Scan Url

Public URL scan. Rate-limited tighter than any other public endpoint in this app (10/hour, vs 20-60/hour elsewhere) since each call now costs real compute (headless browser render + up to two more HTTP fetches + one Haiku call), not just a database read.

Results are cached for FINDINGS_CACHE_MINUTES per URL -- this is the primary defense against someone deliberately flooding one target with repeated scans (from one IP or many): after the first real scan, every subsequent request for the same URL within the window is an instant, free cache hit, not additional render/SSL/AI cost.