AI Prompt Injection Red Flag Checker
Review pasted untrusted text for common prompt injection signals, without uploading data. A human-review aid, not a safety guarantee.
Inspect untrusted text for red flags
Paste website text, retrieved documents, tool results or messages that an AI assistant may read. This privacy-conscious scanner matches visible patterns; it does not execute the text or send it to a server.
Review notes
How to interpret the results
Warnings may suggest an attempt to redirect an AI agent, steal secrets, misuse tools or blur trust boundaries. Ordinary quoted examples can also trigger matches. A clean scan does not prove content is safe, and this tool is not a vulnerability scanner.
Review suspicious passages with a human, keep external content separate from system instructions, and ensure agents require approval for sensitive actions. See our agent security guide for practical boundaries.