Real-Time Evaluation Stream
Streaming live...
Interactive Guardrails Playground
Select a pre-made exploit pattern or input a custom adversarial prompt to see Securum analyze, block, or redact the request in real time.
Securum Security Diagnostic
Exploit Classification
N/A
Firewall Action
N/A
Severity Rating
N/A
Refusal Logic Applied
N/A
Agent Response Output
Active Policy Configuration
Prompt Obfuscation Detector
Inspects inputs for Base64, Hex encodings, and Morse code formats.
System Prompt Leak Guard
Blocks requests querying internal instructions, parameters, or file structures.
Executor Sandbox Isolation
Restricts arbitrary file writes or terminal runs; intercepts excessive agency tags.
Active Self-Defense Prompts
Injects adversarial instructions directly into compiled agent system prompts.
Active Self-Defense Prompt Block