Run2 multi format extraction

Skill cxcscmu/SkillLearnBench/skills/b2-self-feedback-claude-haiku-4-5/organize-messy-files/run2_multi-format-extraction

[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.

Install

npx -y skills add cxcscmu/SkillLearnBench --skill run2_multi-format-extraction

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Unified multi-format extraction for PDF, DOCX, and PPTX with consistent output and quality metrics

SKILL.md

3.4 KB, as published. Nobody here has run it

Multi-Format Document Extraction Skill (Improved)

Overview

Unified framework for extracting text from multiple document formats with quality metrics and consistent output. Handles edge cases and provides confidence scores for extraction quality.

Installation

pip install pdfplumber python-docx python-pptx PyPDF2 -q

Unified Extraction Framework

from pathlib import Path
from dataclasses import dataclass
import pdfplumber
from docx import Document
from pptx import Presentation

@dataclass
class ExtractionResult:
    """Container for extraction results with quality metrics"""
    text: str
    format: str
    length: int
    quality_score: float  # 0.0-1.0
    success: bool

def extract_document_unified(file_path, max_content_units=5):
    """
    Unified extraction for PDF, DOCX, PPTX.
    Returns ExtractionResult with quality score.
    """
    file_path = Path(file_path)
    ext = file_path.suffix.lower()

    if ext == '.pdf':
        return extract_pdf_robust(file_path, max_content_units)
    elif ext == '.docx':
        return extract_docx_robust(file_path, max_content_units)
    elif ext == '.pptx':
        return extract_pptx_robust(file_path, max_content_units)
    else:
        return ExtractionResult("", "unknown", 0, 0.0, False)

def extract_pdf_robust(pdf_path, max_pages=5):
    """Extract from PDF with quality scoring"""
    text = ""
    try:
        with pdfplumber.open(pdf_path) as pdf:
            for page in pdf.pages[:max_pages]:
                extracted = page.extract_text()
                if extracted:
                    text += extracted + " "
    except Exception as e:
        pass

    quality = min(1.0, len(text) / 1000.0)  # 1000 chars = perfect quality
    return ExtractionResult(text, "pdf", len(text), quality, len(text) > 50)

def extract_docx_robust(docx_path, max_paragraphs=30):
    """Extract from DOCX with quality scoring"""
    text = ""
    try:
        doc = Document(docx_path)
        for para in doc.paragraphs[:max_paragraphs]:
            if para.text.strip():
                text += para.text + " "
    except Exception as e:
        pass

    quality = min(1.0, len(text) / 1000.0)
    return ExtractionResult(text, "docx", len(text), quality, len(text) > 50)

def extract_pptx_robust(pptx_path, max_slides=5):
    """Extract from PPTX with quality scoring"""
    text = ""
    try:
        prs = Presentation(pptx_path)
        for slide in prs.slides[:max_slides]:
            for shape in slide.shapes:
                if hasattr(shape, "text") and shape.text.strip():
                    text += shape.text + " "
    except Exception as e:
        pass

    quality = min(1.0, len(text) / 1000.0)
    return ExtractionResult(text, "pptx", len(text), quality, len(text) > 50)

Key Improvements from Run 1

Quality scoring: Numeric confidence in extraction quality
Unified interface: All formats return same ExtractionResult
Better error handling: Graceful degradation
Metrics tracking: Can identify problematic extractions
Reusable dataclass: ExtractionResult for downstream processing

Usage Pattern

result = extract_document_unified(file_path)
if result.quality_score > 0.3:  # Reasonable confidence
    category = classify_text(result.text)
else:
    category = 'music_history'  # Safe default