Dataset Generator Skill

Generate high-quality benchmark evaluation datasets with adjustable difficulty levels from custom PDF documents. Perfect for testing RAG systems, knowledge graphs, and Q&A models.

Usage

Invoke this skill with:

/dataset-generator <pdf_directory> [output_file] [num_questions] [difficulty]

Arguments:

• $1 (required) - Path to PDF directory containing source documents
• $2 (optional) - Output JSON file path (default: benchmark_dataset.json)
• $3 (optional) - Number of questions to generate (default: 20)
• $4 (optional) - Difficulty level: easy, medium, hard, or mixed (default: mixed)

Examples