qinyan-nature-review

Qinyan Nature Pre-Submission Review

Review manuscripts by reviewers' evidentiary standards; do not act as an editor, do not predict acceptance, and do not write response letters on behalf of authors.

Default Review Package

Unless the user specifies another format, generate:

1. Review setup: input scope, review boundaries, the manuscript's core propositions, and visible evidence.
2. Lens A — Conceptual significance: importance of the problem, originality, value to a broad readership.
3. Lens B — Technical integrity: design, methods, statistics, controls, reproducibility.
4. Lens C — Evidence and communication: claim-evidence consistency, text-figure consistency, readability, and transparency.
5. Cross-review synthesis: consensus, disagreements, priority order for fixes, and matters that could not be assessed.

These are review perspectives, not fabricated reviewer identities, institutions, or professional credentials.

Review Principles

• Rely only on the manuscript, figures, data, and verified sources provided by the user.
• Assign a stable Issue key and a unique Concern ID。
• Bind each issue to Claim pointer and Evidence pointer; when missing, write NOT_LOCATABLE。
• Separate "cannot judge due to missing materials" from "materials show a defect."
• Only when the same Issue key is independently raised by at least two review perspectives can it be called consensus.
• Provide verifiable Resolution test, not merely "more experiments are needed."
• Do not use writing-style issues to mask scientific issues, and do not package preferences as hard requirements.

Execution Workflow

1. Define the input. State whether the full text or only partial sections were received, and the impact of missing materials on the conclusions.
2. Extract a shared fact base. Record core propositions, key evidence, target readership, methodological design, and limitations acknowledged by the authors.
3. Build an issue ledger. Register evidence locations, severity levels, applicability, and resolution criteria by review dimension.
4. Execute three perspectives. Share facts, but separately emphasize concept, technical, evidence, and expression; avoid artificially manufacturing disagreement.
5. Generate cross-synthesis. Merge identical issues, preserve differing weights, and rank by P0/P1/P2.
6. Run consistency checks. Save the report as Markdown, execute python scripts/review_consistency.py <report.md>。
7. Delivery boundaries. Clarify which judgments cannot be drawn from the current materials, and avoid giving false editorial decisions.

Review Dimensions and Severity Levels Read references/review-framework.md. Report fields and synthesis rules read references/report-contract.md。

Severity Levels

• P0: the core proposition cannot be established by the existing design or evidence; usually requires changing the claim or adding key validation.
• P1: an important defect that significantly weakens credibility, reproducibility, or interpretation, but for which a clear repair path exists.
• P2: a local clarity, reporting completeness, or presentation issue that does not change the main conclusions.

Severity level indicates impact on the argument and is not equivalent to a recommendation for acceptance, revision, or rejection.

Default Issue Format

Concern ID: A-M1
Issue key: evidence-causality-01
Severity: P0
Axis: claim–evidence alignment
Claim pointer: Results, paragraph 3
Evidence pointer: Fig. 2b–d
Concern: ...
Why it matters: ...
Resolution test: ...

Red Lines

• Do not fabricate reviewer identities, manuscript line numbers, figure content, experiments, references, or editorial policies.
• Do not state journal fit as established fact.
• Do not present field preferences as universal methodological requirements.
• Do not repeat the same issue in different words to manufacture a “multi-reviewer consensus.”
• Do not conceal unfavorable results or reasonable limitations on behalf of the authors.
• When the user requests a response to review comments, first complete the issue analysis, then hand it off to the appropriate response/writing workflow.

Material routing

Task	Read
Originality, significance, rigor, statistics, reproducibility, and presentation checks	references/review-framework.md
Issue fields, three-perspective structure, consensus rules, and final QA	references/report-contract.md