autoresearch-fleet

Official

Run an autonomous research loop with eval gates.

Authorquickcall-dev
Version1.0.0
Installs0

System Documentation

What problem does it solve?

It solves the problem of exploring a large optimization/search space without manual babysitting by autonomously iterating on code changes, evaluating them with a fixed harness, and keeping only improvements.

Core Features & Use Cases

  • Karpathy-style autonomous loop: iteratively edits a single mutable file, runs a deterministic evaluation harness, and commits or reverts changes based on results.
  • Plateau-triggered web search: detects consecutive discards/crashes and switches to a web-search-augmented prompt to break through knowledge ceilings.
  • Budgeted, provider-flexible orchestration: runs with claude, codex, or pi providers using per-iteration budget caps and stop conditions.
  • Git as state machine: improvement advances history while regressions reset hard, keeping a clean experimental trail.

Quick Start

Prepare a fleet root directory with fleet.json and a NEVER STOP program.md, then run: launch.sh <fleet-root> to start the autonomous overnight optimization loop.

Dependency Matrix

Required Modules

None required

Components

scripts

💻 Claude Code Installation

Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.

Please help me install this Skill:
Name: autoresearch-fleet
Download link: https://github.com/quickcall-dev/skills/archive/main.zip#autoresearch-fleet

Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
View Source Repository

Agent Skills Search Helper

Install a tiny helper to your Agent, search and equip skill from 471,000+ vetted skills library on demand.