/** * Detects inbound messages that are reflections of assistant-originated content. * These patterns indicate internal metadata leaked into a channel and then * bounced back as a new inbound message — creating an echo loop. */ export type ReflectionDetection = { isReflection: boolean; matchedLabels: string[]; }; /** * Check whether an inbound message appears to be a reflection of * assistant-originated content. Returns matched pattern labels for telemetry. */ export declare function detectReflectedContent(text: string): ReflectionDetection;