agentleFS
Yang1BaiYang1Bai/nature-paper-hub/skills/nature-reviewer-sim/SKILL.md

nature-reviewer-sim

Skill20 starsChanged 49 days agoMITView on GitHub

Pre-submission peer review simulation calibrated to top-tier journals (Nature, Science, Cell, PNAS, ACS, Wiley families). Deploys 7 specialist agents (field analyst, EIC, methodology reviewer, domain reviewer, perspective reviewer, devils advocate, editorial synthesizer) to simulate the full peer review process. Outputs structured reviewer reports, an Editorial Decision Letter, and a Revision Roadmap. Supports three modes: full (all 7 agents, default), quick (EIC only, ~15 min), and re-review (R&R traceability matrix for post-revision verification). Trigger keywords: 模拟审稿, reviewer simulation, pre-submission review, simulate peer review, 审稿模拟, 投稿前审稿, review my paper, 帮我审稿, 预审稿.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Key difference from a generic reviewer: Every reviewer is calibrated to the specific Nature-series journal's standards, loaded from …

    line 19
  2. Distinctive:

    All reviewer calibrations use these specs as the baseline.

    line 202
  3. Distinctive:

    Do NOT italicize: df, SS, MS, OR, CI, VIF, AIC, BIC, CFI, RMSEA, ICC, ANOVA, SEM

    line 270
  4. Very specific:

    DA CRITICAL BLOCKS ACCEPT: If devils_advocate raises any CRITICAL finding → Editorial Decision cannot be Accept or Minor Revision.

    line 282
  5. Distinctive:

    SPECIFICITY REQUIRED: Every criticism must state: what is wrong, where it appears (section/paragraph), and a concrete suggested fix.

    line 286
  6. Very specific:

    After re-review passes: trigger nature-integrity-check in final-check mode

    line 511
  7. Distinctive:

    Only after final integrity check PASS → proceed to Stage 7 export

    line 512

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

nature-reviewer-sim — Nature-Series Pre-Submission Peer Review Simulator

Simulates a complete Nature-journal peer review process before you submit. Automatically identifies the paper's field and target journal, configures 7 specialist reviewer agents, runs independent multi-perspective reviews, and produces an Editorial Decision Letter with a prioritized Revision Roadmap.

Key difference from a generic reviewer: Every reviewer is calibrated to the specific Nature-series journal's standards, loaded from ${CLAUDE_PLUGIN_ROOT}/templates/journal-specs.json. Acceptance thresholds, scope judgments, and statistical standards all reflect the target journal's real expectations.


Trigger Keywords

LanguagePhrases
Chinese模拟审稿、审稿模拟、投稿前审稿、帮我审稿、预审稿、模拟同行评审
Englishreviewer simulation, pre-submission review, simulate peer review, review my paper, peer review simulation, editorial review

Non-Trigger Scenarios

ScenarioUse Instead
Responding to actual reviewer commentsStage 8 rebuttal in nature-paper-hub
Checking citations/references onlynature-citation
Checking integrity before reviewnature-integrity-check
Writing or revising the manuscriptnature-paper-hub Stages 2–5

Agent Team (7 Agents)

#Agent IDRolePhaseNature-Specific Focus
1field_analystAnalyzes paper field; configures all 5 reviewer personasPhase 0Identifies journal fit, scope, paper maturity
2eic_reviewerEditor-in-Chief — journal fit, novelty, significance, readershipPhase 1Loads journal-specs.json for exact scope/IF calibration
3methodology_reviewerPeer Reviewer 1 — research design, statistics, reproducibilityPhase 1Applies statistical_reporting_standards (see §Statistical Standards)
4domain_reviewerPeer Reviewer 2 — literature coverage, theory, field contributionPhase 1Checks for missing seminal/recent Nature-portfolio citations
5perspective_reviewerPeer Reviewer 3 — cross-disciplinary connections, practical impactPhase 1Flags societal/policy relevance gaps for broad-scope Nature journals
6devils_advocateDevil's Advocate — core argument challenges, logical fallacy detectionPhase 1Frame-lock detection; CRITICAL finding blocks Accept decision
7editorial_synthesizerSynthesizes all 5 reviews; produces Editorial Decision + RoadmapPhase 2Consensus/divergence mapping; DA CRITICAL escalation

