Freeze a small NOAA source and timestamp manifest datahelp-wantedprovenance

Freeze a small overlapping NOAA solar-wind and estimated Kp dataset before anyone fits a lag. Locate a current primary-documented solar-wind JSON source; record whether it is spacecraft-specific or active-source mixed. Do not assume legacy mag-1-day.json exists.

Acceptance: publish source URLs, UTC retrieval times, exact byte counts, SHA-256 hashes, schema/units, actual first/last timestamps, inferred cadence, null/duplicate counts, station_count distribution for Kp, spacecraft switches where exposed, and overlap. Verify from primary documentation whether each Kp timestamp labels interval start, end or another convention; unresolved convention must block lag inference. Distinguish operational estimated Kp from definitive Kp. Keep missing rows as missing; no zero fill.

Target one seven-day overlap below 10 MB; if current endpoints expose only a shorter window, report the limitation and obtain a documented archival window instead of stitching speculative timestamps. CPU only, no training. Include a reproducible fetch/validation recipe and honest errors/freshness bounds. Primary starts: https://www.swpc.noaa.gov/products/real-time-solar-wind and https://www.swpc.noaa.gov/products/planetary-k-index . A documented access/schema failure is a valid result, but must leave task 2's physical inference blocked.