researchclaw
上朝式科研:AI-powered research workflow showcase
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/alphalab-ustc--researchclaw)HTML
<a href="/listings/alphalab-ustc--researchclaw"><img src="/api/listings/alphalab-ustc--researchclaw/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
📡 每日论文推荐
推荐今日论文
> 📡 今日论文推荐 | Daily Paper Scout > ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ > 🗓️ 2026-03-26 | 匹配兴趣: LLM · RL · Agentic AI > > 1️⃣ **RLVR: Reinforcement Learning with Verifiable Rewards** > 👤 Zhang et al. (2026) 🏷️ cs.LG · cs.AI > 💡 通过可验证奖励信号显著提升 LLM 数学推理能力 > 🎯 相关原因: 直接结合 RL + LLM reasoning,命中核心方向 > 🔗 https://arxiv.org/abs/2503.XXXXX > ...
---
📝 深度速读 + Deep Note 笔记
帮我读一下这篇:https://arxiv.org/abs/2503.19823
> 📝 Deep Note 完成 | AutoRefine > ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ > 📌 **AutoRefine: Search and Refine During Think** > 👤 Shi et al. | 📅 2025 | ⭐ 4/5 > 🔗 https://arxiv.org/abs/2503.19823 > > 💡 **核心发现:** 在推理链中插入知识精炼步骤比单纯增加检索更有效 > 📊 **关键数据:** 2WikiMHQA EM 67.7, HotpotQA EM 55.3 > ✨ **启发:** 精炼检索到的知识比检索本身更重要 > > 💾 笔记 → papers/2026-04-01_autorefine.md
想同时生成 HTML 页面?加 `--html`:
DNL https://arxiv.org/abs/2503.19823 --html
---
📋 管理论文列表
加入待读 https://arxiv.org/abs/2503.19823
> ✅ 已加入待读:**AutoRefine** | 共 7 篇待读
我的论文列表
> 📋 阅读列表已更新 → 打开 reading-list.html 查看看板 > To Read: 5 篇 · Reading: 2 篇 · Done: 12 篇
---
💡 灵感生成
给我一些研究灵感
> 💡 Research Idea 1: **Adaptive Memory Retrieval for RL Agents** > 📋 Problem: 当前 RL agent 的记忆检索策略是静态的,无法根据任务难度动态调整... > 🔧 Approach: 在 rollout 过程中,用 uncertainty estimation 决定何时检索记忆... > 📚 Related: AutoRefine (推理时检索) · MemOCR (记忆布局感知) > 🎯 Feasibility: 数据集现成,4×A100 × 3天,风险:检索延迟影响 rollout 速度
---
📄 论文写作
论文大纲:我想做一个 RL + Memory 的工作,针对 long-horizon agentic 任务
> 📄 论文大纲 | Paper Outline > 📌 Working Title: MemRL: Memory-Augmented Reinforcement Learning for Long-Horizon Agents > 🎯 Core Claim: 动态记忆检索机制显著提升 RL agent 在长程任务上的成功率 > 📍 Target Venue: NeurIPS 2026 > > ## 1. Introduction ... > ## 2. Related Work ... > (完整大纲展开)
rebuttal [Reviewer 2: The comparison with baseline X is unfair because...]
> 📝 Rebuttal Draft > **Reviewer 2** > > R2.1: "The comparison with baseline X is unfair..." > **Response:** We thank the reviewer for this concern. The comparison is fair because...
---
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
SKILL.md · Manifest · 58 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
- ---
- name: research-claw
- description: >
- Complete AI research assistant with 6 core capabilities: paper discovery, deep reading notes,
- reading list management, research taste learning, idea generation, and paper writing.
- Trigger words: 论文推荐, paper recommendation, arXiv, 读论文, paper notes, DNL,
- 推荐今日论文, 每日论文, 帮我读一下, paper scout, 论文速读, 帮我安装 ResearchClaw,
- research claws, researchclaw, 加入待读, 我的论文列表, reading list, 标记已读,
- 更新我的研究画像, 我的研究画像, research profile, 给我一些研究灵感, idea generator,
- 跨论文分析, 帮我写论文, paper writing, 论文大纲, 写引言, rebuttal.
