paper-card-analyzer

What This Skill Does

The paper-card-analyzer is a specialized OpenClaw skill designed to transform raw research paper outputs (generated by paper-parse) into highly structured, actionable, and analytical summaries known as "paper cards." Instead of performing a simple surface-level summary, this agent performs a critical reading of the paper's full content and metadata. It systematically breaks down the research into ten predefined core research dimensions—including contributions, methodological novelty, experimental rigor, and reproducibility. By maintaining a strict reliability protocol, the tool ensures that every claim is grounded in the source artifacts, explicitly separating author claims from analytical assessments to prevent hallucination.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/chen-li-17/paper-card-analyzer

Use Cases

• Literature Reviewing: Quickly create standardized summaries across dozens of papers to compare performance metrics and experimental setups consistently.
• Journal Club Preparation: Generate structured talking points that highlight methodological limitations and validity threats before a presentation.
• Rep reproducibility Checks: Use the "Reproducibility Notes" section to quickly determine if the paper provides enough data to warrant an attempt at code reproduction.
• Research Synthesis: Consolidate knowledge across multiple papers within a domain by generating uniform paper cards that allow for easy side-by-side comparison of future work.

Example Prompts

1. "Analyze the paper-parse output in the ./neurips-2023 folder and create a paper-card. Focus specifically on the limitations section."
2. "I've reviewed the first draft of the paper-card for 'Attention is All You Need'. Please refine the 'Core Contributions' section to be more concise and add more detail to the 'Reproducibility Notes'."
3. "Generate a paper-card for the parsed files in my current directory, ensuring that the 'Main Results' section only includes metrics explicitly supported by the tables in the _parsed.json file."

Tips & Limitations

• Data Dependency: The quality of the card is entirely dependent on the quality of the preceding paper-parse execution. Ensure your source files are clean.
• Honest Uncertainty: If the paper is opaque about its methodology, the skill will explicitly state "Not clearly stated in parsed content" rather than guessing. Trust this output as a sign that the paper itself may be underspecified.
• Feedback Loop: Treat the first generated card as a draft. The tool is designed to learn from your iterative feedback; the more you specify where to be stricter, the more accurate the final output will be.
• File Management: Ensure you keep the output files (paper-card.md, paper-card.json, and the feedback log) together to maintain the history of your research analysis.