The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents methods for remote command execution across multiple protocols, including SMB, WMI, WinRM, RDP, and MSSQL. It specifically details flags for executing shell commands (-x) and PowerShell commands (-X), as well as methods to bypass security controls such as the --amsi-bypass flag.
[PRIVILEGE_ESCALATION]:
The documentation includes the use of chmod 777 on remote NFS shares to grant global permissions.
It details the use of the --chroot flag to escape to the root filesystem on remote systems.
The skill provides instructions for delegation abuse, such as Resource-Based Constrained Delegation (RBCD) and S4U2Self, to impersonate administrative users.
Examples include path traversal patterns (e.g., TARGET=../../targets.txt) when performing DACL analysis.
[DATA_EXFILTRATION]: The skill provides extensive guidance on extracting sensitive credentials and secrets, including:
Dumping local SAM hashes and LSA secrets.
Extracting the full Active Directory NTDS.dit database containing domain-wide credentials.
Harvesting DPAPI secrets, browser cookies, and stored credentials from applications like KeePass, Putty, and WinSCP.
Dumping Azure/M365 token caches (WAM).
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The tool processes untrusted data from network targets, such as SMB share contents, LDAP attributes, and SQL query results (found in SKILL.md).
Boundary markers: There are no instructions or boundary markers to prevent the agent from interpreting malicious instructions that might be embedded in the output from these remote systems.
Capability inventory: The skill enables highly privileged actions including remote shell execution, file transfers, and system-wide credential harvesting.
Sanitization: There is no evidence of validation or sanitization of data retrieved from remote network targets before it is processed by the agent.

offensive-netexec