Agent Boundaries (Non-Overlap Rules)

AgentDOESDOES NOT
eic_reviewerScope, novelty, overall significance, readership interestDeep methodology, literature gaps
methodology_reviewerStats, design rigor, reproducibility, EQUATOR complianceScope judgment, literature breadth
domain_reviewerLiterature coverage, theoretical framework, citation completenessStatistics, interdisciplinary angles
perspective_reviewerBroader impact, policy implications, cross-disciplinary connectionsLiterature audit, statistical details
devils_advocateLogical consistency, evidence gaps, strongest counter-arguments, cherry-pickingStats design, literature coverage, practical impact

Operational Modes

ModeTrigger PhraseAgents ActiveOutputTime
fullDefault / "full review" / "完整审稿"All 7 agents5 review reports + Editorial Decision + Revision Roadmap~45–60 min
quick"quick review" / "快速审稿" / "15 min"field_analyst + eic_reviewerEIC assessment + top-10 issues list~15 min
re-review"re-review" / "验证审稿" / "check revisions" / "R&R"field_analyst + eic_reviewer + editorial_synthesizerR&R Traceability Matrix + residual issues + new Decision~30 min

Mode Selection Logic

"Review this paper"                        → full
"Help me review my paper"                  → full
"快速看一下这篇论文"                          → quick
"15-minute check"                          → quick
"Did I address all reviewer comments?"     → re-review
"Verification review after revision"       → re-review
"Check if my revisions are sufficient"     → re-review

Orchestration Workflow (3 Phases + Nature-Specific Hooks)

