extraction-skill

What problem does it solve?

Extracting readable text from PDFs and EPUBs is often slow, unreliable, or requires manual processing; this skill provides a fast, deterministic way to obtain representative text samples for downstream classification and analysis, while gracefully handling errors and timeouts.

Core Features & Use Cases

• Efficient text sampling from PDFs (first 5 pages) and last page, and from EPUBs (first chapter + TOC), limited to 10,000 characters for quick analysis.
• Robust fallback: uses system utilities first (pdftotext, pdfinfo) and falls back to Python libraries (pdfplumber, ebooklib) when needed.
• Timeouts and error logging to prevent pipeline hangs and enable manual review of failures.
• Integrates with discovery-skill and downstream analysis-skill for end-to-end PARA/Enneagram classification.

Quick Start

Run the extraction-skill on a batch of PDFs/EPUBs discovered by the discovery-skill to produce text samples in extractions/ for downstream analysis.