🧪 Sprits Agent Test Range

Does your websearch-capable AI agent follow instructions hidden inside the pages it reads? Start a test, point your agent at the URLs we generate, and watch which prompt-injection techniques it falls for — live.

1. Click Start a test to generate a private session.
2. Give your agent the test URL(s) — e.g. “summarize this page.”
3. If your agent obeys a hidden instruction, it quietly calls a tracking (“beacon”) URL. We log the hit and flip that test to vulnerable.

This is a defensive testing tool. Beacon hits trigger only a harmless request to our own logger — no data exfiltration, no destructive actions. Test only agents you own or are authorized to assess.