document-figures

What problem does it solve?

It extracts visual figures from PDFs, Office files, and web pages with provenance, and it creates Mermaid-first structural diagrams when suitable figures are missing.

Core Features & Use Cases

• Figure extraction with provenance: Pulls embedded images and/or renders page/slide visuals while preserving source reference information.
• Caption recovery and integrity: Recovers captions and alt-text when possible, explicitly flags reconstructed captions, and outputs a Figure Ledger for auditability.
• Mermaid-first figure creation: Generates deterministic Mermaid diagrams for sequential, decision, state, entity-relationship, and other structural claims using an information test so diagrams aren’t just decorated bullets.
• Ledger chaining for document summarization: Produces artifacts that can chain into document-summary via linked Claim Ledger IDs.

Example use case: You have a Claim Ledger describing evidence-by-figure claims from a PDF report; this Skill extracts the relevant figures with page/slide provenance, recovers captions, filters out decorative noise, and emits FIGURES.md plus manifest.json so downstream summarization can cite each figure.

Quick Start

Use the document-figures skill to extract figures from 'report.pdf' and produce a Figure Ledger that links each extracted figure to the matching claims in your Claim Ledger.