golden-dataset-curation

What problem does it solve?

This Skill automates the quality assessment and multi-agent curation of documents for a golden dataset, enabling scalable, auditable data curation.

Core Features & Use Cases

• Content-type classification: automatically categorize incoming documents (article, tutorial, research_paper, etc.) to guide curation.
• Quality assessment & consensus: run parallel agent evaluations on accuracy, coherence, depth, and relevance, then aggregate into a final decision.
• Test query generation: produce retrieval test queries to validate coverage and retrieval performance.
• End-to-end workflow: ingest new documents, run multi-agent quality checks, tag domains, and commit publish-ready entries when thresholds are met.

Quick Start

Provide a new document URL or content snippet to seed the golden-dataset-curation workflow, then trigger the curator to classify, evaluate, and generate a consensus-based inclusion decision. Use the Langfuse prompts and consensus outputs to guide the inclusion or review of the document.