Operational Artifacts Supporting the WorldWideWatchers Bounded-Agentic AI Architecture for Automated Open-Source Intelligence Synthesis

This repository contains representative operational artifacts generated during pilot deployments (2024–2026) of the WorldWideWatchers platform. WorldWideWatchers is a governance-aligned, bounded-agentic AI architecture for automated Open-Source Intelligence (OSINT) synthesis, integrating data ingestion, NLP-based enrichment, and LLM-assisted intelligence synthesis under bounded orchestration. This dataset comprises evaluation materials from a 25-day operational deployment (98 pipeline runs, May 25–June 18, 2026). The artifacts substantiate the empirical findings reported in the associated manuscript. Files: evaluation_pipeline_runs.csv (98-run telemetry), evaluation_nli_scores.csv (NLI faithfulness scores), evaluation_governance_outcomes.csv (governance decisions), evaluation_category_summary.csv (per-category aggregates), paper_metrics.json (summary statistics), zenodo_README.txt (schema documentation). Raw ingestion corpora not included due to copyright. All data derive from publicly available RSS and News API sources.