Verified:GitHub

autoresearch

Coordinate autonomous AI research loops to run experiments and synthesize results.

Updated Apr 3, 2026

Search Tags:#workflow-orchestration #progress-reporting #continuous-learning #autonomous-research #autoresearch #two-loop #domain-skill-routing

Ready to run

npx skills add https://github.com/handsomelong922/my-codex-skills --skill autoresearch-handsomelong922

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: autoresearch
Source: https://github.com/handsomelong922/my-codex-skills/tree/main/skills/0-autoresearch-skill
Command: npx skills add https://github.com/handsomelong922/my-codex-skills --skill autoresearch-handsomelong922

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Coordinate autonomous AI research loops to run experiments and synthesize results.

Core Features & Use Cases

Orchestrates a two-loop architecture (inner loop for rapid experiments, outer loop for synthesis) and routes tasks to domain-specific skills.
Supports continuous operation via heartbeat mechanisms (Claude Code /loop, OpenClaw) to maintain momentum across sessions.
Generates human-ready progress presentations and papers and keeps an auditable trail of experiments.

Quick Start

Bootstrap autoresearch on your project by defining a research question and letting the two-loop workflow begin.

Frequently Asked Questions about autoresearch

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

FAQPage Schema

How do I coordinate autonomous AI research loops to run experiments and synthesize results?▼

A two-loop architecture separates research into rapid inner-loop experimentation and outer-loop synthesis, routing tasks to domain-specific skills to maintain momentum and synthesize end-to-end research results.

How do I keep continuous operation for autonomous research workflows across sessions?▼

Continuous operation for autonomous research workflows is maintained via heartbeat mechanisms like Claude Code /loop or OpenClaw, which keep the two-loop experimentation and synthesis processes running across sessions.

What is the best way to generate progress presentations and papers from automated experiments?▼

Generate human-ready progress presentations and papers directly from the outer-loop synthesis phase, which automatically compiles an auditable trail of experiments into research outputs.

Do I need specific domain skills to route autonomous research tasks?▼

You need domain-specific skills available for routing, as the two-loop workflow orchestrator dynamically assigns inner-loop experimentation tasks to specialized skills based on the research question.

How do I bootstrap an autonomous research project with a two-loop workflow?▼

Bootstrap an autonomous research project by defining a research question, which triggers the two-loop workflow to begin routing rapid experimentation and synthesis tasks automatically.

When should I not use a two-loop architecture for autonomous research?▼

A two-loop architecture is not suited for isolated, single-step tasks or research projects that lack a defined question, as it relies on continuous inner-loop experimentation and outer-loop synthesis to produce auditable results.