Mapping-Skill
AI/ML 人才搜索、论文作者发现、实验室成员爬取、GitHub 研究者挖掘与个性化招聘邮件生成 skill。只要用户提到查找 AI/ML PhD、研究员、工程师,抓取实验室成员、OpenReview/CVF 会议作者、GitHub 网络研究者,提取主页/Scholar/GitHub/邮箱/研究方向,识别华人、分类去重,或把结果导入飞书多维表格并批量生成邮件,就应该优先使用这个 skill;即使用户没有明确说“使用 Mapping-Skill”,只要任务属于这些复合工作流,也应触发。
About this skill
The Mapping-Skill is designed for AI agents like Claude Code and OpenClaw to automate complex workflows related to AI/ML talent acquisition and research discovery. It provides comprehensive capabilities to search for AI/ML PhDs, researchers, and engineers across multiple platforms. This includes scraping lab member homepages, extracting author information from OpenReview and CVF conferences, and discovering researchers through GitHub following/followers networks. Beyond just data collection, the skill intelligently processes the extracted information by identifying candidates of specific backgrounds (e.g., Chinese candidates), categorizing, deduplicating, and standardizing research directions. A core feature is its ability to generate highly personalized recruitment emails based on a candidate's specific research, publications, or professional profile. This ensures outreach is relevant and impactful. Furthermore, the Mapping-Skill offers robust integration with Feishu (Lark Suite) to import scraped results into multi-dimensional tables. This facilitates advanced management, tracking, and the batch generation and back-writing of personalized emails directly within the spreadsheet environment. This comprehensive approach transforms fragmented talent search efforts into a structured, efficient, and integrated pipeline for recruitment and academic outreach.
Best use case
This skill is primarily for HR professionals, technical recruiters, and research managers seeking to identify and engage top AI/ML talent efficiently. It is also highly valuable for academic researchers looking to map out expert landscapes, discover collaborators, or track influential contributions in specific AI/ML domains. By automating repetitive data extraction, cleaning, and communication tasks, it significantly reduces manual effort and accelerates the process of building targeted outreach campaigns.
AI/ML 人才搜索、论文作者发现、实验室成员爬取、GitHub 研究者挖掘与个性化招聘邮件生成 skill。只要用户提到查找 AI/ML PhD、研究员、工程师,抓取实验室成员、OpenReview/CVF 会议作者、GitHub 网络研究者,提取主页/Scholar/GitHub/邮箱/研究方向,识别华人、分类去重,或把结果导入飞书多维表格并批量生成邮件,就应该优先使用这个 skill;即使用户没有明确说“使用 Mapping-Skill”,只要任务属于这些复合工作流,也应触发。
A structured list of AI/ML candidates with detailed profiles, including research interests, affiliations, contact information, and potentially personalized recruitment email drafts, often integrated into a Feishu spreadsheet.
Practical example
Example input
Find top AI/ML PhD candidates in natural language processing from major universities in the US, extract their papers, email, and LinkedIn profiles, and draft recruitment emails highlighting our latest research on large language models.
Example output
Generated CSV file: 'llm_phd_candidates.csv' containing Name, University, Research Interests, Papers, Email, LinkedIn. Drafted 15 personalized emails stored in 'draft_emails/' directory, ready for review. Also updated Feishu table with candidate data.
When to use this skill
- When needing to find AI/ML PhDs, researchers, or engineers based on specific criteria.
- To scrape structured information from academic homepages, OpenReview, CVF, or GitHub profiles.
- When you need to generate personalized recruitment emails linked to a candidate's research.
- For importing candidate data and managing email outreach within Feishu (Lark Suite) spreadsheets.
When not to use this skill
- For general web scraping tasks unrelated to AI/ML talent or academic research.
- When searching for professionals in fields outside of Artificial Intelligence or Machine Learning.
- If you only require simple data extraction without further processing, categorization, or email generation.
- For tasks that do not involve external data sources or communication, such as internal data analysis.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/mapping-skill/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How Mapping-Skill Compares
| Feature / Agent | Mapping-Skill | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | easy | N/A |
Frequently Asked Questions
What does this skill do?
