compaction?▼

To preserve TODO task state across context compaction, you must write structured mission state and progress details into the active session directory rather than relying on conversation memory. This ensures sub-task details survive summarization.

How does checkpoint-based session recovery work for long-running agent workflows?▼

Checkpoint-based session recovery works by generating scheduled checkpoints of your mission state and resuming deterministically from the latest verified checkpoint. This prevents implementation progress from being lost during long-running agent workflows.

Why does my agent lose sub-task details after context window summarization?▼

Your agent loses sub-task details because standard context compaction summarizes away active files and implementation status. Without writing self-contained state to a session directory, TODO entries lose their meaning after compression.

Can I resume a multi-step feature implementation deterministically after a session ends?▼

Yes, you can resume multi-step feature implementation deterministically by loading a verified checkpoint from the active session directory. This restores mission-critical context without relying on the previous conversation history.

Do I need a specific orchestrator to maintain task progress during context compaction?▼

This state persistence mechanism is designed for Qwen Orchestrator runs. It requires writing structured state to the active session directory to ensure your mission and TODO breakdown survive across long-running agent sessions.

What is context compaction recovery for agent orchestration?▼

Context compaction recovery is the process of preserving mission-critical task state and TODO structure when an LLM context window is summarized. It ensures long-running agent workflows resume from verified checkpoints without losing implementation progress.