User: "模拟审稿" / "Review this paper"
       |
       [Input check: Does manuscript exist? Has nature-integrity-check been run?]
       ⚠️ If integrity check not run → remind user to run nature-integrity-check first
       |
=== PHASE 0: FIELD ANALYSIS & JOURNAL CALIBRATION ===
       |
       +→ [field_analyst]
          1. Read complete manuscript
          2. Identify: primary field, methodology type, paper maturity
          3. Load journal specs: read ${CLAUDE_PLUGIN_ROOT}/templates/journal-specs.json
             → Extract: scope, word limit, figure limit, ref limit, IF, acceptance rate
          4. Assess journal fit (1–5 scale)
          5. Configure 5 reviewer personas with specific:
             - Academic identity (institution, seniority, research focus)
             - Review preferences and known sensitivities
             - What they specifically look for in this journal
       |
       ** CHECKPOINT 1: Present Reviewer Configuration Card to user **
       ** User may adjust reviewer identities before proceeding **
       ** Show: journal specs loaded, reviewer personas, any scope concerns **
       |
=== PHASE 1: PARALLEL MULTI-PERSPECTIVE REVIEW ===
       |
       ⚠️ IRON RULE: All 5 reviewers work independently. No cross-referencing.
       |
       |→ [eic_reviewer]
       |  - Journal scope fit (using loaded journal-specs.json)
       |  - Novelty & significance for target journal tier
       |  - Readership interest
       |  - Data/Code Availability statement presence
       |  - CRediT author contributions presence
       |  - Competing interests declaration presence
       |  - EQUATOR reporting guideline compliance (if clinical/observational)
       |  → Output: EIC Review Report (400–600 words)
       |
       |→ [methodology_reviewer]
       |  - Research design rigor and appropriateness
       |  - Statistical reporting (apply §Statistical Standards below)
       |  - Reproducibility: methods detail, software versions, parameters
       |  - Data transparency: raw data, code, materials availability
       |  - Scan §Red Flags list (see §Statistical Standards)
       |  → Output: Methodology Review Report (600–900 words)
       |
       |→ [domain_reviewer]
       |  - Literature coverage completeness
       |  - Theoretical framework appropriateness
       |  - Missing key citations (especially recent Nature-portfolio papers)
       |  - Incremental vs. substantial contribution to field
       |  - Academic argument accuracy
       |  → Output: Domain Review Report (500–700 words)
       |
       |→ [perspective_reviewer]
       |  - Cross-disciplinary connections
       |  - Practical applications and policy implications
       |  - Broader societal/ethical implications
       |  - Challenges fundamental assumptions from outside the field
       |  → Output: Perspective Review Report (400–600 words)
       |
       +→ [devils_advocate]
          - Strongest counter-argument (200–300 words — most important part)
          - Cherry-picking / confirmation bias detection
          - Logic chain validation (premise → conclusion)
          - Overgeneralization check
          - Alternative explanations analysis
          - Missing stakeholder perspectives
          - "So what?" test
          - Frame-lock detection (unstated premises)
          → Output: Devil's Advocate Report (special format, see §Output Formats)
       |
       ** CHECKPOINT 2: All 5 reports complete. Present summary to user. **
       ** User may ask clarifying questions before Phase 2. **
       |
=== PHASE 2: EDITORIAL SYNTHESIS & DECISION ===
       |
       +→ [editorial_synthesizer]
          1. Read all 5 Phase 1 reports
          2. Map consensus (≥3 reviewers agree) vs. divergence (conflicting opinions)
          3. Arbitrate divergent opinions with reasoning
          4. Flag all DA CRITICAL findings in the Decision Letter
             ⚠️ IRON RULE: If DA CRITICAL findings present → Decision CANNOT be Accept
          5. Produce Editorial Decision Letter (Accept / Minor Revision / Major Revision / Reject)
          6. Produce Revision Roadmap (prioritized, ready for nature-paper-hub Stage 3 revision input)
       |
       ** CHECKPOINT 3: Editorial Decision delivered. **
       ** If Decision = Minor/Major Revision → offer Socratic Revision Coaching **
       ** User can say "just give me the roadmap" to skip coaching **

Nature-Specific Calibration

Journal Specs Loading

At Phase 0, field_analyst reads ${CLAUDE_PLUGIN_ROOT}/templates/journal-specs.json and extracts for the target journal:

{
  "journal": "Nature Catalysis",
  "scope": "...",
  "word_limit_body": 3000,
  "word_limit_abstract": 150,
  "figure_limit": 6,
  "reference_limit": 50,
  "acceptance_rate": "~5%",
  "impact_factor": 37.8,
  "methods_location": "after_references",
  "special_rules": [...]
}

All reviewer calibrations use these specs as the baseline.

Mandatory Nature Compliance Checks (EIC Reviewer)

The EIC reviewer checks these Nature-specific requirements in every review:

CheckPass ConditionSeverity if Missing
Data Availability StatementPresent and specific (not generic "available upon request")MAJOR
Code Availability StatementPresent if computational work includedMAJOR
CRediT Author ContributionsAll 14 CRediT roles considered, at least contributor roles filledMAJOR
Competing Interests DeclarationPresent (even if "none")MAJOR
EQUATOR ComplianceCONSORT/ARRIVE/STROBE checklist attached if applicableMAJOR
Ethics ApprovalPresent for human/animal studiesCRITICAL
Funding/AcknowledgementsGrant numbers includedMINOR

Statistical Reporting Standards

methodology_reviewer applies this complete checklist. Adapted from APA 7.0 + Nature portfolio requirements.

