A powerful workflow for extracting structured data and evidence-backed insights from academic papers with built-in hallucination verification.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install paper-summary-json
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install paper-summary-json using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Paper Analysis Evidence?
The Paper Analysis Evidence skill is an advanced research automation tool for the Openclaw Skills ecosystem, designed to perform deep, structured reading of academic documents. It processes local files or direct URLs to break down complex papers into granular technical components like methodology, datasets, and experimental results.
By leveraging the Openclaw Skills framework, this tool ensures high-fidelity information retrieval by enforcing a strict evidence-based extraction process. Every claim extracted by the agent is mapped back to direct quotes or evidence spans from the original source text, providing a level of transparency and reliability essential for technical and academic work.
Paper Analysis Evidence Use Cases
- Accelerating literature reviews by extracting key metrics and baselines across multiple papers using Openclaw Skills.
- Verifying the consistency of experimental results and identifying potential hallucinations in AI-generated summaries.
- Converting unstructured PDF or DOCX academic papers into structured JSON data for research databases.
- Generating professional, multi-format (Markdown, HTML, DOCX) analysis reports for academic peer review or team knowledge sharing.
How Paper Analysis Evidence Works
- The process begins by preparing the paper inputs, where the skill downloads URLs or copies local files into a specific workspace directory on the Ubuntu desktop.
- The system performs raw text extraction and cleaning, then splits the document into logical sections such as Abstract, Introduction, Method, and Conclusion.
- A structured extraction phase uses specialized prompt templates to generate a JSON object containing the title, task, problem, datasets, and at least 6 direct evidence spans.
- The skill runs a consistency verification step, comparing the extracted JSON against the original text to assign a hallucination risk score and a final quality verdict.
- Finally, the extracted data and verification results are rendered into editable Markdown, visual HTML, and Word-compatible DOCX reports.
Paper Analysis Evidence Setup
To utilize this skill within your Openclaw Skills environment, ensure you have the necessary python scripts available in your workspace.
Prepare your papers using the following command:
python scripts/prepare_papers.py --language 中文 --files /path/to/paper.pdf --urls "https://example.com/paper.pdf"
After the extraction and verification steps are complete, render the final reports by pointing to the generated manifest:
python scripts/render_report.py --manifest ~/Desktop/paper_analysis_results/<YYYYMMDD_HHMMSS>/manifest.json
Paper Analysis Evidence Data Schema & Taxonomy
The skill organizes research data into a structured directory under ~/Desktop/paper_analysis_results/. The key data points captured in the structured_result.json include:
| Field | Type | Description |
|---|---|---|
| title | String | The official title of the academic paper |
| method_core | String | The core technical approach or architecture described |
| main_results | Array | Quantitative data points including datasets, metrics, and values |
| evidence_spans | Array | Direct quotes from the paper used to support the extracted claims |
| verification_result | Object | Evaluation of data accuracy including hallucination_risk and overall_score |
Paper Analysis Evidence Advanced Features
- Evidence-enhanced extraction logic that prioritizes direct quotes to eliminate model hallucinations.
- Built-in verification scoring system (1-5) to audit the technical accuracy of every generated report.
- Multi-format export pipeline supporting Markdown, HTML, and DOCX for seamless integration into various workflows.
- Native support for both Chinese and English output, making it a versatile tool for global Openclaw Skills users.
- Automated section splitting that targets specific paper parts like Ablation Studies and Limitations for deeper analysis.
SKILL.md
Loading
Related Openclaw Skills
Engineering as Marketing
A strategic framework for building free, high-value tools and micro-sites that serve as evergreen customer acquisition channels.
Content and Email Marketing
A strategic framework for building an integrated content and email marketing engine that moves users through the acquisition, activation, retention, and revenue lifecycle.
Aetna Healthcare Industry Analysis
A comprehensive data skill for analyzing Aetna's evolution from a 19th-century insurer to a cornerstone of the CVS Health vertical ecosystem.
Bilibili App Skill
An AI agent skill designed to analyze and research Bilibili, the leading Chinese video platform for ACG culture and Gen Z community engagement.
Bullseye Channel Selection
A systematic framework to identify, rank, and test the most effective customer acquisition channels for your startup.
Business Development Pipeline
A strategic framework for startups to build high-impact partnership pipelines using attribute-based selection and a 9-step negotiation process.
Featured*
