Repolode
Submit
wanshuiyin

wanshuiyin/Auto-claude-code-research-in-sleep

Markdown-only skill files that let Claude Code, Codex and other LLM agents run autonomous ML research loops: idea discovery, cross-model review and experiment automation.

View on GitHub
Stars
18,796
Forks
1,527
Open issues
73
Last commit
23 hours ago

What is it?

ARIS (Auto-Research-In-Sleep) is a set of Markdown skill files for autonomous machine learning research. Each skill is a plain prompt-style workflow that an agent such as Claude Code or Codex CLI can run, so there is no separate framework or runtime to install and no lock-in to one vendor.

The core pattern is an executor model that does the work and an independent reviewer model that critiques it in a loop. The project covers idea discovery, paper and code reading, experiment execution and writing support. The default reviewer runs through Codex MCP, and the README describes alternative model combinations, including ModelScope-hosted models that need no Claude or OpenAI API.

The repository also ships ARIS-Code, a standalone CLI, along with plugins for Claude Code and Codex CLI, and an ARIS-Monitor macOS widget that flags sessions waiting for approval.

Key features

  • 01Markdown-only skills that run inside Claude Code, Codex CLI, Cursor and other agent environments
  • 02Cross-model review loop where an executor model is critiqued by an independent reviewer model
  • 03Idea discovery, paper reading and experiment automation workflows
  • 04Default reviewer via Codex MCP, with alternative model combinations supported
  • 05Standalone ARIS-Code CLI and plugins for Claude Code and Codex CLI

Use cases

  • Running an unattended literature-driven idea search and experiment loop for an ML project
  • Getting a second model to critique a research plan or draft before submission
  • Automating repetitive experiment runs while an agent works overnight
  • Using an agent workflow without committing to a single model vendor

Things to know

  • The default setup relies on Claude Code as the executor and Codex CLI as the reviewer, which need access to paid model services.
  • A recent codex-cli release removed the entry point the reviewer used, so existing installs need one re-registration step.

In collections

Skill packs for Claude Code and CodexClaude Code tools, skills & pluginsAI agent skill libraries

Topics

#ai-research#llm#claude-code#agent-skills#machine-learning#paper-review#research-automation#ai-tools#aris#autonomous-agent

FAQ

What is Auto-claude-code-research-in-sleep?
Markdown-only skill files that let Claude Code, Codex and other LLM agents run autonomous ML research loops: idea discovery, cross-model review and experiment automation. ARIS (Auto-Research-In-Sleep) is a set of Markdown skill files for autonomous machine learning research. Each skill is a plain prompt-style workflow that an agent such as Claude Code or Codex CLI can run, so there is no separate framework or runtime to install and no lock-in to one vendor.
Is Auto-claude-code-research-in-sleep free to use? What license does it have?
Auto-claude-code-research-in-sleep is open source under the MIT license. Check the license terms before using it in commercial products.
Is Auto-claude-code-research-in-sleep still maintained?
Actively maintained. The last commit to Auto-claude-code-research-in-sleep was on Nov 6, 2026.
What can I use Auto-claude-code-research-in-sleep for?
Running an unattended literature-driven idea search and experiment loop for an ML project Getting a second model to critique a research plan or draft before submission Automating repetitive experiment runs while an agent works overnight Using an agent workflow without committing to a single model vendor
Where can I get Auto-claude-code-research-in-sleep?
The source code and installation instructions are on GitHub: https://github.com/wanshuiyin/Auto-claude-code-research-in-sleep

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects