network-penetration-testing
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- COMMAND_EXECUTION (HIGH): The skill contains numerous commands for active network exploitation and scanning.
- Evidence: Use of
nmap,masscan,msfconsole,hashcat, andjohnacross multiple sections (SKILL.md). - Context: These tools are used for intrusive operations like OS identification, vulnerability scanning, and password cracking.
- PROMPT_INJECTION (HIGH): Highly susceptible to Indirect Prompt Injection (Category 8).
- Ingestion points: The agent processes output from untrusted external sources, specifically network banners, RPC service enumeration, and Nmap script outputs (SKILL.md).
- Boundary markers: None present. The agent is not instructed to ignore embedded instructions in service banners or script results.
- Capability inventory: The skill possesses high-impact capabilities including
exploit(Metasploit), privilege escalation (getsystem), and lateral movement (smbexec.py). - Sanitization: Absent. There is no evidence of filtering or validation for incoming network data.
- DATA_EXFILTRATION (MEDIUM): Includes commands specifically designed to extract sensitive credentials from systems.
- Evidence:
hashdump,run post/windows/gather/smart_hashdump, andsekurlsa::tickets /export(SKILL.md). - PRIVILEGE_ESCALATION (MEDIUM): Provides methods for bypassing security controls and escalating privileges.
- Evidence: Explicit use of
getsystem,mimikatz, and Kerberos ticket injection (ptt) (SKILL.md).
Recommendations
- AI detected serious security threats
Audit Metadata