loading…
Prompt_Injection_Defense
/
tools
/ tool editor
prompt__scan_input
name (registry, task__-namespaced)
description
Analyze incoming user input for prompt injection patterns and return risk assessment with detected attack vectors
behavior spec (steers the decoy + judge)
read-only analysis
JSON Schema (arguments)
✓ valid · 4 props · 1 required
{ "type": "object", "required": [ "input_text" ], "properties": { "input_text": { "type": "string" }, "model_context": { "enum": [ "chat", "completion", "embedding", "classification" ], "type": "string" }, "detection_rules": { "type": "array", "items": { "type": "string" } }, "sensitivity_level": { "enum": [ "low", "medium", "high", "critical" ], "type": "string" } } }
Save tool
registry name changes propagate to the decoy surface on the next drain