← All Skills
A Markdown-only skill pack that lets an agent do machine-learning research "in its sleep": from idea discovery through running experiments to writing the paper, with cross-model review loops keeping each step honest. No framework, no lock-in — runs on Claude Code, Codex, Cursor and more.
Key Features
- Markdown-only skill pack
- Idea-to-paper research flow
- Cross-model anti-fabrication review
- Runs on many agents
Installation
git clone https://github.com/wanshuiyin/Auto-claude-code-research-in-sleep
cd your-project
bash Auto-claude-code-research-in-sleep/tools/install_aris.sh
(Windows: install_aris.ps1; add --groups / --skills for selective install.)
How to Use
Call the slash-command pipelines in your agent, e.g. /research-pipeline "a research topic", or /idea-discovery, /run-experiment, /paper-writing, /proof-checker, /research-wiki. Control intensity with lite|balanced|max|beast.
Requirements
Claude Code (or Codex CLI / Cursor, etc.), Git, Python 3.9+, and at least one API key (Anthropic, or OpenAI/Codex, Kimi, DeepSeek, GLM, …).
Related Skills