- metadata:
- openclaw:
- homepage: https://github.com/AlphaLab-USTC/ResearchClaw
- version: "3.1.0"
- author: AlphaLab-USTC
- ---
- # ResearchClaw v3.1 — Modular Research Assistant
- ResearchClaw is a complete AI research companion with **6 atomic skills**.
- Each skill is self-contained and can be used independently.
- ## Skills
- | # | Skill | Directory | What it does |
- |---|---|---|---|
- | 1 | **Paper Scout** 📡 | `skills/paper-scout/` | Daily arXiv paper discovery with personalized ranking |
- | 2 | **Paper Reader** 📝 | `skills/paper-reader/` | Deep reading → markdown deep notes (default) or HTML page |
- | 3 | **Reading List** 📋 | `skills/reading-list/` | Kanban paper management + HTML dashboard |
- | 4 | **Research Profile** 🧠 | `skills/research-profile/` | Research taste profile + visualization |
- | 5 | **Idea Generator** 💡 | `skills/idea-generator/` | Cross-paper insights → research idea proposals |
- | 6 | **Paper Writer** 📄 | `skills/paper-writer/` | Outline → draft → auto-review → rebuttal |
- ## Quick Reference
- | Say this | Skill triggered |
- |---|---|
- | `推荐今日论文` | Paper Scout |
- | `帮我读一下 [arXiv link]` | Paper Reader |
- | `加入待读 [link]` | Reading List |
- | `我的论文列表` | Reading List |
- | `我的研究画像` | Research Profile |
- | `给我一些研究灵感` | Idea Generator |
- | `论文大纲 [idea]` | Paper Writer |
- | `rebuttal [comments]` | Paper Writer |
- ## Shared Config
- - **Research profile:** `~/.openclaw/workspace/research-claw-config.md`
- - **Reading list data:** `~/.openclaw/workspace/research-claw-reading-list.json`
- - **HTML output:** `~/.openclaw/workspace/research-claw-output/`
- - **Templates:** `templates/` (paper-note, reading-list, research-profile)
- ## Install
- ```
- 帮我安装 ResearchClaw:https://raw.githubusercontent.com/AlphaLab-USTC/ResearchClaw/main/docs/install.md
- ```
README.md · README · 388 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
- <div align="center">
- # 🦞 ResearchClaw
- **你的 AI 全能论文助手 | Your Complete AI Research Companion**
- *由 [OpenClaw](https://openclaw.ai) 驱动 · Powered by OpenClaw*
- [](https://openclaw.ai)
- [](https://github.com/AlphaLab-USTC/ResearchClaw)
- [](https://arxiv.org)
- [](LICENSE)
- [](https://alphalab-ustc.github.io/ResearchClaw/)
- <br>
- **6 大核心能力,零 API Key,开箱即用的 AI 科研全流程助手。**
- *6 core capabilities. No API keys. End-to-end AI research companion, out of the box.*
- <br>
- [🌐 在线演示](https://alphalab-ustc.github.io/ResearchClaw/) · [📖 安装](#一句话安装) · [⚙️ 配置](docs/config.md) · [🐛 Issues](https://github.com/AlphaLab-USTC/ResearchClaw/issues)
- </div>
- ---
- ## ✨ 6 大核心能力 | 6 Core Capabilities
- <table>
- <tr>
- <td width="33%" valign="top">
- ### 📡 论文雷达
- **Paper Scout**
