wtf is paper-contribution-helper?

This repository is an AI-assisted tool aimed at academic researchers, particularly graduate students and early-career authors who need to present their work more convincingly to peer reviewers. The specific problem it targets is a common pattern in research papers where the contribution is written as "we combine methods A, B, and C," which often leads reviewers to dismiss the work as incremental or lacking novelty. The goal is to take the existing experiments and methods and reorganize them into a stronger, more defensible narrative without inventing new claims. The tool works as a skill for OpenAI Codex and ChatGPT. It has two main modes. In the first mode, you give it a paper PDF and it diagnoses where reviewers are likely to attack, surfaces strengths that the paper already contains but has not explained well, and suggests how to rewrite the abstract, introduction, and contribution bullets. In the second mode, it reads the target paper, collects related work from the same research area, and generates a domain-specific helper skill that can be reused for future papers in the same field. The second mode requires Codex, ChatGPT on the web can only use the first mode or a pre-generated domain skill. The example used throughout the README is a paper called SemiDFL, which combines three existing techniques for semi-supervised federated learning. The tool reframes that combination not as "we stacked three modules" but as "there are three missing consensus interfaces in this setting, and we close each one." The README includes a before-and-after table showing how this shift in framing changes how a reviewer reads the contribution. The stated principle is to be strong but honest. The tool does not encourage making up experiments or exaggerating claims. Instead it tries to surface what is genuinely novel in the paper but was explained poorly, and to prepare the author for specific reviewer objections around novelty, baseline fairness, mechanism evidence, and reproducibility. The repository includes example outputs for the SemiDFL paper in both version 1.0.2 and 1.0.8, including a full reframing report, a generated domain skill zip, screen recordings from Codex, and saved ChatGPT session files.