email-security-deep
Fail
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: CRITICALOBFUSCATIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPERSISTENCEPRIVILEGE_ESCALATIONREMOTE_CODE_EXECUTIONTIME_DELAYED_CONDITIONALDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [OBFUSCATION]: Use of homoglyph-based typosquatting in domains such as
micros0ft-login.com,micros0ft-secure.com, andm1crosoft-login.comto impersonate legitimate services. - [OBFUSCATION]: Employment of the Right-to-Left Override (RTLO) character (U+202E) to disguise malicious file extensions (e.g., making
invoice[U+202E]txt.scrappear asinvoice scr.txt). - [DATA_EXFILTRATION]: Comprehensive instructions for deploying
evilginx2andmodlishkato perform Adversary-in-the-Middle (AiTM) attacks, specifically targeting the capture of plaintext credentials, MFA tokens, and authenticated session cookies. - [REMOTE_CODE_EXECUTION]: Provides commands to download and execute remote scripts directly in the shell using piped bash execution (e.g., for
king-phisherinstallation). - [DYNAMIC_EXECUTION]: Uses HTML smuggling to assemble and trigger the download of malicious binaries client-side within the victim's browser, bypassing gateway sandbox detection.
- [COMMAND_EXECUTION]: Includes tools and scripts for automated email spoofing and 'email bombing' (denial-of-service) against specific email targets.
- [PERSISTENCE]: Instructions for maintaining access by configuring phishing tools as persistent systemd services on a VPS.
- [PRIVILEGE_ESCALATION]: Extensive use of
sudocommands for system configuration, firewall manipulation, and tool installation. - [TIME_DELAYED_CONDITIONAL]: Implementation of execution delays and user-interaction checks to prevent payload detonation within automated sandbox environments.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from email headers and bodies, creating a vulnerability surface for prompt injection attacks if not properly sanitized.
Recommendations
- AI detected serious security threats
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata