Peer Review Check

Peer review check: author self-check against Elsevier prompts

Angela hands over her own manuscript draft. This skill checks it, section by section, against Elsevier's official Structured peer review prompts (April 2026) and returns a gap list: for each prompt a reviewer will be asked to answer, whether the draft satisfies it, and if not, the exact fix. The point is to catch every reviewer objection before she submits, not to referee someone else's paper.

Depth is multi-agent deep: one sub-reviewer per manuscript section runs in parallel, each focused on its Elsevier prompt, then results are synthesised into one ranked report.

The full prompt tables (both article types) live in references/elsevier-prompts.md. Read that file at the start of every run.

Flow

1. Locate and read the manuscript

• If Angela gave a path, use it. Otherwise ask for the file (PDF, .docx, or .md/.tex).
• Read the whole thing:
• .md / .tex / .txt → Read directly.
• .pdf → Read via the pages parameter (whole document).
• .docx → convert with the markitdown skill or pandoc, then read.
• Do not review from the abstract or memory alone. Read the source first (OPUS hard rule 3):

every gap you report must cite something actually in the manuscript.

2. Detect the article type, confirm in one line

Decide original research vs review article (the two tables in the reference file). Signals: a Methods/Results with new data and a study design → original research; a synthesis of the field with search strategy / PRISMA / meta-analysis → review article. State the call in one line and proceed; only pause if genuinely ambiguous.

3. Fan out one sub-reviewer per section (parallel)

For the detected article type, launch a general-purpose Agent per row in that table (Abstract, Introduction, Methods, ... for original research; Overall relevance, Historical development, ... for review). Launch them in a single message so they run concurrently. Give each agent:

• the manuscript file path (it reads the file itself),
• its section name and the exact Elsevier prompt text for that section,
• this instruction: *"You are pre-empting a peer reviewer who will answer this prompt about

Angela's own draft. Judge whether the manuscript would satisfy that reviewer. Return, as a compact structured list: (a) verdict — PASS / MINOR GAP / MAJOR GAP; (b) evidence — quote or cite the specific place in the manuscript, with section/page; (c) what a reviewer would object to; (d) the concrete fix Angela should make before submitting. Only flag what is genuinely missing or weak. Do not invent problems, do not praise. If the prompt asks for something not applicable to this study (e.g. MCID for a non-clinical paper), say so and mark PASS."*

Always include the Overall (integrity/novelty) and Title prompts for original research; always include Language quality and Essential reporting for review articles.

4. Synthesise the gap list

Collect all sub-reviewer returns into one report. Structure:

1. Verdict line — submission-ready / minor revisions needed / major gaps before submitting.
2. Major gaps first, then minor gaps, each as:

• Section (Elsevier prompt it maps to)
• What the prompt asks
• What your draft has (the evidence the sub-reviewer found)
• Fix (specific, actionable)

1. Passed — a short list of prompts the draft already satisfies, so she knows what not to touch.

Rank by how likely each gap is to draw a reviewer objection. Keep it scannable. Follow Angela's style rules: sentence-case headings, no em dashes, no "not X but Y" construction.

5. Save and state the path

Save the report as Markdown next to the manuscript by default: /_peer-review-check.md. If the manuscript is inside the Research vault or OneDrive, keep it there; do not move it. Always state the absolute save path and whether it is in the Obsidian vault or OneDrive (OPUS hard rule 4). Then give a 3-5 line summary in chat: verdict + the top gaps.

Notes

• This is self-check, so frame everything as "before you submit, a reviewer will say X — here is

how to close it." Not a referee report to send to a journal.

• If Angela instead wants to referee someone else's manuscript, or wants a full simulated

multi-reviewer panel, route to the academic-paper-reviewer skill.

• The reference file is verbatim Elsevier text — do not paraphrase the prompts when handing them

to sub-reviewers; a reviewer will answer the exact wording.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: narcoco
• Source: narcoco/peer-review-check
• License: MIT

Install and usage instructions live in the source repository linked above.