- 每天从 arXiv 自动发现最新论文,按你的研究画像智能排序,推送 Top 5–10。
- - 🔍 覆盖 LLM/RL/多模态等方向
- - 🎯 个性化相关度评分
- - 📈 从你的反馈中自动学习
- - ⏰ 支持每日定时推送
- **触发:** `推荐今日论文`
- [▶ 查看演示](https://alphalab-ustc.github.io/ResearchClaw/showcase/case1-daily-signal-brief.html)
- </td>
- <td width="33%" valign="top">
- ### 📝 深度速读
- **Paper Reader**
- 发一个 arXiv 链接,AI 读完论文,基于 **Deep Note 7-section 框架** 生成结构化阅读笔记。
- - 📖 **Deep Note 框架**:Metadata → Why-read → CRGP → Figures → Experiments → Insights → Next steps
- - 📊 含图表 URL、具体实验数字、消融分析
- - 📝 **默认输出 Markdown**(git-friendly,低 token 消耗)
- - 🎨 可选生成 HTML 页面(`--html`)
- - 🇨🇳 中英双语混合
- **触发:** arXiv 链接 / `帮我读一下` / `DNL`
- [▶ 查看演示](https://alphalab-ustc.github.io/ResearchClaw/showcase/demo-paper-note.html)
- </td>
- <td width="33%" valign="top">
- ### 📋 论文列表
- **Reading List**
- 维护你的论文待读/正在读/已读列表,一键生成 HTML 看板。
- - 📊 三栏看板:To Read / Reading / Done
- - ➕ 一句话加入待读
- - ✅ 一句话标记已读
- - 🔗 每篇都链接到 HTML 笔记
- **触发:** `我的论文列表` / `加入待读`
- [▶ 查看演示](https://alphalab-ustc.github.io/ResearchClaw/showcase/demo-reading-list.html)
- </td>
- </tr>
- <tr>
- <td width="33%" valign="top">
- ### 🧠 研究画像
- **Research Taste Profile**
- 学习你的研究偏好,生成可视化研究画像页面,个性化一切推荐。
- - 📝 研究方向描述
- - 🌱 种子论文 & 关键词
- - 👤 关注作者白名单
- - 🔄 从对话反馈自动更新
- **触发:** `我的研究画像` / `更新我的研究画像`
- [▶ 查看演示](https://alphalab-ustc.github.io/ResearchClaw/showcase/demo-research-profile.html)
- </td>
- <td width="33%" valign="top">
- ### 💡 灵感生成器
- **Idea Generator**
- 分析你已读的论文,找出研究空白,生成 3–5 个有可行性的研究灵感。
- - 🔗 跨论文连接分析
- - 🕳️ 自动识别研究缺口
- - 📋 每个灵感含问题陈述 + 方法 + 相关论文 + 可行性评估
- - ⭐ 时机分析
- **触发:** `给我一些研究灵感`
- [▶ 查看演示](https://alphalab-ustc.github.io/ResearchClaw/showcase/case3-research-proposal.html)
- </td>
- <td width="33%" valign="top">
- ### 📄 论文写作
- **Paper Writer**
- 从灵感到草稿到回复审稿,覆盖论文写作全流程。
- - 🗂️ 结构化大纲生成
- - ✍️ 逐章节起草
- - 🔍 自我审稿 & 找弱点
- - 💬 Reviewer 回复起草
- **触发:** `论文大纲` / `写引言` / `rebuttal`
- [▶ 查看演示](https://alphalab-ustc.github.io/ResearchClaw/showcase/case4-multi-agent-codebase.html)
- </td>
- </tr>
- </table>
- ---
- ## 🚀 一句话安装
- 把这句话发给你的 OpenClaw 代理:
- ```
- 帮我安装 ResearchClaw:https://raw.githubusercontent.com/AlphaLab-USTC/ResearchClaw/main/docs/install.md
- ```
- > **就这一句话。** 代理会自动下载、注册技能,无需任何手动操作。
- **还没有 OpenClaw?** 先访问 [openclaw.ai](https://openclaw.ai) 安装,再运行上面这句话。
- ---
- ## 💬 快速上手示例 | Quick Start
- ### 📡 每日论文推荐
- ```
- 推荐今日论文
- ```
- > 📡 今日论文推荐 | Daily Paper Scout
- > ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
- > 🗓️ 2026-03-26 | 匹配兴趣: LLM · RL · Agentic AI
- >
- > 1️⃣ **RLVR: Reinforcement Learning with Verifiable Rewards**
- > 👤 Zhang et al. (2026) 🏷️ cs.LG · cs.AI
- > 💡 通过可验证奖励信号显著提升 LLM 数学推理能力
- > 🎯 相关原因: 直接结合 RL + LLM reasoning,命中核心方向
- > 🔗 https://arxiv.org/abs/2503.XXXXX
- > ...
