Manuscript Referee Review

What This Does

Before submitting a paper, run Claude as a critical referee. It will identify weaknesses, raise objections, and suggest improvements — helping you fix issues before actual reviewers see them.

Prerequisites

• Claude Code installed and configured
• A draft manuscript (PDF, Word, or text)

The CLAUDE.md Template

Copy this into a CLAUDE.md file in your research folder:

# Manuscript Review System

## Command
`/review-paper [file]` — Full referee-style review of a manuscript

## Review Protocol

### Role
Claude acts as a thorough, critical academic referee. The goal is NOT to be nice — it's to identify every potential weakness before real reviewers do.

### Review Dimensions

**1. Contribution Assessment**
- Is the contribution clear?
- Is it significant enough?
- How does it advance beyond prior work?

**2. Framing & Positioning**
- Is the research question well-motivated?
- Is the literature review adequate?
- Are claims appropriately scoped?

**3. Methodology**
- Is the method appropriate for the question?
- Are there identification concerns?
- Is there internal validity?
- Is there external validity / generalizability?

**4. Results Interpretation**
- Are conclusions supported by evidence?
- Are alternative explanations considered?
- Is statistical inference appropriate?

**5. Presentation**
- Is the writing clear?
- Is the structure logical?
- Are tables/figures effective?

**6. Technical Issues**
- Any errors in equations/analysis?
- Are robustness checks adequate?
- Is replication possible from the description?

### Referee Report Format

Referee Report: [Paper Title]

Summary

[2-3 sentence summary of the paper]

Overall Assessment

Recommendation: [Accept / Minor Revision / Major Revision / Reject] Main strengths: [2-3 bullets] Main weaknesses: [2-3 bullets]

Major Comments

(Issues that must be addressed)

1. [Comment 1]

   • Concern: [What's wrong]
   • Suggestion: [How to fix]

2. [Comment 2] ...

Minor Comments

(Issues that would improve the paper)

1. [Comment]
2. [Comment] ...

Typos / Small Errors

• Page X, line Y: [correction] ...

Questions for Authors

1. [Question]
2. [Question]

### Severity Calibration

**Major Comment (must address)**:
- Threatens validity of main conclusions
- Missing critical information
- Methodological flaw
- Unsupported claims

**Minor Comment (should address)**:
- Presentation issues
- Missing robustness check
- Unclear explanation
- Missing citation

**Typo/Small (nice to fix)**:
- Grammar
- Formatting
- Number discrepancies

### Common Objections by Type

**Empirical Papers:**
- Selection bias: "How do you account for non-random selection into treatment?"
- Omitted variable: "Have you considered that Z might explain both X and Y?"
- Measurement: "How valid is your proxy for construct X?"
- External validity: "Would this hold in context Y?"

**Theoretical Papers:**
- Assumption scrutiny: "Why is assumption X reasonable?"
- Boundary conditions: "When does the theory not apply?"
- Empirical implications: "What would falsify this theory?"

**Review Papers:**
- Scope: "Why these papers and not others?"
- Synthesis: "What's the contribution beyond summarizing?"
- Gaps: "What's missing from the literature?"

Step-by-Step Setup

Step 1: Add to your CLAUDE.md

Add the manuscript review template to your research folder's CLAUDE.md.

Step 2: Prepare your manuscript

Save your paper as a PDF, Word doc, or plain text file that Claude can read.

Step 3: Run the review

/review-paper paper-draft.pdf

Or:

Review my paper manuscript at [file path]. Act as a critical referee. Don't be nice — I need to find the weaknesses.

Example Usage

User:

"/review-paper working-paper-v3.pdf"

Claude:

Referee Report: "Remote Work and Innovation: Evidence from Patent Data"

Summary

The paper examines whether firms' adoption of remote work policies affects their innovation output, measured by patents. Using a diff-in-diff design around COVID-19, the authors find a 12% increase in patents for firms that went remote, driven by increased inventor collaboration across geographies.

Overall Assessment

Recommendation: Major Revision

Strengths:

• Important, timely question
• Creative use of patent data
• Compelling mechanism story (geographic collaboration)

Weaknesses:

• Identification concerns with COVID timing
• Patent quality not addressed
• Selection into remote work not random

Major Comments

1. COVID Confounds

• Concern: COVID affected innovation through many channels beyond remote work: government R&D funding surged, demand shifted, supply chains disrupted. Any of these could explain patent changes.
• Suggestion: Identify pure remote work variation (e.g., firms that went remote before COVID vs during). Or, include controls for COVID-affected industries and test for pre-trends.

2. Patent Quantity vs Quality

• Concern: The 12% increase is in patent counts. If remote work enables filing more incremental patents while reducing breakthrough innovations, the net effect on innovation is ambiguous.
• Suggestion: Examine patent quality (forward citations, claims breadth, or technology class novelty). Show the effect isn't driven by low-quality patents.

3. Selection into Remote Work

• Concern: Firms that adopted remote work may be systematically different (more tech-savvy, more flexible culture). These traits might independently predict higher innovation.
• Suggestion: Use instrumental variable (e.g., pre-pandemic office lease expiration) or show balance on observables between remote and non-remote firms.

Minor Comments

1. The literature review omits Atkin et al. (2022) on remote work and productivity — directly relevant.

2. Table 3 shows results for all firms together. Consider breaking out by industry — the effect may be concentrated in software.

3. The mechanism analysis (geographic collaboration) is compelling but could use visualization. A figure showing collaboration network changes would strengthen this.

4. Abstract claims "causal effect" but the identification isn't bulletproof. Soften to "relationship" or address endogeneity more thoroughly.

Typos

• Page 7, line 14: "effet" → "effect"
• Table 2: Column headers misaligned
• Page 15: Citation format inconsistent (APA vs Chicago)

Questions for Authors

1. Can you show pre-trends in patent activity for treatment vs control groups?
2. What percentage of your sample had explicit remote work policies before COVID?
3. Do results hold if you exclude Big Tech firms (which may drive the average)?

Tips

• Ask for specific focus: "Review the methodology section with particular attention to identification" gets more targeted feedback.
• Request multiple passes: "First review for big-picture issues, then a second pass for presentation" ensures comprehensive coverage.
• Ask for devil's advocate: "What's the strongest objection a hostile reviewer could make?"
• Compare to standards: "How does this compare to recently published papers in [target journal]?"

Troubleshooting

Problem: Review is too gentle

Solution: Explicitly request critical review: "Be harsh. Pretend you're a reviewer looking for reasons to reject."

Problem: Review misses domain-specific issues

Solution: Tell Claude the field: "Review this as an economics paper. Watch for identification issues, econometric concerns, and policy relevance."

Problem: Review catches things but not the real weaknesses

Solution: Add context: "The weakness I'm most worried about is X. Is it really a problem? What else might I be missing?"