autoresearch

Coordinate autonomous three-team codebase improvement sessions in clean-room isolation.

52|2|Updated Mar 14, 2026

Search Tags:#testing #codebase #branching #session-management #experimentation #autonomous

Ready to run

npx skills add https://github.com/ehmo/autoresearch-skill --skill autoresearch-ehmo

💬 Prompt for ChatGPT / Claude Chat▼

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

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Coordinating autonomous, three-team codebase improvement sessions to safely discover, fix, and simplify issues without touching the main branch.

Core Features & Use Cases

Three-mode orchestration: narrow (goal-directed), broad (hypothesis-divergent), and sweep (quality sweep) with clean-room barriers between Red, Green, and Refactor teams.
Automated session management: per-session version stamping, area scoping, branch tracking, and results logging to support reproducibility and auditing.

Quick Start

Start a new session with /autoresearch narrow <path> to drive a measurable goal in a repository.

Frequently Asked Questions about autoresearch

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

FAQPage Schema

How do I run autonomous codebase improvement sessions without touching the main branch?▼

Run autonomous codebase improvement in clean-room isolation using three teams that operate on separate branches to find, fix, and simplify code. Strict revert procedures, version stamping, and branch-naming conventions keep your main branch untouched while per-track artifacts preserve experiment integrity.

What is the best way to automate code fixes and refactoring across different programming languages?▼

Automate code fixes across languages by applying a three-team orchestration that detects repository stacks and runs tests automatically. It applies to target repositories regardless of language, logging per-cycle results to support reproducibility and auditing for your codebase experimentation.

Can I scope automated codebase refactoring to specific directories or goals?▼

Yes, you can scope automated refactoring to specific areas. Use the narrow mode for goal-directed improvements on a target path, or choose broad mode for hypothesis-divergent exploration and sweep mode for comprehensive quality sweeps across your codebase.

Do I need configuration files to manage autonomous testing and experimentation tracks?▼

No, configuration is optional. Autonomous testing and experimentation runs by default with automated session management, but you can add an optional .autoresearch.yml file to customize team assignments and operating modes for your codebase improvement sessions.

How does session management work for isolated codebase experimentation and testing?▼

Session management handles per-session version stamping, area scoping, branch tracking, and results logging. This automated process ensures reproducibility and auditing by keeping clean-room barriers between the Red, Green, and Refactor teams during codebase experimentation.

What are the limitations of using autonomous teams for codebase refactoring?▼

Autonomous refactoring operates safely within clean-room barriers using strict revert procedures and version stamping, but requires adequate test coverage to validate fixes. Per-cycle results logging and branch tracking help monitor limitations and ensure experiment integrity across isolated tracks.