The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous examples of tools used to execute commands on remote Windows systems, such as psexec.py, smbexec.py, and wmiexec.py. These are legitimate tools for administrative and security auditing tasks.
[REMOTE_CODE_EXECUTION]: Instructions are included for identifying and exploiting critical vulnerabilities like MS17-010 (EternalBlue) and PrintNightmare using established security frameworks. This is consistent with a penetration testing methodology.
[CREDENTIALS_UNSAFE]: Includes a reference table of common default passwords and describes techniques like password spraying and Pass-the-Hash used to evaluate the security of SMB services.
[DATA_EXFILTRATION]: Outlines procedures for collecting sensitive data from open shares and extracting system credentials (SAM/LSA/LSASS) as part of a security assessment.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Processes output from remote SMB shares and network scanning tools as identified in SKILL.md (Phase 2 and 3).
Boundary markers: No specific boundary delimiters or safety instructions are provided to the agent for handling data retrieved from targets.
Capability inventory: Features powerful execution capabilities across multiple scripts, including wmiexec.py and crackmapexec as referenced in SKILL.md (Phase 6).
Sanitization: Lacks explicit sanitization for data ingested from untrusted external systems during the enumeration process.

smb-pentesting