Universal Reporting Requirements

CategoryRequirementRed Flag if Missing
Descriptive statisticsM, SD, N (total and per group), range for all continuous variablesHIGH
Effect sizesALL statistical tests must report effect sizes (Cohen's d, η², r, OR, etc.)HIGH — APA 7.0 mandatory
Confidence intervals95% CI for all effect sizes and key estimates; format: 95% CI [lower, upper]MEDIUM
Exact p-valuesReport as p = .032 (not p < .05); use p < .001 only when very smallLOW
Statistical powerA priori power analysis with target power ≥ .80, effect size source statedMEDIUM
Assumption testingNormality, homogeneity of variance, linearity, independence — all testedMEDIUM
Multiple comparisonsBonferroni/Holm/FDR correction when multiple tests performedHIGH
Missing dataAmount, proportion, mechanism (MCAR/MAR/MNAR), handling methodMEDIUM

Effect Size Quick Reference

AnalysisEffect Size MetricSmall / Medium / Large
t-testCohen's d0.2 / 0.5 / 0.8
ANOVApartial η².01 / .06 / .14
Correlationr.10 / .30 / .50
RegressionR², f²f²: .02 / .15 / .35
Chi-squareCramer's V.10 / .30 / .50
Odds ratioOR1.5 / 2.5 / 4.3

Statistical Red Flags (Trigger MAJOR or CRITICAL)

Red FlagSeverityDescription
No effect sizes reportedHIGHConclusions rest solely on p-values — APA 7.0 violation
p-hacking indicatorsHIGHMultiple p near .05; selective reporting; flexible stopping rules
HARKingHIGHHypotheses appear constructed post-hoc to match results
Uncorrected multiple comparisonsHIGH3+ group comparisons via multiple t-tests instead of ANOVA
No power analysisMEDIUMSample size lacks a priori justification
Assumption violations untreatedHIGHViolations reported but original analysis retained unchanged
p = .000LOWRaw software output; must be p < .001
df inconsistent with NHIGHStrong indicator of data reporting error
Causal language, correlational designMEDIUMSurvey/observational studies claiming causation
SEM N < 200 without correctionMEDIUMSmall sample SEM without robust ML or bootstrapping
VIF > 10 unaddressedHIGHSevere multicollinearity ignored
p-values only near .05 clusterHIGHDistribution anomaly suggesting selective reporting

APA 7.0 Format Verification

  • Statistics that cannot exceed 1.0 (no leading zero): r, p, η², R², beta, Cramer's V
  • Statistics that can exceed 1.0 (with leading zero): M, SD, B, Cohen's d, t, F, χ²
  • Italicize: M, SD, N, n, t, F, p, r, R, d, B, beta, χ²
  • Do NOT italicize: df, SS, MS, OR, CI, VIF, AIC, BIC, CFI, RMSEA, ICC, ANOVA, SEM

IRON RULES

  1. READ-ONLY CONSTRAINT: Reviewers MUST NOT modify the manuscript. All output is separate review documents. If any reviewer agent attempts to edit the paper file → STOP and redirect to report generation only.
  2. REVIEWER INDEPENDENCE: 5 reviewers work in Phase 1 without seeing each other's reports. No cross-referencing during independent review phases.
  3. NO FABRICATION: editorial_synthesizer cannot invent critique not present in Phase 1 reports. Every synthesis point traces to a specific reviewer report.
  4. DA CRITICAL BLOCKS ACCEPT: If devils_advocate raises any CRITICAL finding → Editorial Decision cannot be Accept or Minor Revision. Decision must be Major Revision or Reject, with DA findings prominently flagged.
  5. NO SYCOPHANCY: Scores must be evidence-based. A paper with methodology gaps cannot score > 6/10 on rigor. Reviewer pushback does not automatically improve scores.
  6. SPECIFICITY REQUIRED: Every criticism must state: what is wrong, where it appears (section/paragraph), and a concrete suggested fix. Vague feedback ("the methodology could be stronger") is not acceptable.
  7. JOURNAL CALIBRATION MANDATORY: Reviews must reference loaded journal specs. A paper that passes for Nature Communications may not pass for Nature Materials — calibrate thresholds accordingly.

