Проверка конфиденциальных данных перед выгрузкой

security

disclosure-check is a Claude Code skill that pre-screens analysis outputs — tables, figures, and logs — built on restricted or confidential data for statistical-disclosure-limitation problems before any release from a secure enclave or RDC. It scans for small cell counts (default n<10), complementary-suppression gaps, p-percent and (n,k) dominance failures, re-identifying exact counts, PII leakage, and unrounded sensitive statistics, classifying each finding as CRITICAL, WARNING, or OK and gating on any CRITICAL. Provider profiles for census, irs, irb, and generic are supported, with thresholds overridable via --threshold. Researchers working with Census FSRDC, IRS SOI, administrative registers, or linked health records can use this skill to catch disclosure problems on their own machine before submitting an export request or depositing results to openICPSR or a journal.