Run Parallel Referee Reviews

Execute reviews from all referee personas simultaneously using subagents.

Your Task

1. Identify what to review: Find the paper, documentation, and code in the current directory

2. Load referee personas: Use the referee personas previously generated (or defaults if none exist)

3. Launch parallel reviews: Spawn each referee as a subagent using the Task tool:

   • Each referee reviews independently
   • Each produces a structured report
   • Reviews run in parallel for efficiency

4. Collect and synthesize: After all reviews complete:

   • Compile all feedback
   • Identify common themes
   • Highlight critical issues vs. minor suggestions
   • Note any conflicting recommendations

5. Generate editorial summary: Produce a summary like a journal editor would:

   • Overall recommendation (Accept / Minor Revisions / Major Revisions / Reject)
   • Key issues that must be addressed
   • Suggestions for improvement
   • Praise for strengths

Default Referee Types (if not specified)

If no referee personas have been generated, use these defaults:

• methodology: Technical/methods expert
• domain: Application domain specialist
• practitioner: End-user perspective
• reproducibility: Code/data/replication focus
• stylistic: Writing style and presentation expert
• reference-validator: Citation accuracy checker (uses essential:reference-validator agent to verify all citations support their claims)

Referees to Run

{{ referees | default: "all" }}

Output Format

# Peer Review Summary

## Editorial Decision: [DECISION]

## Referee Reports

### Referee 1: [Title]
**Recommendation**: Accept / Minor / Major / Reject
**Summary**:
**Major Issues**:
**Minor Issues**:
**Strengths**:

### Referee 2: [Title]
...

## Synthesis

### Critical Issues (Must Address)
1.
2.

### Suggested Improvements
1.
2.

### Noted Strengths
1.
2.

## Next Steps
To address feedback: [specific guidance]
To re-run review after changes: `/run-review`