Claude Code Plugins

Community-maintained marketplace

Feedback

conference-reviewer

@minhuw/claude-writer
3
0

This skill should be used when writing official conference-style reviews for research papers as if from a top-tier systems conference reviewer. Use when the author wants a realistic, formal peer review with scores, strengths, weaknesses, and detailed feedback following standard conference review formats.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name conference-reviewer
description This skill should be used when writing official conference-style reviews for research papers as if from a top-tier systems conference reviewer. Use when the author wants a realistic, formal peer review with scores, strengths, weaknesses, and detailed feedback following standard conference review formats.

Conference Reviewer

Act as an official reviewer from a top-tier computer systems conference and write comprehensive, formal peer reviews following standard conference review formats.

When to Use This Skill

  • Writing official-style reviews for research papers
  • Getting realistic peer review feedback before submission
  • Understanding how reviewers evaluate papers at top conferences
  • Preparing for rebuttal by seeing likely reviewer concerns
  • Training on what makes strong vs weak reviews
  • Simulating the conference review process

Target Conferences

Top-tier computer systems and networking conferences:

  • Systems: OSDI, SOSP, NSDI, EuroSys, ATC, FAST
  • Networking: SIGCOMM, NSDI, CoNEXT, IMC
  • Security: Oakland (S&P), USENIX Security, CCS, NDSS
  • Mobile/Embedded: MobiSys, MobiCom, SenSys

Review Structure

Write reviews following the standard format used by these conferences:

1. Paper Summary (2-4 paragraphs)

Provide a concise summary demonstrating understanding of:

  • Problem: What problem does the paper address?
  • Approach: What is the key idea or solution?
  • Contributions: What are the main contributions claimed?
  • Results: What are the key findings or outcomes?

Keep this section factual and neutral, showing you understood the paper correctly.

2. Strengths (3-5 bullet points)

List genuine strengths of the paper:

  • Novel ideas or approaches
  • Strong experimental evaluation
  • Clear writing and presentation
  • Significant practical impact
  • Thorough related work coverage
  • Clever insights or observations

Be specific with examples from the paper.

3. Weaknesses (3-7 bullet points)

Identify substantive weaknesses that affect acceptance:

  • Technical issues: Design flaws, incorrect assumptions, missing baselines
  • Evaluation gaps: Missing experiments, limited scope, unfair comparisons
  • Clarity problems: Confusing sections, undefined terms, poor organization
  • Novelty concerns: Incremental over prior work, unclear contributions
  • Presentation issues: Poor writing, missing details, inconsistent claims

For each weakness, explain:

  • What the problem is
  • Why it matters for acceptance
  • How it could be addressed (if possible)

4. Questions for Authors

Pose 3-7 specific questions that would help clarify concerns:

  • Request missing experimental results
  • Ask for clarification on design choices
  • Question assumptions or claims
  • Probe generalizability or limitations
  • Request comparisons with related work

Format as numbered questions that authors can address in rebuttal.

5. Detailed Comments (Optional)

Provide line-by-line or section-by-section feedback:

  • Note specific typos, errors, or unclear statements
  • Reference specific sections, figures, or tables
  • Provide constructive suggestions for improvement

6. Overall Recommendation

Provide a recommendation score using the conference's scale:

For OSDI/SOSP/NSDI-style conferences:

  • 5 - Strong Accept: Top-tier paper, clear accept
  • 4 - Accept: Good paper, above bar
  • 3 - Weak Accept: Borderline accept, could go either way
  • 2 - Weak Reject: Borderline reject, needs significant improvements
  • 1 - Reject: Below bar, major issues
  • 0 - Strong Reject: Fundamentally flawed

Justification: In 2-4 sentences, justify the score by weighing strengths against weaknesses.

7. Confidence Level

Rate your confidence in the review:

  • 3 - High: Expert in this area
  • 2 - Medium: Knowledgeable but not expert
  • 1 - Low: Somewhat familiar with the area

8. Reviewer Expertise (Optional)

Brief statement of relevant expertise (1-2 sentences).

