Goal

• Rate the passed paper from 1 (low) to 5 (best) by its page length, its authors' affiliations, and its citation count
• Follow .claude/skills/paper.rules.md, # Paper Rating

Workflow

Get the Paper

• Use the PDF the user passes
• If the input is a URL, an arXiv ID, or a DOI, download only the PDF with download_academic_paper_to_md.py --input <INPUT> --clear_actions --action download

Collect the Signals

• Count the pages of the PDF
• Read the affiliations of all authors on the first page
• Look up the citation count in Semantic Scholar for the DOI or arXiv ID and note the date
• Do not use memory for any of these values

Rate

• Score each signal and compute the rating with the tables in .claude/skills/paper.rules.md

Write the Rating File

• Save <BASE>.rating.md next to the paper, where <BASE> is the PDF name without .pdf
• Follow .claude/templates/paper_rating.template.md
• State in the file how it was generated, right after the title:

  - Generated by `.claude/skills/paper.rate/SKILL.md`
  

• Report the rating and the three scores to the user

Constraints

• Use whole numbers only
• If a signal is not found, write unknown and leave it out of the mean

Verification

• The rating is an integer from 1 to 5
• Each signal has a value and a score
• The citation count has a source and a date
• The rating equals the mean of the known scores, rounded half up
• The rating file says it was generated by .claude/skills/paper.rate/SKILL.md