# Public Statistics Suppression Tool Local aggregate-count disclosure workflow. Node 22+, no packages/account or other purchase. `node app.mjs`, open printed full session URL, load sample or JSON, build report and download JSON/CSV. CLI: `node cli.mjs examples/input.json new-report.json` (optional final `csv`). Existing output files refused. Test: `node --test tests/product.test.mjs`. Input: explicit threshold 2–1000, 1–100 unique count columns, up to 50 disjoint groups with members and optional total, 1–1000 records with unique public id and exact counts. Counts must be nonnegative whole numbers <=1 billion. Every declared partition must sum to its total. Overlap/nesting/repeated group fields are refused; the tool will not silently process a problem outside its model. Positive counts below threshold are null in JSON and "withheld" in CSV. When a partition member is small, all nonzero members are hidden. Complementary cells may be large: the CSV does not falsely label them <5. Small totals are hidden; zero stays visible. A visible total cannot leave exactly one hidden member. Original counts and their hashes are absent from output. Both UI and downloads use only the suppressed result. Input remains in your local editor until closed/replaced. This is a one-table transformation, not publication approval or a privacy guarantee. No linked-table, margin, longitudinal differencing, external-knowledge or differential privacy protection. The threshold and zero policy are explicit policy choices, not universal legal/public-health requirements. Use public aggregate labels only; the tool does not remove personal information from your id fields. All data stays local. Browser visual acceptance pending independently of automated tests.