Static security scan of a single text or prompt blob: hidden unicode, prompt injection, data-exfiltration directives, dangerous-capability and tool-shadowing language, and obfuscation. Returns a 0-100 risk score, category and findings. Security indicators, not a guarantee.