Peer Review Check — Development skill for Claude Code

peer-review-check

A [Claude Code / Agent SDK skill](https://docs.claude.com/en/docs/claude-code/skills) that does an **author self-check** of a manuscript draft against Elsevier's official [*Structured peer review prompts*](https://www.elsevier.com/) (April 2026), before you submit.

Instead of waiting for reviewers to tell you what's missing, it pre-empts them: it reads your draft, detects the article type, and runs one focused sub-reviewer per manuscript section, each answering the exact Elsevier prompt a real reviewer will be asked. It then synthesises a **severity-ranked gap list** and a submission-readiness verdict.

What you get

For every Elsevier prompt, the report tells you:

• Verdict — pass / minor gap / major gap
• What your draft has — the specific evidence, cited from your manuscript
• What a reviewer would object to
• The concrete fix to make before submitting

Ranked most-likely-objection first, with a short "already passes" list so you know what not to touch.

Coverage

Follows Elsevier's two prompt sets verbatim:

• Original research articles — Abstract, Introduction, Methods, Results, Discussion, References, Overall (novelty + ethics/integrity), and Title.
• Review articles — Overall relevance, Historical development, References, Methods & reproducibility, Statistical analyses, Essential reporting (PRISMA), Language quality.

Non-applicable prompts (e.g. MCID for a non-clinical paper) are marked pass rather than invented as problems.

Install

Clone into your skills directory:

git clone https://github.com/narcoco/peer-review-check.git ~/.claude/skills/peer-review-check

Use

Point the assistant at a draft (PDF, `.docx`, or `.md`/`.tex`) and ask any of:

• /peer-review-check
• "check my paper before submission"
• "is my manuscript submission-ready"
• "will this pass review"

How it works

• Multi-agent deep review: one sub-reviewer per section runs in parallel, each scoped to its Elsevi