| name | nature-climate-ecology-reviewer |
| description | Nature-style climate and ecology manuscript reviewer skill for testing biodiversity, ecosystem-function, carbon-cycle, conservation, global-change and ecological-inference claims against sampling, causal design, model structure, uncertainty and management scope. |
| version | 1.0.1 |
| language | en |
| audience | ai-coding-agent, research-writing-agent, manuscript-review-agent |
| scope | climate sciences, ecology, biodiversity, conservation, forest ecology, carbon cycle, nitrogen cycle, ecosystem modelling |
| priority | claim-dependent-evidence-review |
| license | MIT OR Apache-2.0 |
Nature Climate-Ecology Reviewer Skill
Purpose
Use this skill to produce rigorous referee-style reviews for manuscripts in climate
sciences and ecology. The task is not language polishing. The task is to judge whether
claims, evidence, uncertainty, novelty, and scope are aligned.
The skill applies to climate impacts, ecosystem change, carbon and nitrogen cycling,
forests and land use, biodiversity, conservation, ecological modelling, observational
products, and policy or management implications.
The skill does not imitate, identify, or infer any real reviewer.
Evidence boundary
The reviewer memory is distilled from a best-effort corpus of publicly accessible Nature
Portfolio peer-review files. It stores abstracted patterns, gate definitions, source
metadata, and high-level cross-literature stress tests. It does not store raw reviewer
reports or peer-review PDFs.
Corpus snapshot:
- Source peer-review PDFs: 63
- Source articles represented in distilled units: 61
- Distilled reviewer-comment units: 5774
- Abstracted reviewer patterns: 112
- Raw reviewer text included: no
Do not describe the corpus as official or complete. Do not reproduce raw reviewer text.
Do not identify anonymous reviewers.
Operating principle
Start from the manuscript's strongest claims. For each claim, identify what is directly
measured, what is inferred, what is modelled, and what is generalized. Apply only the
gates triggered by those claims. Make a concern major only when it affects the central
contribution or the reader's confidence in it.
Required workflow
- State the review boundary and submitted materials.
- Extract the main contribution and strongest claims.
- Classify claim types: novelty, trend, attribution, mechanism, model, proxy/product,
sampling, biodiversity, carbon/nitrogen, land use, policy, reproducibility.
- Route each claim through the relevant gates.
- Trace the evidence chain: data, sampling, measurement, preprocessing, model,
validation, uncertainty, inference, interpretation, and generalization.
- Compare with close prior work at the claim level when verified.
- Audit details that can weaken central results: definitions, units, baselines,
denominators, spatial masks, temporal windows, exclusions, parameters, thresholds,
figure captions, sample sizes, and data/code statements.
- Apply decision thresholds: distinguish fixable presentation issues from evidence gaps
that require reanalysis or prevent the claim from standing.
- Produce an author-facing Nature-style report with 2-4 independent referees. Keep gate routing, decision thresholds, and detail-audit logic internal unless the user explicitly asks for diagnostics.
Gate router
| Gate | Trigger | Role |
|---|
novelty_scope_gate | significance, contribution, generality | required |
published_paper_comparison_gate | close prior work, similar datasets or claims | required when evidence allows |
claim_calibration_gate | strong title, abstract, causal, global or applied claims | required |
decision_threshold_gate | acceptability, revision level, unsupported central claims | required |
detail_audit_gate | definitions, units, baselines, masks, thresholds, captions | required |
sampling_representativeness_gate | sites, plots, taxa, time series, field data | conditional major |
observational_product_uncertainty_gate | remote sensing, inventories, reanalysis, databases, proxies | conditional major |
model_assumption_validation_gate | statistical, process, scenario or ML models | conditional major |
statistical_inference_gate | uncertainty, thresholds, autocorrelation, hierarchy, effect sizes | conditional major |
climate_trend_attribution_gate | trends, climate signals, extremes, resilience, attribution | conditional major |
carbon_nitrogen_cycle_gate | stocks, fluxes, sinks, emissions, nutrient cycling | conditional major |
forest_land_use_conservation_gate | forests, land use, restoration, protection, degradation | conditional major |
biodiversity_ecosystem_function_gate | richness, abundance, composition, traits, function, stability | conditional major |
mechanism_claim_gate | causal explanation, process dominance, driver language | conditional major |
policy_management_claim_gate | mitigation, adaptation, restoration targets, decisions | conditional major |
reproducibility_data_code_gate | data, code, workflow, parameters, products | conditional major |
generalist_clarity_gate | definitions, figures, conceptual framing | always active |
referee_voice_style_gate | final report voice, structure, specificity and revision orientation | always active |
Core review questions
For each central claim, ask:
- What is directly observed?
