← 返回 Skills 市场
biociao

Literature Daily Report

作者 Fang, Chao · GitHub ↗ · v1.1.0 · MIT-0
cross-platform ⚠ suspicious
81
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install biociao-literature-daily-report
功能描述
每日科研文献日报生成与管理。当用户请求生成科研日报、更新文献收集脚本或分析研究趋势时激活。支持:(1) 自动抓取 PubMed/bioRxiv/arXiv 最新文献,(2) 语义筛选 AI/生信/病原/真菌等领域,(3) LLM 智能总结与编辑排版,(4) 中文格式报告输出,(5) Zotero 自动录入,(6)...
使用说明 (SKILL.md)

每日科研文献日报

🎯 核心功能

自动化生成 生命科学×AI 交叉领域 的每日科研文献日报,包含:

  1. 📊 文献采集 - 实时抓取 PubMed, bioRxiv, arXiv 当天最新发表
  2. 🔍 智能筛选 - 语义检索 AI/生信/病原/真菌等关注领域
  3. 🤖 LLM 总结 - 自动生成 100-250 字核心摘要(中英文混排)
  4. 📰 编辑排版 - 前言 + 重点推荐 + 完整列表 + 趋势总结

🚀 快速开始

生成今日日报

# 方式 1:直接使用脚本
~/.openclaw/skills/literature-daily-report/scripts/literature_collector.py

# 方式 2:通过项目目录运行
cd ~/.openclaw/workspace/projects/literature-collector && ./run.sh

自定义配置

编辑配置文件 config.yaml

output_dir: ~/.openclaw/workspace/literature
search_queries:  # 搜索关键词组合
  - "metagenomics AND machine learning"
  - "fungal pathogen AND bioinformatics"
  - "single-cell AND deep learning"
high_impact_journals:
  - Nature
  - Cell
  - Science
  - Bioinformatics

🔧 工作流程

Step 1: 文献采集 (fetch_articles_with_abstracts)

从三大数据库获取最新文献及摘要:

来源 API 查询范围 特点
PubMed EUtils 最近 1 天 已发表论文,有 PMID
bioRxiv REST API 最近 1 天 预印本,DOI 格式
arXiv Export API 最近 1 天 CS/Q-Bio 类别

执行逻辑:

for query in SEARCH_QUERIES:
    articles = fetch_pubmed(query)      # PubMed
    biorxiv = fetch_biorxiv()           # bioRxiv  
    arxiv = fetch_arxiv()               # arXiv
    all_articles.extend(...)

Step 2: 语义筛选 (categorize_article)

根据关键词匹配筛选目标领域:

CATEGORY_KEYWORDS = {
    "单细胞组学": ["single-cell", "scRNA-seq", "spatial transcriptomics"],
    "宏基因组学": ["metagenomics", "microbiome", "16S"],
    "病原真菌": ["fungal", "pathogen", "Candida", "Aspergillus"],
    "生信方法": ["bioinformatics", "algorithm", "pipeline", "tool"],
    "AI/ML": ["machine learning", "deep learning", "transformer"],
    "基因组学": ["genomics", "genome", "pan-genome"],
}

优先级评分:

  • 高影响力期刊(Nature/Cell/Science)+10 分
  • 方法开发类文章 +5 分
  • 热门话题(单细胞/AI)+3 分

Step 3: LLM 总结 (generate_summary)

基于摘要生成结构化中文总结:

【研究目的】
【样本与方法】
【研究结果】
【创新性】

优化要点:

  • 识别英文摘要 → 添加 【】 标签包裹各部分
  • 长度控制在 5000 字以内
  • 提取关键句子而非简单截断

Step 4: 编辑排版 (generate_mark_report_v2)

生成完整的 Markdown 日报结构:

# 📚 每日文献速递 - YYYY-MM-DD

## 📰 编辑前言
- 日期统计
- 来源分布
- 热点领域概览

## ⭐ 重点推荐 (8 篇)
- 带标签分类
- 结构化摘要
- DOI 链接

## 📖 完整文献列表
- 按来源分组
- 详细元数据

## 📝 编辑总结
- 今日趋势分析
- 编者点评
- 关注建议

📂 文件组织

literature-daily-report/
├── SKILL.md                    # 本说明文档
├── scripts/
│   └── literature_collector.py # 主收集脚本
└── references/
    ├── categories.md           # 领域分类标准
    ├── workflows.md            # 工作流指南
    └── api_docs.md             # API 参考文档

