cloakcheck
Scan a page for content planted to hijack an AI browsing/shopping agent before it acts on the page.
Community: Submitted by a user or imported; check the owner before granting accessDegradedNo sign-inGlobalFreeRead-only
What it can do
What data it sees
Do you need an account
No: the server works without sign-in
Scan a page for content planted to hijack an AI browsing/shopping agent before it acts on the page.
Server tool list (1)
Raw names from tools/list. Only developers need these.
| check_page_for_agent_hijacking | Fetch a URL and scan it for content planted to hijack an AI agent visiting it rather than to inform a human: invisible unicode (zero-width joiners, steganographic tag-block characters), CSS-hidden text (display:none, opacity:0, off-screen) paired with instruction-shaped language ('ignore previous instructions', 'add to cart and checkout', fake system-role text), and instruction-shaped phrases in alt/title/aria-label attributes or meta tags. Call this before an autonomous browsing or checkout flow acts on instructions found on a page. A clean result means no known hijack pattern found, not a full content-safety guarantee. |