figure-comparison

Compares paper reference figures, tables, and values with reproduced results.

18|2|Updated Feb 15, 2026
One-click install
npx skills add https://github.com/LightconeResearch/lightcone-cli --skill figure-comparison
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: figure-comparison
Source: https://github.com/LightconeResearch/lightcone-cli/tree/main/claude/lightcone/skills/figure-comparison
Command: npx skills add https://github.com/LightconeResearch/lightcone-cli --skill figure-comparison

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It helps you quickly verify whether an ASTRA reproduction matched the paper’s reported figures, tables, and numerical claims by producing a single portable HTML comparison.

Core Features & Use Cases

  • Side-by-side visual comparison: Places paper reference artifacts from work/reference on the left and reproduced results from results/<universe> on the right, flagging missing counterparts.
  • Scope-aware matching: Uses comparison-report.yaml or targets/targets.md when available, otherwise builds a best-effort inventory from astra.yaml and work/reference artifacts.
  • Static, non-computational value checking: Displays paper vs. project values and performs only a simple ±1σ tolerance status check without re-deriving numbers.
  • Portable, self-contained output: Embeds images as base64 (PDFs rasterized to PNG) and writes a single HTML file to .lightcone/comparison.html for easy sharing.

Quick Start

Run the skill from the project folder that contains astra.yaml and point it at the paper reference directory by telling the assistant the argument path, e.g. the folder that contains work/reference/.

Frequently Asked Questions about figure-comparison

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a side-by-side comparison report for reproduced research results?

To generate a side-by-side comparison report, the Skill reads astra.yaml alongside optional comparison-report.yaml or targets.md to match reference figures and tables against reproduced results, outputting a single portable HTML file to .lightcone/comparison.html.

Can I verify numerical claims from a paper against my reproduction without re-deriving the numbers?

Yes, you can verify numerical claims through static, non-computational value checking that displays paper versus project values and applies a simple ±1σ tolerance status check without re-deriving the underlying calculations.

Does the comparison report support embedding PDF figures or does it require image formats?

The comparison report supports PDF references by automatically rasterizing them to PNG format, then embedding those images as base64 directly into the self-contained HTML output for portable sharing.

What is the best way to scope which figures and tables are included in a paper reproduction verification?

Scoping is handled by providing a comparison-report.yaml or targets/targets.md file to define specific artifacts, otherwise the Skill builds a best-effort inventory from astra.yaml and available work/reference files.

How do I start a paper reproduction audit from a project folder containing reference materials?

Run the Skill from the project folder containing astra.yaml and point it at the paper reference directory, such as work/reference/, to automatically inventory artifacts and emit the comparison HTML report.

What happens if a reproduced result is missing a counterpart figure or table from the paper?

When a reproduced result lacks a counterpart, the side-by-side visual comparison surfaces the mismatch by displaying the paper reference artifact alongside a flagged missing artifact notification in the HTML report.