peer-review

Structured manuscript and grant review with checklist-based evaluation covering methodology assessment, statistical validity, reporting standards compliance (CONSORT/STROBE), and constructive feedback for formal peer review writing.

Write formal peer reviews with structured checklist evaluation

Source K-Dense AI
License MIT
First documented

Trigger phrases

Phrases that activate this skill when typed to Claude Code:

  • peer review this manuscript
  • review this paper
  • write a formal review
  • evaluate this manuscript
  • critique this study

What it does

peer-review is a Claude Code skill from K-Dense AI’s scientific-agent-skills repo. It turns Claude into a structured manuscript reviewer that works through a checklist covering major scientific concerns — hypothesis clarity, study design, statistical validity, reporting standards compliance, and writing quality — then produces a formal review document with major and minor comments formatted for journal submission.

A session produces a review letter: a summary paragraph, numbered major concerns, numbered minor concerns, and a recommendation (accept/minor revision/major revision/reject). The checklist ensures nothing obvious is missed.

When to use it

Reach for it when:

  • You’ve been assigned a peer review and want a structured first pass before adding your expert judgment
  • You’re revising your own manuscript and want to anticipate what reviewers are likely to flag
  • You’re mentoring junior researchers and want to show what a complete review looks like

When not to reach for it:

  • Evaluating the quality of evidence for a clinical decision — use scientific-critical-thinking
  • Scoring a set of manuscripts with a numerical rubric — use scholar-evaluation

Install

Copy the SKILL.md from K-Dense AI’s peer-review folder into .claude/skills/peer-review/ in your project.

Trigger phrases: “peer review this manuscript”, “review this paper”, “write a formal review”.

What a session looks like

A typical session has three phases:

  1. Manuscript intake. Paste the manuscript text or provide the file path. Claude identifies the study type (RCT, observational, review) and selects the appropriate reporting checklist.
  2. Checklist evaluation. Claude works through each checklist item systematically, flagging concerns with supporting quotes from the manuscript.
  3. Review letter draft. Findings are structured into a formal review document with summary, major concerns, minor concerns, and a recommendation. You edit and add domain expertise before submitting.

Receipts

Where it works well:

  • Methods and statistics sections — Claude catches missing details (sample size justification, blinding procedures, multiple comparison corrections) reliably
  • Formatting and reporting guideline compliance — CONSORT flow diagrams, STROBE checklists, PRISMA flow

Where it backfires:

  • Novelty assessment — Claude cannot judge whether a finding advances a field without domain knowledge you supply
  • Detecting subtle data fabrication or image manipulation — this is not a research integrity tool

Pattern that works: use the skill output as a structured scaffold, then layer your own expert commentary on top rather than treating the generated review as final.

Source and attribution

Originally authored by K-Dense Inc.. The canonical SKILL.md lives in the peer-review folder of their public scientific-agent-skills repository.

License: MIT. Install, adapt, and redistribute with attribution preserved.

This page documents the skill from a practitioner’s perspective. For the formal spec and any updates, defer to the source repo.