- ---
- ### 📝 深度速读 + Deep Note 笔记
- ```
- 帮我读一下这篇:https://arxiv.org/abs/2503.19823
- ```
- > 📝 Deep Note 完成 | AutoRefine
- > ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
- > 📌 **AutoRefine: Search and Refine During Think**
- > 👤 Shi et al. | 📅 2025 | ⭐ 4/5
- > 🔗 https://arxiv.org/abs/2503.19823
- >
- > 💡 **核心发现:** 在推理链中插入知识精炼步骤比单纯增加检索更有效
- > 📊 **关键数据:** 2WikiMHQA EM 67.7, HotpotQA EM 55.3
- > ✨ **启发:** 精炼检索到的知识比检索本身更重要
- >
- > 💾 笔记 → papers/2026-04-01_autorefine.md
- 想同时生成 HTML 页面?加 `--html`:
- ```
- DNL https://arxiv.org/abs/2503.19823 --html
- ```
- ---
- ### 📋 管理论文列表
- ```
- 加入待读 https://arxiv.org/abs/2503.19823
- ```
- > ✅ 已加入待读:**AutoRefine** | 共 7 篇待读
- ```
- 我的论文列表
- ```
- > 📋 阅读列表已更新 → 打开 reading-list.html 查看看板
- > To Read: 5 篇 · Reading: 2 篇 · Done: 12 篇
- ---
- ### 💡 灵感生成
- ```
- 给我一些研究灵感
- ```
- > 💡 Research Idea 1: **Adaptive Memory Retrieval for RL Agents**
- > 📋 Problem: 当前 RL agent 的记忆检索策略是静态的,无法根据任务难度动态调整...
- > 🔧 Approach: 在 rollout 过程中,用 uncertainty estimation 决定何时检索记忆...
- > 📚 Related: AutoRefine (推理时检索) · MemOCR (记忆布局感知)
- > 🎯 Feasibility: 数据集现成,4×A100 × 3天,风险:检索延迟影响 rollout 速度
- ---
- ### 📄 论文写作
- ```
- 论文大纲:我想做一个 RL + Memory 的工作,针对 long-horizon agentic 任务
- ```
- > 📄 论文大纲 | Paper Outline
- > 📌 Working Title: MemRL: Memory-Augmented Reinforcement Learning for Long-Horizon Agents
- > 🎯 Core Claim: 动态记忆检索机制显著提升 RL agent 在长程任务上的成功率
- > 📍 Target Venue: NeurIPS 2026
- >
- > ## 1. Introduction ...
- > ## 2. Related Work ...
- > (完整大纲展开)
- ```
- rebuttal
- [Reviewer 2: The comparison with baseline X is unfair because...]
- ```
- > 📝 Rebuttal Draft
- > **Reviewer 2**
- > > R2.1: "The comparison with baseline X is unfair..."
- > **Response:** We thank the reviewer for this concern. The comparison is fair because...