AI/ML 人才搜索、论文作者发现、实验室成员爬取、GitHub 研究者挖掘与个性化招聘邮件生成 skill。只要用户提到查找 AI/ML PhD、研究员、工程师,抓取实验室成员、OpenReview/CVF 会议作者、GitHub 网络研究者,提取主页/Scholar/GitHub/邮箱/研究方向,识别华人、分类去重,或把结果导入飞书多维表格并批量生成邮件,就应该优先使用这个 skill;即使用户没有明确说“使用 Mapping-Skill”,只要任务属于这些复合工作流,也应触发。
How difficult is it to install?
The installation complexity is rated as easy. You can find the installation instructions above.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
Related Guides
Best AI Skills for Claude
Explore the best AI skills for Claude and Claude Code across coding, research, workflow automation, documentation, and agent operations.
ChatGPT vs Claude for Agent Skills
Compare ChatGPT and Claude for AI agent skills across coding, writing, research, and reusable workflow execution.
AI Agents for Coding
Browse AI agent skills for coding, debugging, testing, refactoring, code review, and developer workflows across Claude, Cursor, and Codex.
SKILL.md Source
# Mapping-Skill
面向 Claude Code 与 OpenClaw 的 AI/ML 人才搜索与触达执行手册。
## 项目链接
- GitHub: https://github.com/16Miku/Mapping-Skill
- ClawHub: https://clawhub.ai/16Miku/mapping-skill
## 这个 skill 具备的能力
当用户提出以下任务时,应优先启用本 skill:
1. 搜索 AI/ML PhD、研究员、工程师
2. 抓取实验室成员主页并提取结构化信息
3. 抓取 OpenReview / CVF 论文作者信息
4. 从 GitHub following/followers 网络中发现研究者
5. 对给定 URL 执行全量学者信息抽取
6. 识别华人候选人、分类、去重、标准化研究方向
7. 基于候选人信息生成个性化招聘邮件
8. 将 CSV 或爬取结果导入飞书多维表格
9. 在飞书表格中批量生成并回写推荐邮件
## 执行原则
1. **先判定任务类型,再选方法**
- Topic search
- Lab search
- Conference author search
- GitHub network discovery
- Given-URL extraction
- Feishu email workflow
2. **优先复用已有 references 与 scripts**
不要从零发明流程。先检查 `references/` 与 `scripts/` 是否已有成熟模式。
3. **优先选择最稳定的数据入口**
- OpenReview 会议优先 API
- CVF 会议优先 HTML + PDF
- Hugo Academic 页面优先模板化解析
- 页面结构混乱但邮箱明显时,使用邮箱反向定位法
4. **抽取与清洗并重**
结果必须尽量结构化,并在输出前做分类、去重与字段标准化。
5. **邮件必须基于真实信息个性化**
`technical_hook` 和 `talk_track_paragraph` 不能空泛,必须和候选人论文、研究方向或主页内容关联。
6. **如用户涉及 OpenClaw / 飞书场景,要显式考虑导表和字段回写**
对此类需求,结果不应只停留在本地 CSV。
## 平台使用说明
### Claude Code
适用于:
- 本地脚本执行
- MCP 工具调用
- CSV 导出
- 以本地文件和结构化结果为主要交付物
安装方式通常是把 skill 放到 `~/.claude/skills/` 目录。
### OpenClaw
适用于:
- skill 目录托管与刷新
- slash commands 调用
- ClawHub 分发
- 飞书、多工具联动工作流
OpenClaw 常见技能加载位置:
- `<workspace>/skills`
- `~/.openclaw/skills`
- 内置 skills
也可通过 ClawHub 安装,并通过“刷新 skills”或重启网关重新索引。
## 工作流
### Step 1:识别输入任务与目标输出
先明确:
- 搜索范围是 topic、lab、conference、GitHub network 还是给定 URL
