autoresearch

Automate goal-directed codebase iteration with atomic changes and verification.

9|1|Updated Mar 31, 2026
One-click install
npx skills add https://github.com/Maleick/AutoResearch --skill autoresearch-maleick
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: autoresearch
Source: https://github.com/Maleick/AutoResearch/tree/main/plugins/autoresearch/skills/autoresearch
Command: npx skills add https://github.com/Maleick/AutoResearch --skill autoresearch-maleick

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Autonomous, goal-directed iteration optimizes codebases by autonomously proposing small, atomic changes, verifying outcomes, and retaining only improvements.

Core Features & Use Cases

  • Autonomous loop: Modify → Verify → Keep/Discard → Repeat to drive progressive improvements.
  • Branch isolation and deterministic verification: each iteration creates an isolated autoresearch/<timestamp> branch and logs results for traceability.
  • Learn-and-record workflow: supports a root SKILL bundle with references, scripts, and assets to guide the Codex plugin.

Quick Start

Provide a Goal, Scope, Metric, and Verify command to start an autonomous autoresearch run.

Frequently Asked Questions about autoresearch

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

FAQPage Schema
How do I automate code improvement iterations on a git-backed project?

Autonomous code improvement iterations are driven by providing a goal, scope, and verify command. The engine proposes small atomic changes, verifies outcomes using a numeric metric, and retains successful improvements on an isolated git branch.

What is autonomous goal-directed iteration for codebases?

Autonomous goal-directed iteration is an automated loop that modifies code, verifies the outcome, and keeps or discards changes. It progressively optimizes codebases by retaining only verified improvements based on a deterministic metric.

Do I need a specific verification setup to start automated code iteration?

Yes, automated code iteration requires a Verify command that outputs a numeric metric to evaluate changes. An optional Guard command and a Scope glob to define affected files are also supported for deterministic branch isolation.

How does the automated iteration engine handle git branch isolation?

The iteration engine handles git branch isolation by creating a dedicated autoresearch branch for each run. It logs results and verifies outcomes deterministically on this isolated branch to ensure traceability and prevent untested changes.

Can I use automated iteration without a deterministic verify command?

No, a deterministic verify command is strictly required to automate codebase iteration. The engine relies on the numeric metric output from this command to decide whether to keep or discard each atomic modification.

What is the best way to define the affected files during an autonomous code iteration run?

The best way to define affected files is by providing a Scope glob during setup. This restricts the autonomous modifications to specific files, ensuring the verify command and guard checks only evaluate relevant codebase changes.