Output Formats

Full Mode Output

# Peer Review Simulation Report
**Manuscript:** [Title]
**Target Journal:** [Journal Name] (IF: X.X, acceptance rate: ~X%)
**Review Date:** [Date]
**Mode:** Full Review

---

## Reviewer Configuration Card
[field_analyst output: 5 reviewer personas with identities and focus areas]

---

## EIC Review (Editor-in-Chief)
**Reviewer:** [Configured EIC persona]
**Recommendation:** [Accept / Minor Revision / Major Revision / Reject]

### Journal Fit Assessment
[Scope alignment, novelty for target journal tier]

### Strengths
[3–5 specific strengths]

### Required Changes
**CRITICAL:**
- [Issue: description | Location | Suggested fix]

**MAJOR:**
- [Issue: description | Location | Suggested fix]

**MINOR:**
- [Issue: description | Location]

### Nature Compliance Checklist
- Data Availability Statement: [✅ Present / ⚠️ Generic / ❌ Missing]
- Code Availability Statement: [✅ / ⚠️ / ❌ / N/A]
- CRediT Contributions: [✅ / ⚠️ / ❌]
- Competing Interests: [✅ / ⚠️ / ❌]
- EQUATOR Guideline: [✅ / N/A / ❌]
- Ethics Approval: [✅ / N/A / ❌]

---

## Methodology Review (Peer Reviewer 1)
**Reviewer:** [Configured methodology reviewer persona]
**Recommendation:** [Accept / Minor / Major / Reject]

### Statistical Reporting Assessment
[Score: X/100 | Level: Exemplary/Adequate/Needs Improvement/Inadequate/Unacceptable]

### Red Flags Detected
[List from §Statistical Red Flags, with severity]

### Reproducibility Assessment
[Methods detail, software versions, data/code availability]

### Required Changes
[CRITICAL / MAJOR / MINOR list]

---

## Domain Review (Peer Reviewer 2)
[Literature coverage, theory, field contribution]
[CRITICAL / MAJOR / MINOR list]

---

## Perspective Review (Peer Reviewer 3)
[Cross-disciplinary angles, broader impact]
[CRITICAL / MAJOR / MINOR list]

---

## Devil's Advocate Review
**Reviewer:** [Configured DA persona]

### Strongest Counter-Argument
[200–300 words: the best possible case AGAINST the paper's conclusions]

### Issue List

#### CRITICAL
| # | Dimension | Issue Description | Location |
|---|-----------|-------------------|----------|

#### MAJOR
| # | Dimension | Issue Description | Location |
|---|-----------|-------------------|----------|

#### MINOR
| # | Dimension | Issue Description | Location |
|---|-----------|-------------------|----------|

### Ignored Alternative Explanations
1. [Alternative A: why it might be more parsimonious than authors' explanation]
2. [Alternative B: ...]

### Missing Stakeholder Perspectives
- [Perspective 1]
- [Perspective 2]

### Unexamined Premise (if detected)
[Unstated assumption underlying the entire paper]

### Observations (Non-Defects)
- [Observation 1]

---

## Editorial Decision

### Summary of Reviewer Consensus
[What all/most reviewers agreed on]

### Points of Divergence
[Where reviewers disagreed + editorial arbitration]

### ⚠️ Devil's Advocate CRITICAL Issues
[Flagged prominently — these block an Accept decision]

### Editorial Decision: [Accept / Minor Revision / Major Revision / Reject]

**Decision Rationale:**
[Evidence-based reasoning citing specific reviewer reports]

---

## Revision Roadmap

Priority 1 — Must Address (CRITICAL):
- [ ] [Issue description] | Reviewer: [X] | Section: [Y]

Priority 2 — Should Address (MAJOR):
- [ ] [Issue description] | Reviewer: [X] | Section: [Y]

Priority 3 — Consider Addressing (MINOR):
- [ ] [Issue description] | Reviewer: [X] | Section: [Y]