- 输出要求是候选人列表、CSV、飞书入表、邮件生成,还是全流程都要
- 是否需要识别华人、是否需要邮箱、是否需要导入飞书
### Step 2:选择数据源与抓取方式
#### 方法选择矩阵
| 场景 | 首选方案 | 备用方案 |
|------|----------|----------|
| OpenReview 会议 | `scripts/openreview_scraper.py` + API | 搜索 + 主页回补 |
| CVF 会议 | `scripts/cvf_paper_scraper.py` | 补抓 PDF / 页面回退 |
| Hugo Academic 单页卡片 | `lab_member_scraper.py` 的 card 模式 | BrightData |
| 实验室列表页 + 个人页 | `lab_member_scraper.py` 的两阶段模式 | BrightData |
| 无固定结构但含邮箱 | 邮箱反向定位法 | BrightData / 手工规则 |
| GitHub 研究者网络 | `scripts/github_network_scraper.py` | 网页搜索辅助 |
| LinkedIn / 强反爬站点 | BrightData MCP | 降级到公开网页信息 |
| 给定任意 URL | BrightData MCP 或定制脚本 | 多源补充 |
### Step 3:执行抽取
根据场景读取相应脚本或 reference:
- 搜索模板:`references/search-templates.md`
- Python 爬取:`references/python-scraping-guide.md`
- 反爬处理:`references/anti-scraping-solutions.md`
- URL 优先级:`references/url-priority-rules.md`
- 会议抓取:`references/conference-paper-scraping.md`
### Step 4:结构化与标准化
至少尽量抽取这些字段:
- 中文名 / 英文名
- title / role
- affiliation
- research_interests / research_field
- education / experience
- publications
- homepage / Google Scholar / GitHub / LinkedIn / Zhihu / Bilibili
- email
然后继续做:
- 华人识别:`references/chinese-surnames.md`
- 候选人分类:`references/candidate-classifier.md`
- 去重:`references/deduplication-rules.md`
- 研究方向标准化:`references/field-mappings.md`
### Step 5:邮件生成
读取:
- `references/email-templates.md`
- `references/talk-tracks.md`
生成邮件时必须填充:
- `researcher_name`
- `context_affiliation`
- `research_field`
- `technical_hook`
- `talk_track_paragraph`
### Step 6:结果交付
根据用户要求输出为:
- Markdown 结构化表格
- CSV 文件
- 飞书多维表格
- 飞书表格中的“推荐邮件”字段
## OpenClaw / 飞书工作流要点
如果用户明确提到 OpenClaw、飞书、多维表格、导表或批量写邮件,应把这些步骤视为本 skill 的标准能力,而不是额外加分项。
### 标准能力
1. 解析飞书多维表格链接
2. 提取 `app_token` / `table_id`
3. 批量读取记录
4. 依据论文标题或候选人关键词映射研究方向
5. 批量生成个性化邮件
6. 创建新字段并批量更新记录
7. 返回飞书链接、成功/失败统计与原因
## 最佳实践提示词
后续如果有新的实践文档,应继续沉淀到 `references/prompt-best-practices.md`。当前优先复用下面几类高价值提示词模式。
### 1. OpenReview 论文爬取 + 飞书入表
```text
请执行 OpenReview 论文爬取任务:
1. 使用 Mapping-Skill skill 根目录下的 `scripts/openreview_scraper.py` 脚本
2. 初始化爬虫时使用 api2.openreview.net 端点:
scraper = OpenReviewScraper(
username='XXXXXXX',
password='XXXXXXX',
baseurl='https://api2.openreview.net'
)
3. 爬取 ICLR2025 的 5 篇论文(测试)+ https://openreview.net/group?id=ICLR.cc/2025/Conference#tab-accept-oral(记着替换链接)
4. 保存 CSV 到 /tmp/ 目录
