Agent Armor API Reference
    Preparing search index...

    Interface SessionScanResult

    Result of scanning a multi-turn conversation. turns holds the per-turn single-string results (identical to scanning each turn on its own); crossTurnThreats holds threats that only the session view reveals.

    Note: a session has no single sanitized form — per-turn sanitized text lives on each entry in turns; cross-turn threats are advisory.

    interface SessionScanResult {
        clean: boolean;
        crossTurnThreats: CrossTurnThreat[];
        durationMs: number;
        stats: {
            crossTurnThreatsFound: number;
            highestSeverity: Severity | null;
            threatsFound: number;
            turnsScanned: number;
            windowChars: number;
        };
        turns: ScanResult[];
    }
    Index
    clean: boolean

    Whether the session is free of both per-turn and cross-turn threats.

    crossTurnThreats: CrossTurnThreat[]

    Threats that only emerge when turns are considered together.

    durationMs: number

    Time taken in milliseconds for the whole session scan.

    stats: {
        crossTurnThreatsFound: number;
        highestSeverity: Severity | null;
        threatsFound: number;
        turnsScanned: number;
        windowChars: number;
    }

    Summary stats for the session.

    Type Declaration

    • crossTurnThreatsFound: number
    • highestSeverity: Severity | null
    • threatsFound: number
    • turnsScanned: number
    • windowChars: number

      Total chars considered in the cross-turn split-payload window.

    turns: ScanResult[]

    Per-turn scan results, in the order the turns were supplied.