Peer Reviewer Skill

AI-powered academic rigor for your research.

Peer Reviewer is an OpenClaw skill that simulates a multi-agent academic peer review process. It employs a "Council of Agents" to deconstruct arguments, find contradictions in established literature, and render a final judgment on the merit of a scientific paper or claim.

🧠 How It Works

The system uses three specialized AI agents:

1. The Deconstructor: Parses raw text into a formal Logic Graph (Toulmin Model), extracting claims, premises, and evidence without judgment.
2. The Devil's Advocate: Takes the claims and actively searches for contradictions (using Google Serper & ArXiv). It looks for theoretical conflicts, empirical contradictions, and prior art.
3. The Judge: Evaluates the Logic Graph against the Devil's Advocate's objections. It scores the paper on logical coherence, foundational integrity, and empirical falsifiability, ignoring "consensus" in favor of strict logical validity.

🚀 Installation

npm install @sschepis/peer-reviewer

Note: This package is designed to be used as an OpenClaw skill but can also run standalone.

🛠️ Usage

As an OpenClaw Skill

Trigger the skill by asking OpenClaw:

"Review this paper for flaws." "Analyze the logic of this argument."

(Ensure the skill is loaded in your OpenClaw configuration).

Standalone CLI

You can run the reviewer directly on a text file or a raw string:

# Review a file
node dist/index.js "/path/to/paper.txt"

# Review a raw claim
node dist/index.js "Claim: We can exceed the speed of light by..."

⚙️ Configuration

The reviewer requires access to LLM and Search providers. Create a .env file or set environment variables:

# Google Vertex AI (Gemini)
GOOGLE_APPLICATION_CREDENTIALS="path/to/google.json"
VERTEX_AI_MODEL="gemini-3-pro-preview"
VERTEX_AI_LOCATION="us-central1"

# Search Provider (Google Serper)
SERPER_API_KEY="your_serper_key"

📊 Output: The Merit Report

The tool generates a JSON report containing:

• Overall Score (0-10): A weighted metric of logical and empirical strength.
• Defense Strategy: How the author might address the strongest objections.
• Suggestions: Concrete steps to improve the paper's rigor.
• Dimensions: Detailed scores for Logical Coherence, Causal Robustness, etc.

License

MIT