sihengtao/phd-application-planner

PhD Application Planner is a skill for Claude Code or Codex that helps someone plan PhD applications by turning a short conversation into an interactive dashboard. Instead of manually researching schools and advisors, you answer a few questions about your field, target regions, minimum stipend, and preferences, and the tool researches programs and potential advisors for you, then builds a dashboard you can browse and filter. The dashboard is built with marimo, a tool for making interactive Python apps. It shows key numbers at a glance, a list of top-fit picks, and a program browser you can filter by region, fit, stipend, or application strategy. Each program gets a detail card covering stipend, cohort size, deadlines, application restrictions, and notes for international students. There is also an advisor browser where you can filter by category, research area, region, and citation count, hide advisors you are not interested in, and write notes that are saved even after you close the app. A slider-based ranker lets you weight what matters most to you, such as fit, stipend, or location, and the dashboard re-ranks live. It also includes statistical analysis, such as clustering and correlation charts, to help spot patterns across programs. To use it, you need Claude Code or Codex with local skills enabled, plus Python 3.9 or newer with a few libraries installed, including marimo, pandas, and scikit-learn. You install it by cloning the repository into your skills folder, then asking your AI assistant to help you find programs in your field. The tool runs through intake questions, writes a config file, researches programs and advisors, builds the dashboard's data files, and launches it in your browser. The README is honest that the research results are snapshots pulled from the web and may be out of date, so it recommends double-checking stipends and deadlines on official program pages before making decisions. The project works for any academic field or region and does not include any built-in personal data. It is released under the MIT license.