nature-reviewer

Nature Reviewer

Assess a manuscript before submission. Use nature-response after a real editor decision or reviewer report.

Contract

Default to three reviewer reports with different emphases plus one synthesis; follow another requested count or structure when it preserves independent assessment and synthesis.

Evaluate:

• originality;
• scientific importance;
• interdisciplinary readership;
• technical soundness and reproducibility;
• readability for non-specialists.

Read manifest.yaml. Load the current local source basis and only the workflow, axes, domain gates, structure, role, or QA references needed at the current stage.

Workflow

1. Establish source coverage and extract one shared manuscript fact base.
2. Separate author claims, visible evidence, limitations, external checks, and facts not assessable from the supplied material.
3. Build a concern ledger. Each substantive concern needs a stable ID, claim pointer, evidence pointer, severity, and resolution test.
4. Draft reviewers from the same facts with different emphases, not invented identities or biographies.
5. Synthesize consensus only when multiple reports raise the same underlying concern; preserve legitimate weighting differences.
6. Run groundedness, coverage, issue-key consistency, synthesis, role-boundary, and non-invention QA.

Boundaries

• Missing material means not assessable, not proof that the authors omitted the work.
• Do not invent experiments, controls, citations, figure panels, line numbers, reviewer identities, institutions, specialties, or selection history.
• Do not infer field novelty solely from the manuscript's related work.
• Do not make the editor's final decision or guarantee fit to Nature.
• The bundled editorial criteria are a local snapshot. Verify current official policy when the user needs policy-level certainty, or state the limitation.

Completion

Finish when every substantive concern is evidence-anchored, unsupported and not-assessable claims are labelled, repeated concerns share stable issue keys, and the synthesis distinguishes supported consensus from differences in emphasis.