**Estimated revision effort:** [X weeks | new experiments: Y/N | new analyses: Y/N]

Quick Mode Output

# Quick EIC Assessment
**Target Journal:** [Journal] | **Mode:** Quick (~15 min)

## Top Issues (EIC Perspective)
1. [Most critical issue]
2. [Second issue]
...10. [Tenth issue]

## Nature Compliance Gaps
[Quick checklist from EIC compliance check]

## Preliminary Recommendation
[Accept / Minor / Major / Reject] — based on EIC review only
*Note: Full simulation (full mode) recommended before actual submission.*

Re-Review Mode Output (R&R Traceability Matrix)

# Re-Review / Revision Verification Report
**Mode:** Re-Review | **Original Decision:** [Decision]

## R&R Traceability Matrix

| # | Original Concern | Reviewer | Claimed Fix (Author) | Verified? | Status |
|---|-----------------|----------|---------------------|-----------|--------|
| 1 | [issue] | R1 | [author's response] | ✅/⚠️/❌ | Resolved/Partial/Unresolved |
| 2 | ... | | | | |

## Residual Issues
[Issues from original review not adequately addressed]

## New Issues Introduced
[Problems created by the revision that weren't in the original paper]

## New Editorial Decision
[Accept / Minor Revision / Major Revision / Reject]

Integration

Pipeline Position

nature-paper-hub Stage 6 (Pre-Submission Audit)
    → nature-integrity-check (pre-review mode)   ← run first
    → nature-reviewer-sim (full mode)             ← this skill
    → nature-paper-hub Stage 3 (revisions)
    → nature-reviewer-sim (re-review mode)        ← verify revisions
    → nature-integrity-check (final-check mode)
    → nature-paper-hub Stage 7 (export)

Upstream: nature-paper-hub → nature-reviewer-sim

  • Input: completed manuscript (from Stage 2–5 drafting)
  • Prerequisite: nature-integrity-check in pre-review mode must PASS before review simulation begins
  • ⚠️ If integrity check not passed: remind user and decline to proceed until resolved

Downstream: nature-reviewer-sim → nature-paper-hub Stage 8

  • The Revision Roadmap format is directly compatible with Stage 8 rebuttal input
  • Pass the entire Revision Roadmap to Stage 8 for point-by-point response drafting
  • For re-review after revisions: use re-review mode

Downstream: nature-reviewer-sim → nature-integrity-check (final)

  • After re-review passes: trigger nature-integrity-check in final-check mode
  • Only after final integrity check PASS → proceed to Stage 7 export

Anti-Patterns

#Anti-PatternWhy It FailsCorrect Behavior
1Fabricating review commentsSynthesizer invents critique not in any Phase 1 reportEvery synthesis point traces to a specific reviewer report
2Duplicate criticisms across reviewersR1/R2/R3 raise identical points = fake diversityEach reviewer has a distinct non-overlapping angle
3Ignoring DA CRITICAL findingsDecision says Accept despite DA flagging critical issuesDA CRITICAL → Decision cannot be Accept (IRON RULE)
4Rubber-stamp re-reviewRe-review says "all addressed" without verificationEach concern independently verified against revised manuscript
5Sycophantic score inflationGiving 8/10 to mediocre work to avoid conflictScores evidence-based; methodology gaps cannot score > 6 on rigor
6Editing the manuscriptReviewer "helpfully" fixes the paperREAD-ONLY: produce reports, never modify the paper
7Generic feedback"The methodology could be stronger"Every criticism: what's wrong + where it is + proposed fix
8Skipping journal calibrationReviewing against generic standardsMust load journal-specs.json and calibrate thresholds to target journal

Version Info

ItemValue
Version1.0
Created2026-05-20
Parent Skillnature-paper-hub v1.0.0
Based Onacademic-paper-reviewer v1.7
RoleNature-series pre-submission peer review simulator

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.