What problem does it solve? Investigating people and companies across fragmented public records — corporate filings, government contracts, lobbying disclosures, sanctions lists, property deeds, and court records — requires manually querying a dozen disconnected databases and reconciling inconsistent entity names. This Skill automates acquisition, entity resolution, and evidence-chain construction across 11 public-records sources. ## Core Features & Use Cases - Multi-source data acquisition: Fetch normalized CSVs from SEC EDGAR, USAspending, Senate lobbying disclosures, OFAC SDN sanctions, ICIJ Offshore Leaks, NYC ACRIS property records, OpenCorporates, CourtListener, Wayback Machine, Wikipedia/Wikidata, and GDELT news — Python stdlib only, most sources need no API key. - Entity resolution with confidence tiers: Match names across sources using exact, fuzzy (word-bag), and token-overlap tiers, each labeled high/medium/low confidence so leads are never confused with confirmed facts. - Statistical timing analysis: Run permutation tests to detect whether events (e.g., lobbying filings) cluster suspiciously near contract awards, then compile everything into a findings.json where every claim traces to a specific source row. - Use Case: A journalist investigating pay-to-play patterns fetches a company's Senate lobbying filings and USAspending contract awards, resolves entity name variants between the two datasets, tests timing correlation, and produces a verifiable evidence chain for the story. ## Quick Start Ask the agent to investigate a company by fetching its SEC EDGAR filings and USAspending contracts, then cross-link the results and build a findings report.
