# Peer Review > Use when reviewing an academic journal manuscript, referee report, peer-review PDF, manuscript plus supplementary materials, or when the user wants a structured journal review with critical, important, and minor revisions. This skill uses the academic-research reviewer workflow, local geography/GIScience/health geography field context when relevant, and pdf/docx tools for manuscript and visual-material inspection. Chinese outputs default to Simplified Chinese. - Skill: `lingwei-zheng/peer-review` (Agent Skill, multi-file: 4 files) - Install (CLI): `npx skillmds@latest add lingwei-zheng/peer-review` - Raw SKILL.md: https://api.skillmd.com/api/skills/lingwei-zheng/peer-review/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: lingwei-zheng (https://skillmd.com/u/lingwei-zheng) - Updated: 2026-09-21 - Page: https://skillmd.com/skills/lingwei-zheng/peer-review --- # Journal Peer Review Produce a rigorous external-reviewer assessment from the manuscript and all available supplementary materials. This is a submission-stage external review and therefore uses comprehensive rigor. Do not export its full robustness burden backward into exploratory or first-draft workflows; those are calibrated by `../shared/research-calibration.md`. ## Review Stack 1. Read `../academic-research/SKILL.md` and use its `academic-paper-reviewer` workflow as the main review framework. 2. Use `../shared/field-context/` only for geography, human geography, GIScience, GeoAI, remote sensing, spatial analysis, environmental exposure, or health geography. Do not assume NORA is installed. 3. Use the active PDF and DOCX skills for extraction, media inspection, rendering, and visual review. 4. Read [references/report-structure.md](references/report-structure.md) before drafting the final reports. ## Non-Negotiable Visual Rule Do not finalize a review from extracted text alone when figures, maps, diagrams, tables, equations, statistical charts, or supplementary visuals support the claims. Attempt visual inspection first. If it cannot be performed, disclose exactly what was not inspected and how that limits the review. Chinese working reports default to Simplified Chinese. Use Traditional Chinese only when explicitly requested. ## Workflow ### 1. Discover Materials - Identify the main manuscript and every supplementary file. - Read `.codex/project.yaml` when present. - Record what was reviewed, rendered, visually inspected, or unavailable. ### 2. Inspect Text And Evidence - For PDFs, combine text extraction with rendered-page inspection. - For DOCX, inspect OOXML and embedded media when relevant. - Inspect central figures, maps, tables, plots, diagrams, equations, and supplementary evidence supporting key claims. - Track whether each concern arises from text, captions, visual evidence, supplements, methods, statistics, or domain reasoning. ### 3. Run Independent Review Lenses Assess field and journal fit, research question, contribution, method validity, reproducibility, sampling, representativeness, statistical assumptions, figure/table evidence, claim-evidence alignment, causal language, limitations, ethics, and data access. Preserve independent lenses before synthesis. Treat plain association language as already non-causal. Request a causal correction only when the manuscript uses causal verbs, mechanisms, mediation, or policy effects beyond the design; do not demand `exploratory association`, `observational association`, or repeated no-causality disclaimers. For relevant geography manuscripts, read only the needed field references: - `geography-publication-readiness.md` for the five publication gates and desk-reject risk. - `spatial-methods.md` for scale, spatial units, MAUP, dependence, and validity. - `environmental-health.md` for exposure and outcome pathways. - `geoai-domain.md` for GeoAI or geospatial machine learning. - `journal-templates/` for venue fit. Use a devil's-advocate pass to test the strongest plausible rejection risk, but do not manufacture one. If no material critical risk is supported, say so and keep the issue at the appropriate lower severity. Explicitly test whether a location change has cumulative, transferability, policy, measurement, or boundary-condition value; do not reject it for similarity alone. Also test method stacking without inferential gain, theory-as-label, Discussion restatement, and title overpromise when relevant. Apply spatial diagnostics only when the paper's claims or method assumptions trigger them. Mark MAUP, residual spatial dependence, alternative weights, GWR/MGWR, spatial CV, transfer, or boundary checks `N/A` with a reason when they do not apply; their absence is not automatically a weakness. ### 4. Triage And Draft Review central claims, primary figures/tables, and load-bearing method choices before lower-impact details. Separate zero or one highest-risk critical issue from important and minor issues. Follow user-specified counts when provided; otherwise use the defaults in [references/report-structure.md](references/report-structure.md). Base the editorial recommendation on fixability, not tone or preference. Bind every critical and important issue to an inspectable evidence pointer: page, section, paragraph, figure, table, equation, supplement, or exact claim. State whether the concern is directly observed, inferred from the manuscript, or cannot be verified from the available files. Do not use a vague topic label as the evidence for a major criticism. ### 5. Verify Before delivery, confirm that visual inspection or its limitation is disclosed, there is at most one critical issue, issue counts are respected, geography gates were applied when relevant, every critical and important issue has an evidence pointer, and user-edited files were not overwritten. ## Output Create only the report set the user needs. Use [references/report-structure.md](references/report-structure.md) as the source of truth for filenames, section structure, issue hierarchy, decision standards, tone, and final report checks.