# 输出目录
~/.openclaw/workspace/literature/
├── literature-YYYY-MM-DD.md   # 当日报告
└── latest.md                   # 最新报告索引

# ClawLib 同步目录
~/.openclaw/workspace/ClawLib/科研日报/
├── literature-YYYY-MM-DD.md   # 当日报告(自动同步)
└── latest.md                   # 最新报告(自动同步)

🔄 自动同步(v3.3 全套链路)

生成日报后会自动同步到多个目标,一气呵成:

  1. 📝 本地报告~/.openclaw/workspace/literature/literature-{date}.md
  2. 📚 ClawLib~/.openclaw/workspace/ClawLib/科研日报/
  3. 🔍 Zotero → BioCiaoLab Group Library(按分类收藏夹)
  4. 🧠 知识图谱research/literature-{date} 实体

📚 Zotero 集成 (v3.3)

文献收集完成后会自动录入 BioCiaoLab Group Library

功能特点:

  • PubMed 文献:通过 PMID 自动添加
  • bioRxiv/arXiv 预印本:通过 DOI 自动添加
  • 自动去重(已存在的文献会跳过)
  • 添加标签:literature-daily + 日期 + 分类(如 单细胞组学, 宏基因组学
  • 按文献归属类别添加到对应的收藏夹(单细胞组学/宏基因组学/病原真菌/AI+ML/生信方法/基因组学等)

环境变量配置(写入 ~/.zshrc):

export ZOTERO_API_KEY="your-api-key"
export ZOTERO_USER_ID="your-user-id"       # Zotero 用户 ID
export ZOTERO_GROUP_ID="your-group-id"     # BioCiaoLab Group ID (默认 6489333)

⚠️ 脚本会自动从 ~/.zshrc 加载环境变量,无需手动 export。

获取方式:

  1. API Key: https://www.zotero.org/settings/keys/new
  2. User ID: https://www.zotero.org/settings
  3. Group ID: 访问 BioCiaoLab 小组页面,URL 中的数字部分即为 Group ID

🧠 知识图谱同步 (v3.3)

文献收集完成后会自动同步到 OpenClaw 知识图谱

功能特点:

  • 每篇文献作为一条 fact 写入 research/literature-YYYY-MM-DD 实体
  • fact 包含:分类标签、标题、作者、期刊、来源、PMID/URL
  • 调用 kg.py add 逐条写入知识图谱

写入位置:

  • entity: research/literature-\x3C日期>
  • category: 文献主分类(单细胞组学/宏基因组学/病原真菌/AI+ML/生信方法/基因组学等)
  • source: literature-daily-report

🌟 高级用法

自定义搜索领域

SEARCH_QUERIES 中添加新主题:

SEARCH_QUERIES = [
    '(epigenetics[Title/Abstract]) AND ((deep learning[Title/Abstract]))',
    '(CAR-T[Title/Abstract]) AND ((single-cell[Title/Abstract]))',
]

调整输出格式

修改 generate_mark_report_v2() 中的标题层级和标签风格。

定时任务

添加 crontab 自动发送日报:

# 每天早上 9 点生成并发送到飞书
0 1 * * * cd ~/.openclaw/workspace/projects/literature-collector && ./run.sh

🔄 迭代历史

  • v1.0: 基础采集(标题 + 来源)
  • v2.0: 增加摘要获取 + 智能总结
  • v3.0: 中文格式优化 + 编辑结构化排版
  • v3.1: 自动同步到 ClawLib 科研日报目录
  • v3.2: 新增 Zotero 自动录入功能(PMID/DOI 自动添加)
  • v3.3: 新增知识图谱同步;Zotero 环境变量自动从 ~/.zshrc 加载

📞 技术支持

遇到问题?检查:

  1. API 限流 → 增加 time.sleep() 延时
  2. 无结果 → 扩展搜索关键词范围
  3. 格式异常 → 查看 literature/abstract_cache.json 缓存状态
安全使用建议
Before installing or running this skill: 1) Treat it as requiring Zotero credentials even though the metadata omits them — do not place secrets in ~/.zshrc unless you intend to share them. Prefer setting ZOTERO_API_KEY in a restricted environment and verify how the script reads it. 2) Inspect the full literature_collector.py (and any zotero.py / kg.py it calls) for subprocess calls and network endpoints — those helper scripts (paths under ~/.openclaw/workspace/skills/...) could upload data to external services. 3) Verify the default ZOTERO_GROUP_ID (6489333) — if the script uploads to that group by default, you may be sending items to a third‑party collection. 4) If you want to proceed, run the script in a sandbox or isolated environment first, with minimal credentials and with network access limited, and monitor exactly what network requests it makes. 5) Ask the publisher/owner for provenance (homepage, source repo) or request the skill declare its required env vars and endpoints in metadata; lack of provenance combined with automatic uploading behavior is the primary reason this is suspicious.
功能分析
Type: OpenClaw Skill Name: biociao-literature-daily-report Version: 1.1.0 The skill bundle provides a functional tool for scientific literature management but includes high-risk behaviors in `scripts/literature_collector.py`. Specifically, the script executes a shell command (`bash -l -c "env"`) to scrape the user's entire environment for credentials and uses `subprocess.run` to execute external scripts (`zotero.py`, `kg.py`). While these actions are plausibly needed for the stated purpose of syncing data to Zotero and a Knowledge Graph, they represent a broad attack surface for credential harvesting and arbitrary code execution. The script also makes numerous network calls to scientific APIs (PubMed, bioRxiv, arXiv) and interacts with a Zotero group (ID: 6489333).
能力评估
Purpose & Capability
The skill claims no required environment variables or binaries in the metadata, but the SKILL.md asks users to place Zotero credentials in ~/.zshrc and the included script references Zotero and knowledge‑graph helper scripts. Requesting Zotero API keys is consistent with the stated Zotero integration, but the metadata omission (no declared env vars) is an incoherence: the skill actually needs credentials but does not declare them.
Instruction Scope
SKILL.md instructs the agent/script to automatically load environment variables from ~/.zshrc, create reports under ~/.openclaw/workspace, sync to a ClawLib directory, push to a Zotero group (BioCiaoLab default group id), and call a knowledge‑graph CLI (kg.py add). Those actions go beyond simple fetching/summarization because they perform writes and network uploads and rely on other scripts (zotero.py, kg.py) at paths under the workspace. The instructions also tell users to write secrets into ~/.zshrc and promise automatic loading — Python code doesn't automatically source shell rc files, so this is inconsistent and could lead to surprising credential access or accidental exfiltration.
Install Mechanism
This is instruction‑only with a bundled Python script; there is no external install/download step or remote archive. That lowers supply‑chain installation risk. However, the script invokes other local helper scripts (zotero.py, kg.py) via fixed paths which could execute arbitrary code if those helper scripts are present or installed later.
Credentials
The SKILL.md requires sensitive environment variables (ZOTERO_API_KEY, ZOTERO_USER_ID, ZOTERO_GROUP_ID) but the skill metadata did not declare any required env vars or primary credential. The code defaults ZOTERO_GROUP_ID to a specific group id (BioCiaoLab) which means, if the code invokes a Zotero upload helper, collected metadata could be added to that group by default. The script also reads os.environ and may call other scripts that themselves use credentials — overall the credential handling is under‑declared and therefore disproportionate to the metadata.
Persistence & Privilege
The skill is not always:true and does not claim elevated platform privileges. It writes files into the user's workspace and ClawLib directories and calls local helper scripts; these behaviors are within the expected scope for a sync/reporting tool but do require the skill to be allowed file and network access. There's no explicit modification of other skills' configs, but dependence on workspace helper scripts increases the attack surface (those scripts could be swapped to exfiltrate data).
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install biociao-literature-daily-report
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /biociao-literature-daily-report 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.1.0
v3.3: 新增知识图谱同步;Zotero 环境变量自动从 ~/.zshrc 加载;全套同步链路(ClawLib + Zotero + 知识图谱)
元数据
Slug biociao-literature-daily-report
版本 1.1.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Literature Daily Report 是什么?

每日科研文献日报生成与管理。当用户请求生成科研日报、更新文献收集脚本或分析研究趋势时激活。支持:(1) 自动抓取 PubMed/bioRxiv/arXiv 最新文献,(2) 语义筛选 AI/生信/病原/真菌等领域,(3) LLM 智能总结与编辑排版,(4) 中文格式报告输出,(5) Zotero 自动录入,(6)... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 81 次。

如何安装 Literature Daily Report?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install biociao-literature-daily-report」即可一键安装,无需额外配置。

Literature Daily Report 是免费的吗?

是的,Literature Daily Report 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Literature Daily Report 支持哪些平台?

Literature Daily Report 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Literature Daily Report?

由 Fang, Chao(@biociao)开发并维护,当前版本 v1.1.0。

💬 留言讨论