← Back to Skills Marketplace
smseow001

Research Pro

by SMS · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
40
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install research-pro
Description
科研专项工作流 - 参考 AutoResearchClaw 23阶段全流程,从想法到论文。覆盖文献调研、实验设计、代码实现、统计分析、论文写作。
README (SKILL.md)

Research Pro - 科研专项工作流

参考 AutoResearchClaw 思路的完整科研流程


核心思想

像 AutoResearchClaw 一样做科研:

  1. 想法 → 文献调研
  2. 实验设计 → 沙盒执行
  3. 统计分析 → 同行评审
  4. LaTeX 论文输出

触发词

科研模式 / research pro / 做研究中 / 论文模式


完整工作流(23阶段)

Phase 1: 想法生成 🎯

## 🎯 想法定义

### 研究問題
[你想研究什么问题]

### 核心创新点
1. [创新点1]
2. [创新点2]

### 预期贡献
[发表后的贡献]

Phase 2: 文献调研 📚

## 📚 文献调研

### 相关工作
| 论文 | 方法 | 我们的区别 |
|------|------|-----------|
| [Paper A] | [方法A] | [改进点] |
| [Paper B] | [方法B] | [改进点] |

### 数据来源
- [x] OpenAlex
- [x] Semantic Scholar
- [x] arXiv

### 引用要求
- 真实引用(非幻觉)
- 近3年顶会优先

Phase 3: 实验设计 🧪

## 🧪 实验设计

### 对比基线
1. [基线方法1]
2. [基线方法2]

### 实验配置
- 数据集:[数据集名]
- 评估指标:[指标]
- 硬件:[GPU/CPU]

### 假设检验
- H0: [原假设]
- H1: [备择假设]

Phase 4: 代码实现 💻

(引用 code-pro skill)

## 💻 实现

### 核心代码
```python
# 核心实现

依赖

  • Python 3.x+
  • [库1]
  • [库2]

---

### Phase 5: 执行与调试 🔄

🔄 执行

环境检测

[GPU/CPU 检测结果]

运行结果

实验 结果 提升
基线A 0.85 -
我们的 0.89 +4.7%

统计分析

  • p-value: [值]
  • 显著性: [是/否]

---

### Phase 6: 论文写作 📄

📄 论文结构

Abstract

[150-300字摘要]

Introduction

  1. 问题定义
  2. 相关工作
  3. 我们的贡献

Method

[方法描述]

Experiments

[实验设置与结果]

Conclusion

[结论与未来工作]

格式

  • LaTeX (NeurIPS/ICML/ICLR)
  • 双栏排版

---

## 多代理协作

| 角色 | 职责 |
|------|------|
| CodeAgent | 代码生成与调试 |
| BenchmarkAgent | 实验评估 |
| FigureAgent | 图表生成 |
| Reviewer | 同行评审 |

---

## 质量保证

### Anti-Hallucination
- ✅ 引用必须从真实数据库获取
- ✅ 实验结果必须可复现
- ✅ claim 需验证

### Human-in-the-Loop
- 关键节点可人工介入
- checkpoint 审核
- 批准后继续

---

## 输出格式

### 论文模板

```markdown
# Title

## Abstract
[150-300 words]

## 1. Introduction
[1-2 paragraphs]

## 2. Related Work
[2-3 paragraphs]

## 3. Method
[2-3 paragraphs]

## 4. Experiments
[2-3 paragraphs]

## 5. Conclusion
[1 paragraph]

## References
[真实引用]

快速指令表

需求 命令
开始新研究 科研模式:研究[想法]
文献调研 调研[主题]相关文献
设计实验 设计实验:目标[指标]
写论文 生成论文:LaTeX格式
审核 同行评审这个研究

组合使用

完整流程示例:

科研模式:研究"用大模��做代码bug自动修复"
→ 帮我设计对比实验
→ 生成ICML格式论文

Research Pro | 参考 AutoResearchClaw 的全流程科研工具

Usage Guidance
This skill appears safe as an instruction-only research assistant. Before using it for experiments, confirm what code will run, what files or datasets it can access, and whether any referenced helper skill such as code-pro has separate permissions or install steps.
Capability Assessment
Purpose & Capability
The stated purpose—literature review, experiment design, code drafting, analysis, and paper writing—matches the SKILL.md workflow and no hidden capabilities are declared.
Instruction Scope
The workflow includes code implementation and experiment execution/debugging, but frames this as part of the research process and mentions sandbox execution and human checkpoints.
Install Mechanism
There is no install spec or bundled code. The document references using a separate code-pro skill, so behavior would depend on that other skill if invoked.
Credentials
No environment variables, credentials, binaries, or OS permissions are requested. Any environment checks or external literature lookups appear proportionate to the research workflow.
Persistence & Privilege
The artifacts do not request persistence, background workers, elevated privileges, credentials, session stores, or long-term memory.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install research-pro
  3. After installation, invoke the skill by name or use /research-pro
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of Research Pro skill, version 1.0.0. - Provides a comprehensive scientific research workflow inspired by AutoResearchClaw, covering 23 phases from idea to paper writing. - Supports literature review, experimental design, code implementation (with code-pro integration), statistical analysis, and LaTeX paper writing. - Features role-based collaboration (CodeAgent, BenchmarkAgent, FigureAgent, Reviewer) and anti-hallucination measures for research integrity. - Includes quick commands for starting research, literature survey, experiment design, paper writing, and peer review. - Ensures human-in-the-loop checkpoints and quality assurance throughout the workflow.
Metadata
Slug research-pro
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Research Pro?

科研专项工作流 - 参考 AutoResearchClaw 23阶段全流程,从想法到论文。覆盖文献调研、实验设计、代码实现、统计分析、论文写作。 It is an AI Agent Skill for Claude Code / OpenClaw, with 40 downloads so far.

How do I install Research Pro?

Run "/install research-pro" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Research Pro free?

Yes, Research Pro is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Research Pro support?

Research Pro is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Research Pro?

It is built and maintained by SMS (@smseow001); the current version is v1.0.0.

💬 Comments