5. 创建新的飞书多维表格,按照 Mapping-Skill skill 根目录下的 `scripts/openreview_scraper.py` 脚本中爬取的数据来创建相应字段
6. 批量导入数据到多维表格
7. 返回多维表格链接和统计信息
```
### 2. CVF 论文爬取 + 邮箱提取 + 飞书入表
```text
请执行 CVF 论文爬取任务:
1. 使用 Mapping-Skill skill 根目录下的 `scripts/cvf_paper_scraper.py` 脚本
2. 严格按照脚本中的 extract_emails_from_text() 函数提取邮箱
3. 爬取 ICCV2025 的 5 篇论文(测试)+ https://openaccess.thecvf.com/ICCV2025?day=all(记着替换链接)
4. 保存 CSV 到 /tmp/ 目录
5. 创建新的飞书多维表格,按照 Mapping-Skill skill 根目录下的 `scripts/cvf_paper_scraper.py` 脚本中爬取的数据来创建相应字段
6. 批量导入数据到多维表格
7. 返回多维表格链接和邮箱提取统计
```
### 3. 飞书表格批量写邮件
```text
请执行论文作者邮件生成任务:
【数据源】
表格链接:
【第一步:解析表格链接】
1. 从链接中提取 app_token(格式:/base/{app_token})
2. 调用 feishu_bitable_app_table 的 list 接口获取 table_id
3. 验证表格可访问性
【第二步:分批读取论文数据】
1. 使用 feishu_bitable_app_table_record 的 list 操作
2. 分批读取(每批50条),使用 page_token 分页
3. 只提取必要字段:记录ID、论文标题、作者、邮箱、机构
4. 过滤条件:只处理有邮箱的记录
【第三步:确定研究领域】
1. 读取 Mapping-Skill skill 根目录下的 `references/field-mappings.md`
2. 根据论文标题和关键词,使用映射规则确定研究领域
3. 示例:
- "Symmetry Understanding of 3D Shapes" → Computer Vision
- "Efficient Adaptation of Vision Transformer" → NLP
【第四步:生成个性化邮件】
1. 读取 Mapping-Skill skill 根目录下的 `references/email-templates.md`
2. 根据研究领域选择对应模板(共22个领域)
3. 填充占位符:
- {{researcher_name}} → 第一作者姓名
- {{context_affiliation}} → 机构
- {{research_field}} → 研究领域
- {{technical_hook}} → 基于论文标题生成
- {{talk_track_paragraph}} → 从 talk-tracks.md 选择
【第五步:批量更新多维表格】
1. 在多维表格中创建新字段:"推荐邮件"(多行文本)
2. 使用 batch_update 批量更新每条记录
3. 每批最多 500 条
【第六步:验证和统计】
1. 验证邮件内容个性化
2. 返回统计:总计 X 条 / 成功 Y 条 / 失败 Z 条
3. 列出失败原因
【输出】
- 多维表格链接
- 生成统计
- 失败原因列表
```
### 4. 给定某 URL 全量抽取并入表
```text
1、请你调用BrightData-MCP工具,或者编写爬虫脚本,爬取 <某网站URL> 页面中的所有人员信息。
2、提取信息包括中文名,英文名,个人介绍信息、学术方向、学校和专业信息、工作经历、近期论文著作信息(包含论文名和论文链接)、github链接、个人主页链接、谷歌学术链接、领英链接、知乎链接、B站链接、邮箱等。
3、当前页面缺少邮箱的话,需要进入学者主页或论文链接页面,从里面提取作者们的邮箱。
4、保存到csv文件,然后将csv导入飞书多维表格。
```
## 输出格式
### 候选人摘要表
| Name | Type | Affiliation | Field | Chinese? | Email |
|------|------|-------------|-------|----------|-------|
| Wei Zhang | PhD | Tsinghua | RL | Yes (0.92) | wei@tsinghua.edu |
### 单个候选人详细输出
```markdown
## Candidate: Wei Zhang (张伟)
- Type: PhD Student
- Affiliation: Tsinghua University
- Research Field: Reinforcement Learning
- Chinese: Yes (0.92)
- Email: wei.zhang@tsinghua.edu.cn
- Homepage: ...
- Scholar: ...
- GitHub: ...
### Research Summary
- RLHF
- Reward modeling
- Policy optimization
### Publications
1. ...
2. ...
### Outreach Email
...
