An AI skill suite that automates the slow, repetitive parts of applying for a PhD — finding openings, analyzing professors, drafting outreach, tailoring materials, and tracking deadlines — while keeping everything you send personal and authentic.
It is field-agnostic: your research area, target regions, and preferred sources are
configured once in knowledge-base/profile/profile.md, and the same skills work whether
you're applying in machine learning, molecular biology, or medieval history.
可自动化博士申请过程中那些耗时且重复性的工作:包括查找招生机会、分析教授的研究方向、撰写联系邮件、针对不同项目定制申请材料，以及跟踪各项截止日期，同时确保你发送的所有内容都保持个人风格，真实且自然。
你的研究方向、目标地区和偏好的信息来源只需在 knowledge-base/profile/profile.md 中配置一次，之后无论你申请的是机器学习、分子生物学，还是中世纪史，均可使用同一套技能完成整个申请流程。
The system is a set of focused skills that share a knowledge base of Markdown files. Skills don't call each other; they read and write the same files (a "blackboard"). That makes every step inspectable and hand-editable, and the whole thing version-controlled.
skills/ the suite — each dir is one skill with a SKILL.md
knowledge-base/ your data: profile, professors, openings, applications, interactions
shared/schemas/ the agreed-upon shape of each knowledge-base file (+ validate_kb.py)
shared/references/ cross-cutting guidance: data sources, connectors, ethics
ARCHITECTURE.md full design: skills, collaboration, data sources, roadmap
REPORT.md what was built + how it was evaluated
- Fill in
knowledge-base/profile/profile.md(who you are, what you want, where to look) andprofile/cv-master.md(your full CV). - Ask Claude to analyze a professor — e.g. "Do a deep dive on Prof. Jane Smith at MIT
for fit with my background." The professor-analyzer skill writes a profile to
knowledge-base/professors/. - From there, draft outreach, rank opportunities, or tailor materials — each skill picks up the files the previous one wrote.
All nine skills are implemented (each is a SKILL.md under skills/) and evaluated against
realistic prompts — see REPORT.md for the build and evaluation write-up.
| Skill | What it does |
|---|---|
| phd-copilot | front door: surveys the knowledge base and routes to the next best step |
| position-discovery | finds current, matching PhD openings → openings/ |
| professor-analyzer | deep-dives a professor/lab and scores fit → professors/ |
| opportunity-ranker | ranks openings on weighted, transparent dimensions → openings/_ranking.md |
| outreach-email | drafts grounded cold/follow-up emails (never sends) → interactions/ |
| research-proposal | drafts a proposal at the lab's open problems → applications/<id>/proposal.md |
| application-materials | tailors CV / SOP / personal statement → applications/<id>/ |
| application-tracker | surfaces what's overdue, due-soon, or missing across applications |
| interview-prep | generates likely questions + model answers → applications/<id>/interview.md |
Validate a knowledge base against the shared schemas at any time:
python3 shared/schemas/validate_kb.py knowledge-base
See ARCHITECTURE.md for the full design and rationale.