name: paper-review
description: >
Perform systematic peer review of a biomedical research paper.
Extracts claims, validates citations, cross-checks literature,
audits data integrity, detects contradictions, reviews methodology,
scores evidence quality (RCMXT), and generates a structured review
report with major/minor comments and decision recommendation.
metadata:
author: bioteam-ai
version: "1.0"
domain: biomedical-research
allowed-tools: Bash(python:*) Read
Paper Review Skill
When to Activate
- User asks to "review a paper" or "peer review this manuscript"
- User wants to assess a paper's methodology, claims, or evidence quality
- User needs a structured review report for a journal submission
Usage
cd backend && uv run python -m app.skills.w8_cli --pdf /path/to/paper.pdf --budget 3.0
Options
--pdf / -p: Path to the paper PDF (required)--budget / -b: Maximum budget in USD (default: $3.00)--output / -o: Output JSON file path--markdown / -m: Output Markdown report file path
Pipeline (12 Steps)
- INGEST: Read and validate PDF file
- PARSE_SECTIONS: Extract text, split by headings (PyMuPDF)
- EXTRACT_CLAIMS: LLM extracts structured claims per section
- CITE_VALIDATION: Validate citations (DOI/PMID parseable)
- BACKGROUND_LIT: Search PubMed/bioRxiv for supporting/contradicting evidence
- INTEGRITY_AUDIT: Statistical checks, gene name validation, retraction check
- CONTRADICTION_CHECK: Compare paper claims vs. literature
- METHODOLOGY_REVIEW: Deep assessment (study design, controls, statistics)
- EVIDENCE_GRADE: RCMXT scoring of key claims
- HUMAN_CHECKPOINT: Pause for reviewer input
- SYNTHESIZE_REVIEW: Generate structured peer review
- REPORT: Assemble final Markdown report
Output
- Structured review with major/minor comments
- Decision recommendation (accept/minor/major revision/reject)
- Methodology assessment with score
- RCMXT evidence quality scores
- Citation integrity report
- AI disclosure statement
By