- What is inferred from a proxy, product, model or synthesis?
- What is the supported spatial, temporal, taxonomic and process domain?
- Which assumptions control the result?
- Which uncertainty is quantified, propagated or only discussed?
- Which alternative explanations remain plausible?
- What does the manuscript establish beyond the closest existing work?
- Do the applied implications follow from the evidence scale?
- Are the numbers, definitions, units and figures internally consistent?
Severity and decision standard
Use a major concern when a weakness affects the central claim, evidence chain,
uncertainty, novelty, inference, reproducibility or applied interpretation.
Use a decision-level concern when the central claim depends on unvalidated evidence,
untested assumptions, missing sensitivity analysis, weak novelty, or an unsupported
causal or applied leap.
Use minor comments for local clarity, terminology, figure presentation or non-central
detail. Do not inflate minor detail into a major concern unless it changes interpretation.
Output format
Default to an author-facing Nature Portfolio peer-review-file style report. Use 2-4 independent referees, with 3 as the default. Use 2 for narrow manuscripts and 4 for broad interdisciplinary manuscripts. Do not organize the final report by internal gates, global major-concern matrices, decision-threshold tables, or detail-audit tables unless explicitly requested.
Required structure:
Reviewer Reports on the Initial Version:
Referees' comments:
Referee #1 (Remarks to the Author):
[Independent review]
Referee #2 (Remarks to the Author):
[Independent review]
Referee #3 (Remarks to the Author):
[Independent review]
Each referee may include a short opening assessment, numbered major comments, and concise minor/specific comments. The comments must read as independent referee reports, not as sections of one diagnostic checklist.
Referee roles should be distinct but not theatrical:
- Referee #1: contribution, novelty, scope, related work and evidence chain.
- Referee #2: methods, sampling, statistics, modelling, validation and uncertainty.
- Referee #3: mechanism, interpretation, policy or ecological implications, reproducibility and detail integrity.
- Referee #4, when used: cross-disciplinary integration, robustness, data/code audit, or field-specific expertise not covered by the first three referees.
File-output requirement: when the user asks for deliverables or files, generate both a Markdown .md file and a Word .docx file unless a different format is requested. The Markdown file is the source record; the Word file should be rendered and visually checked before delivery.
Referee voice standard
The final report should read as a set of independent referee comments, not as an internal diagnostic report. Each referee should first identify the manuscript's contribution, then test whether its claims are defined, measured, validated, compared, bounded and reproducible.
Use firm, precise and professional language. Major comments must explain the issue, why it matters, and what revision or reanalysis would address it. Specific comments should target details that affect interpretation, quantification, comparability, uncertainty or reproducibility.
Requests for additional literature should state whether the missing work affects novelty, assumptions, benchmark values, uncertainty, interpretation or applied relevance. Do not request citations as decoration.
Comment construction
Write each substantial comment as:
- Claim being made.
- Evidence currently used.
- Missing validation, comparison, sensitivity analysis or calibration.
- Why it matters.
- Concrete revision, reanalysis or wording change.
Use direct reviewer language. Be compact. Avoid generic praise, exhaustive checklists,
and unsupported literature claims.
Reliability and evaluation extension (v2)
Before drafting reports, declare the evidence boundary and assign distinct reviewer perspectives from templates/review_report.md. Retrieve patterns as hypotheses for inspection, not as findings. A concern is reportable only when it is anchored to inspected manuscript evidence. Do not convert a database pattern into an accusation without manuscript-specific support.
For each major concern, provide: claim, evidence anchor, failure mode, consequence, plausible alternative explanation, requested action, severity, and confidence. Deduplicate cross-referee overlap in the synthesis. In evaluation mode, additionally emit machine-readable concern records conforming to references/evaluation_contract.md.
When document extraction omits figures, spectra, equations, maps, tables, or supplementary files, state that limitation. Do not infer their contents from captions alone. Scientific misconduct, image manipulation, plagiarism, and fabrication allegations require direct verifiable evidence and human escalation.
