Super Lab / Benchmark edition 2026.11 / OWASP OAT-011 Scraping
Your site, analyzed. Then the fight, staged on its shape.
Paste a link to your site. The lab actually analyzes it - its listing structure, items per page, how many records it exposes, which fields each record leaks, what API surface shows - with a few read-only fetches, the way a scraper's first reconnaissance would. Then it stages the fight on a fair twin built to your site's shape: no defense vs antiRea, driven by attack and defense behavior measured in recorded benchmark runs. Your real site is never attacked. And every fail is published below, by design.
Site analysis / measured from your site just now - not a guess
The arena / one cell per record, red means taken
Match report / measured, not narrated
How the frame stays honest
The analysis is real and it is reconnaissance-grade: the same handful of read-only fetches a scraper makes first. What it reports - structure, page size, record estimate - always names its basis (the site's own JSON-LD total, pagination math, a sitemap census, embedded page data, or a labeled lower bound). Estimates are labeled estimates.
Twins are basic market-standard sites with a correct basic setup, verified by a fairness probe on both arms; if a twin loses data, the cause is the attack method, never a careless victim. Battle outcomes on this page are simulations: the mechanics (budget stop, the key-gate wall, identity rotation, request counts, legit cost) come from the edition's recorded live runs, cited in every report, and a record counts as taken only under the same marker rule the live range uses. antiRea's known limits are published in its LIMITATIONS.md. The recorded 3.0.0 cells show both sides honestly: a per-identity budget alone loses to identity rotation, so 3.0 puts the tail behind server-issued keys - the listing path closes entirely, and what a key-holding harvester can still draw is capped by the issuance pool (measured at 10,000-record scale in the 3.0 validation run: 1,457 of 10,000 records).