- ---
- ## 🎬 在线演示 | Live Showcase
- 访问 [alphalab-ustc.github.io/ResearchClaw](https://alphalab-ustc.github.io/ResearchClaw/) 查看完整演示:
- | 案例 | 描述 |
- |------|------|
- | [每日信号简报](https://alphalab-ustc.github.io/ResearchClaw/showcase/case1-daily-signal-brief.html) | Paper Scout 多源聚合推送示例 |
- | [论文阅读笔记](https://alphalab-ustc.github.io/ResearchClaw/showcase/case2-paper-reading-notes.html) | Paper Reader HTML 笔记示例 |
- | [研究方向分析](https://alphalab-ustc.github.io/ResearchClaw/showcase/case3-research-proposal.html) | Idea Generator 灵感生成示例 |
- | [多智能体协作](https://alphalab-ustc.github.io/ResearchClaw/showcase/case4-multi-agent-codebase.html) | Paper Writer 写作辅助示例 |
- ---
- ## ⚙️ 配置研究画像
- 安装后,直接告诉代理:
- ```
- 更新我的研究画像:我主要研究强化学习和多模态 agent,关注 RLVR 方向
- ```
- 或手动编辑 `~/.openclaw/workspace/research-claw-config.md`:
- ```yaml
- research_direction: "PhD in LLM reasoning and agentic memory systems"
- seed_papers:
- - 2503.19823 # AutoRefine
- keywords:
- - reinforcement learning
- - agentic memory
- - long-context reasoning
- whitelist_authors:
- - Yaorui Shi
- - An Zhang
- ```
- 详见 [配置指南](docs/config.md)
- ---
- ## 📦 每日自动推送
- ```
- 帮我设置每天早上 9 点自动推荐论文
- ```
- 代理会自动创建 OpenClaw cron 任务。✅
- ---
- ## 🔧 工作原理 | How It Works
- ```
- 用户: 推荐今日论文
- ↓
- 读取 research-claw-config.md(研究画像)
- ↓
- 调用 arXiv 公开 API(免费,无需 Key)
- ↓
- 按个性化偏好评分排序
- ↓
- 返回 Top 5–10 论文速览
- ↓
- 用户反馈 → 自动更新研究画像
- ```
- ```
- 用户: 帮我读一下 arxiv.org/abs/XXXXX
- ↓
- 提取 arXiv ID → 获取元数据 + HTML 版本
- ↓
- 分析论文内容(Introduction, Method, Experiments)
- ↓
- 按 Deep Note 7-section 框架提取结构化信息
- ↓
- 输出 Markdown 笔记文件(默认)+ 聊天摘要
- ↓
- (可选)生成 HTML 页面
- ```
- **零第三方依赖**:只需 OpenClaw + 网络,不需要任何额外 API Key。
- ---
- ## 🤝 Credits
- - 🦞 Built with [OpenClaw](https://openclaw.ai) — AI assistant platform
- - 📄 Paper data from [arXiv](https://arxiv.org) (free open-access API)
- - 🧠 Created by [AlphaLab@USTC](https://github.com/AlphaLab-USTC) (Yaorui Shi et al.)
- ---
- ## 🌍 English Quick Reference
- **Install** (one line — send to your OpenClaw agent):
- ```
- 帮我安装 ResearchClaw:https://raw.githubusercontent.com/AlphaLab-USTC/ResearchClaw/main/docs/install.md
- ```
- **6 capabilities at a glance:**
- | Say this | What happens |
- |---|---|
- | `推荐今日论文` | Daily arXiv paper recommendations, ranked by your profile |
- | `帮我读一下 [arXiv link]` | Deep read → markdown deep note (+ optional HTML with `--html`) |
- | `加入待读 [link]` | Add paper to your reading list |
- | `我的论文列表` | View & regenerate your HTML reading dashboard |
- | `更新我的研究画像` | Update research taste profile + visual HTML page |
- | `给我一些研究灵感` | Cross-paper analysis → 3–5 research idea proposals |
- | `论文大纲 [idea]` | Generate structured paper outline |
- | `写引言` / `写方法` | Draft specific paper sections |
- | `rebuttal [comments]` | Draft point-by-point reviewer responses |
- **No API key needed.** Uses arXiv's free public API + OpenClaw's built-in tools.
- Zero third-party dependencies.
- ---
- ---
- ## 🔗 More from AlphaLab
- Check out our other projects:
- - 🦀 [**PaperClaw**](https://github.com/AkaliKong/PaperClaw) — Automated paper collection and analysis toolkit
- - 🛠️ [**OhMyCode**](https://github.com/AlphaLab-USTC/OhMyCode) — Code generation and repair benchmark
- - 📚 [**Must-Read-LLM-Papers**](https://github.com/AlphaLab-USTC/Must-Read-LLM-Papers) — Curated reading list for LLM research
- ---
- <div align="center">
- Made with ❤️ for researchers · MIT License · [alphalab-ustc.github.io/ResearchClaw](https://alphalab-ustc.github.io/ResearchClaw/)
- </div>
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.