Proofreading — Development skill for Claude Code

Proofreading

A Claude Code skill for proofreading academic papers against a structured checklist covering paper structure, math notation, statistical relevance, figures and tables, grammar and style, and abbreviations.

Experience

[I](https://jakob-thumm.com) went through a hard school with [my professor](https://www.ce.cit.tum.de/en/cps/members/prof-dr-ing-matthias-althoff/) highlighting every single detail that differentiates an average paper from an outstanding scientific work. Over the years of my Ph.D., I received feedback on my paper drafts 44 times! I went through all of this feedback and compiled a comprehensive list of checks that regularly come up during proof reading. That being said, especially Check 5 (Grammar and Style) is highly subjective and should be adapted to your personal needs.

If you find this skill useful, **give it a star** and **share it with your colleagues**. Do not hesitate to open up a pull request to extend this skill further!

Features

• Report mode: full prioritized issue report with line-level findings
• Interactive mode: fix issues step by step together with the agent
• PDF annotation extraction: when given an annotated PDF, extracts highlights, notes, and strikethrough comments from reviewers

Checks

This skill contains over 100 checks, which we categorize into 6 groups. The following list gives a brief summary of the most important checks in each group.

**Check 1 — Paper Structure** (`intro`)

• Abstract follows the four-part structure: motivation, gap, approach, quantitative result
• Introduction contains an explicit, falsifiable contribution list
• Related work is organized thematically and ends with a gap statement

**Check 2 — Math Symbols and Notation** (`math-notation`)

• Every symbol and index is defined before first use
• Scalar/vector/matrix/set convention is detected and applied consistently
• Equations are punctuated and integrated into the surrounding text

**Check 3 — Statistical Relevanc