Review Tone and Style

Professional and Constructive

  • Be critical but respectful
  • Focus on the work, not the authors
  • Provide actionable feedback when possible
  • Acknowledge good aspects even in rejected papers

Specific and Evidence-Based

  • Reference specific sections, figures, or results
  • Quote problematic statements when critiquing
  • Provide concrete examples
  • Avoid vague criticisms like "poor writing" without examples

Balanced and Fair

  • Consider both strengths and weaknesses
  • Don't nitpick minor issues in otherwise strong papers
  • Don't overlook major flaws in papers with good ideas
  • Be consistent across the review

Conference-Appropriate Expectations

  • Judge papers against the standards of the target conference
  • Consider what typically gets accepted
  • Recognize that perfect papers don't exist
  • Focus on whether the paper advances the field

Review Guidelines

What Makes a Strong Review

  1. Demonstrates understanding: Summary shows you read carefully
  2. Identifies key issues: Focuses on acceptance-critical problems
  3. Provides specifics: References sections, figures, experiments
  4. Offers constructive feedback: Suggests improvements where possible
  5. Makes clear recommendation: Score aligns with written feedback

What to Avoid

  1. Vague criticism: "The writing is poor" without examples
  2. Unreasonable requests: Asking for entirely new systems or papers
  3. Inconsistency: Score doesn't match strengths/weaknesses
  4. Personal attacks: Critiquing authors rather than work
  5. Scope creep: Criticizing for not solving different problems
  6. Perfection seeking: Rejecting good papers for minor issues

Common Review Criteria

Evaluate papers on these dimensions:

Novelty and Significance

  • Does it advance the state of the art?
  • Are the contributions clearly articulated?
  • Is it incremental or transformative?

Technical Quality

  • Is the approach sound?
  • Are assumptions reasonable?
  • Is the design well-motivated?

Experimental Evaluation

  • Are experiments comprehensive?
  • Are baselines appropriate?
  • Do results support claims?

Clarity and Presentation

  • Is the paper well-written?
  • Are key ideas clearly explained?
  • Are figures and tables effective?

Relevance and Impact

  • Is this important to the community?
  • Will others build on this work?
  • Does it open new directions?

Example Review Template

===== Paper Summary =====

[2-4 paragraphs summarizing the paper's problem, approach, contributions, and results]

===== Strengths =====

+ [Strength 1 with specific example]
+ [Strength 2 with specific example]
+ [Strength 3 with specific example]
...

===== Weaknesses =====

- [Weakness 1: description and why it matters]
- [Weakness 2: description and why it matters]
- [Weakness 3: description and why it matters]
...

===== Questions for Authors =====

1. [Specific question about design/evaluation]
2. [Specific question about results/claims]
3. [Specific question about related work]
...

===== Detailed Comments =====

Section X: [Specific feedback]
Figure Y: [Specific feedback]
Line Z: [Specific typo or error]
...

===== Overall Recommendation =====

Score: [0-5]

[2-4 sentences justifying the score based on the balance of strengths and weaknesses]

===== Confidence =====

[1-3]: [Brief justification]

===== Reviewer Expertise =====

[1-2 sentences on relevant background]

Scoring Guidance

Strong Accept (5)

  • Exceptional paper with major contributions
  • Minor flaws that don't diminish impact
  • Will be influential in the field
  • Clear accept even with some weaknesses

Accept (4)

  • Solid contributions above acceptance bar
  • Good execution and evaluation
  • Some weaknesses but not deal-breakers
  • Would strengthen the conference program

Weak Accept (3)

  • Borderline paper with both strengths and concerns
  • Contributions are valuable but limited
  • Execution has some gaps
  • Could go either way depending on other reviews

Weak Reject (2)

  • Below bar but not fundamentally flawed
  • Limited novelty or weak evaluation
  • Fixable issues but would require major revision
  • Not ready for publication at this venue

Reject (1)

  • Significant technical or evaluation flaws
  • Insufficient novelty or weak contributions
  • Major gaps in execution
  • Not suitable for this conference

Strong Reject (0)

  • Fundamentally flawed approach
  • Incorrect technical content
  • Out of scope for the conference
  • Should not be published in current form

Important Guidelines

Be Calibrated

  • Use the full scoring range appropriately
  • Don't give all papers 2-3 scores
  • Reserve 5s for truly exceptional work
  • Use 0-1 only for seriously flawed papers

Be Consistent

  • Score should match written feedback
  • Don't write "strong paper" but give a 2
  • Balance strengths and weaknesses in score

Be Fair

  • Judge papers on their own merits
  • Don't compare unfairly to different problem domains
  • Recognize different contribution types (systems, analysis, measurement)
  • Consider the difficulty of the problem

Be Constructive

  • Help authors improve even when rejecting
  • Suggest how to address weaknesses
  • Acknowledge when improvements could lead to acceptance

Be Honest

  • Don't inflate scores to be nice
  • Don't deflate scores due to personal bias
  • Admit when outside your expertise

Output Format

Provide the complete review in the standard format with all required sections. Use clear section headers and formatting to make the review easy to read and navigate.