jxtse/auto-paper-harvester ? reverse-engineered prompt

Build me a small Python app called Auto Paper Harvester that helps me download full text PDFs for a list of papers I already chose.

I want to paste in DOIs, doi.org links, or a text file with references, and have it save one folder per paper and tell me which ones it could not download and why. It should try open access copies first, then publisher text and data mining access if I have those keys, and finally fall back to my own logged in Chrome for papers my institution can access. Please make it rate limited by default so it does not hammer sites or get blocked.

I want a simple command line tool and a local web page for setup and downloads, and if possible a small agent skill for Claude or Cursor too. Please include a doctor or status command that shows what is configured, and keep any saved settings local. If you need current setup details for browser access or publisher APIs, look up the latest docs online.