novelty-matrix_skill

Overview

This skill builds a novelty/prior-work matrix that aligns each submission claim with the closest related works and records overlaps, deltas, and evidence. It produces a structured NOVELTY_MATRIX.md that makes positioning concrete and traceable. The output is evidence-first and designed for peer review use.

How this skill works

It reads the contribution list from output/CLAIMS.md and a provided related-work list (or the paper’s citations) and selects the closest ≥5 prior works when possible. For every claim × related-work pair it records: overlap (what matches), delta (what’s new), and evidence pointers (paper section, quote, citation). Finally it summarizes clearly novel claims and those that appear incremental.

When to use it

• Preparing or writing a peer review that must justify novelty and positioning
• Before submitting rebuttal or response to reviewers to map claims to prior art
• When the paper has claims but you need a structured, evidence-backed comparison
• If you want to flag claims that lack clear novelty or need stronger citations
• During internal author review to identify where stronger claims/evidence are required

Best practices

• Ensure output/CLAIMS.md is complete and granular (one claim per line) before running the matrix
• Provide at least the paper’s reference list; otherwise mark needs_related_work_list and stop
• Force each cell to cite a concrete axis: problem, method component, data, eval protocol, or results
• Prefer short quoted evidence pointers (section, page, figure, or DOI) over freeform opinion
• Include ≥5 related works or explicitly document why fewer were used
• Summarize novelty with clear labels: clearly novel, incremental, unclear/no evidence

Example use cases

• Reviewer assessing whether claimed algorithmic component is original vs. cited baselines
• Author mapping a new dataset claim against existing datasets to show differences in scale/annotations
• Conference meta-reviewer auditing novelty statements across submissions
• Author preparing a rebuttal that enumerates deltas with exact evidence pointers
• Research team prioritizing which claims need stronger experiments or citations before submission