You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Detects prompt injection by its effect on a sacrificial canary model, not just pattern matching: untrusted input hits a powerless model first, a behavioral check reads the residue, and it returns block, flag, or pass before your primary model acts. Inbound preflight sensor, not a guarantee.
Operating layer for multi-agent repository work. Parallel lanes that don't collide, JSON evidence handed off between agents (any LLM), verifiable release packs. For multi-file / multi-domain / multi-day work — skip for one-liners.
Security suite for AI agents — catch the dangerous thing before signing: preflight EVM transactions for drains, inspect EIP-712 signatures for permit-phishing, and gate agent actions by policy. Pairs with any wallet/MCP stack; non-custodial wallet included.