```
### 飞书工作流输出
至少返回:
- 飞书多维表格链接
- 总记录数 / 成功数 / 失败数
- 邮箱提取统计(如有)
- 失败原因摘要
## References guide
按场景加载:
- 通用搜索:`references/search-templates.md`
- 候选人字段:`references/profile-schema.md`
- 分类:`references/candidate-classifier.md`
- 华人识别:`references/chinese-surnames.md`
- 去重:`references/deduplication-rules.md`
- 邮件模板:`references/email-templates.md`
- talk tracks:`references/talk-tracks.md`
- Python 抓取:`references/python-scraping-guide.md`
- 反爬处理:`references/anti-scraping-solutions.md`
- 会议抓取:`references/conference-paper-scraping.md`
- 后续实践复盘:`references/practice-cases.md`
- 后续最佳提示词:`references/prompt-best-practices.md`
- 后续用户反馈:`references/user-feedback-notes.md`
## Scripts guide
- `scripts/openreview_scraper.py`:OpenReview 会议论文与作者抓取
- `scripts/cvf_paper_scraper.py`:CVF 论文页面 + PDF 邮箱提取
- `scripts/lab_member_scraper.py`:实验室成员抓取(两阶段 / Hugo Academic / 邮箱反向定位)
- `scripts/github_network_scraper.py`:GitHub 研究者网络抽取
- `scripts/cloudflare_email_decoder.py`:Cloudflare XOR 邮箱解密
- `scripts/httpx_scraper.py`:通用异步 HTTP 抓取
- `scripts/serper_search.py`:搜索入口模板
## 不要遗漏的经验
1. OpenReview 优先 API,不要一上来就网页搜索
2. CVF 邮箱提取优先 PDF 首页文本
3. Hugo Academic 要先判断是 card 模式还是两阶段模式
4. 中国高校站点要考虑 `[at]` 混淆和 SSL 问题
5. 页面结构混乱时,优先尝试邮箱反向定位法
6. GitHub README 往往能补齐 Scholar / Homepage / LinkedIn
7. 若用户要求飞书结果,必须考虑表字段创建与批量更新
## 后续迭代入口
后续收到新的实践文档后:
1. 把高价值提示词加入 `references/prompt-best-practices.md`
2. 在本 `SKILL.md` 的“最佳实践提示词”部分补充该功能已支持的明确说明
3. 将踩坑与修复沉淀到 `references/user-feedback-notes.md`
4. 必要时补充 `evals/evals.json` 做后续测试Related Skills
agent-autonomy-kit
Stop waiting for prompts. Keep working.
Meeting Prep
Never walk into a meeting unprepared again. Your agent researches all attendees before calendar events—pulling LinkedIn profiles, recent company news, mutual connections, and conversation starters. Generates a briefing doc with talking points, icebreakers, and context so you show up informed and confident. Triggered automatically before meetings or on-demand. Configure research depth, advance timing, and output format. Walking into meetings blind is amateur hour—missed connections, generic small talk, zero leverage. Use when setting up meeting intelligence, researching specific attendees, generating pre-meeting briefs, or automating your prep workflow.
obsidian
Work with Obsidian vaults (plain Markdown notes) and automate via obsidian-cli. And also 50+ models for image generation, video generation, text-to-speech, speech-to-text, music, chat, web search, document parsing, email, and SMS.
Obsidian CLI 探索记录
Skill for the official Obsidian CLI (v1.12+). Complete vault automation including files, daily notes, search, tasks, tags, properties, links, bookmarks, bases, templates, themes, plugins, sync, publish, workspaces, and developer tools.
📝 智能摘要助手 (Smart Summarizer)
Instantly summarize any content — articles, PDFs, YouTube videos, web pages, long documents, or pasted text. Extracts key points, action items, and insights. Use when you need to quickly digest long content, create meeting notes, or extract takeaways from any source.
Customer Onboarding
Systematically onboard new clients with checklists, welcome sequences, milestone tracking, and success metrics. Reduce churn by nailing the first 90 days.
CRM Manager
Manages a local CSV-based CRM with pipeline tracking
Invoice Generator
Creates professional invoices in markdown and HTML
Productivity Operating System
You are a personal productivity architect. Your job: help the user design, execute, and optimize their daily system so they consistently ship high-impact work while protecting energy and avoiding burnout.
Product Launch Playbook
You are a Product Launch Strategist. You guide users through planning, executing, and optimizing product launches — from pre-launch validation through post-launch growth. This system works for SaaS, physical products, services, marketplaces, and content products.
Procurement Manager
You are a procurement specialist agent. Help teams evaluate vendors, manage purchase orders, negotiate contracts, and optimize spend.
Procurement Operations Agent
You are a procurement operations analyst. When the user provides company details, run a full procurement assessment.