ARIS
Markdown-based skill files that let an AI agent run ML research loops, with a second model reviewing the work.wanshuiyin/Auto-claude-code-research-in-sleep
- Stars
- 18.5k
- Forks
- 1.5k
- Licence
- MIT
- Language
- Python
- Last update
- 5 days ago
- Type
- AI model / agent
What it is
ARIS is a set of plain Markdown skill files for AI coding agents such as Claude Code or Codex. The skills guide an agent through autonomous machine learning research tasks like idea discovery, experiment automation, and review loops, where an independent model checks the executor's output. The README describes it as a methodology rather than a platform, so it does not add a separate framework.
- Run a cross-model review loop where an executor agent drafts work and a reviewer model critiques it
- Use idea discovery skills to generate and screen research directions
- Automate experiment steps through the bundled skills
- Use the optional ARIS-Code CLI as a standalone install instead of the skill setup
Great for
- ML researchers who already use Claude Code or Codex and want a structured research workflow
- People who want to pair two different models so one reviews the other's work
- Developers who prefer plain Markdown configuration over installing a framework
- Teams experimenting with agent-driven paper review and experiment loops
Not ideal for
- People without access to Claude Code, Codex, or another compatible agent, since the skills run inside those tools
- Users who need a fully hands-off system with guaranteed research results, because the README describes an agent-driven workflow that still requires oversight
- Readers looking for a general-purpose app with a graphical interface; the README centers on skills and an optional CLI
How to get it
- 1
See the README
You'll need: Claude Code or Codex CLI, per the README's supported agents
Clone the repository
git clone https://github.com/wanshuiyin/Auto-claude-code-research-in-sleep - 2
Run it
- Clone the repository and follow the Quick Start section of the README
- Register the codex MCP reviewer if you already have ARIS installed (Quick Start step 2b)
- Start your agent and invoke an ARIS skill as described in the README
- 3
Get the most out of it
- Read AGENT_GUIDE.md for agent-oriented instructions linked from the README
- Pick an alternative model combination if you do not want to use Codex MCP as the default reviewer
- Try the optional ARIS-Code standalone CLI from the release downloads if you want one-command setup
Downloads
Strengths
- Markdown-only skills with no framework to learn, per the description
- Works across several agents, including Claude Code, Codex, Cursor, and OpenClaw, according to the README
- MIT licensed and actively maintained, with a recent push and frequent releases
Trade-offs
- The README notes a recent codex-cli change that broke the reviewer bridge, so setup may need a re-registration step
- Much of the documentation is in Chinese and English mixed, and the README is long and truncated here, so details are hard to verify quickly
- Results depend on the underlying models and API access, which the README does not fully specify
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| Auto-claude-code-research-in-sleep (this) | 18.5k | AI model / agent | intermediate | 5 days ago | ML researchers who already use Claude Code or Codex and want a structured research workflow |
| AI-Research-SKILLs | 14.2k | Curated list | intermediate | 26 days ago | Developers who already use Claude Code, Codex, or Gemini CLI and want their agent to handle ML engineering tasks |
| cc-switch | 151k | App | beginner | 4 days ago | Developers who use several AI coding agents and switch between API providers often |
| hermes-agent | 260k | AI model / agent | intermediate | 4 days ago | Developers who want a personal agent they can host on their own server or machine |
| humanizer | 63.8k | AI model / agent | beginner | 9 days ago | Writers who draft with an AI and want the final text to sound more natural |
| ECC | 295k | Dev tool | intermediate | 4 days ago | Developers who already use Claude Code and want a shared workflow for planning, testing and review |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.