overnight

Maintain recoverable run state and checkpointing for long-running research workflows.

1|Updated Mar 23, 2026

Search Tags:#checkpointing #state-management #overnight #pause-resume #research-workflow #phase2 #phase1

Ready to run

npx skills add https://github.com/LinkDry/Link-Research --skill overnight

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: overnight
Source: https://github.com/LinkDry/Link-Research/tree/main/skills/overnight
Command: npx skills add https://github.com/LinkDry/Link-Research --skill overnight

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Phase-long research runs often stall due to lost state, manual restarts, and brittle orchestration. Overnight provides a recoverable checkpointing mechanism, strong autonomy, and clean pause/resume so long-running workflows can continue safely.

Core Features & Use Cases

Checkpointing and run summaries: Maintains canonical run state and can resume after interruptions.
Autonomy and sequencing: Coordinates steps across extended phases without constant human input.
Pause/Resume safety: Provides controlled suspension and resumption preserving project integrity.
Use Case: When a project spans multiple days and must recover gracefully after a crash or outage.

Quick Start

Kick off a long-running run with automatic checkpointing and safe pause/resume using Overnight.

Frequently Asked Questions about overnight

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

FAQPage Schema

How do I resume a long-running research workflow after an interruption?▼

To resume a long-running research workflow after an interruption, the mechanism uses automatic checkpointing to maintain recoverable run state. It reads project state and lifecycle fields, allowing paused tasks to safely continue from their last saved point.

What is checkpointing and how does it work for multi-day research phases?▼

Checkpointing for multi-day research phases works by continuously writing canonical run state to files like review-state.json and STATE.md. This preserves project integrity, enabling workflows to recover gracefully after crashes or outages without manual restarts.

How do I safely pause and resume extended Phase 1 or Phase 2 workflows?▼

To safely pause and resume Phase 1 or Phase 2 workflows, the orchestration applies controlled suspension and resumption behaviors. This manages run lifecycle fields and project state, ensuring extended sequencing can stop and start without losing progress.

Do I need any external dependencies to manage project state for autonomous research runs?▼

No external dependencies are needed to manage project state for autonomous research runs. The orchestration operates independently to coordinate steps across extended phases, handling reading and writing of project state through native files like STATE.md.

What's the best way to prevent lost state during long-running research orchestration?▼

The best way to prevent lost state during long-running research orchestration is using automatic checkpointing with clean pause and resume. This approach maintains recoverable run state, eliminating brittle orchestration and manual restarts caused by unexpected outages.