The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the execution of various command-line penetration testing tools including Metasploit, sqlmap, commix, searchsploit, and impacket to perform vulnerability validation and exploitation.
[REMOTE_CODE_EXECUTION]: Provides guidance and templates for delivering and executing remote payloads on target systems, including one-liner reverse shells in Bash, Python, and PowerShell, as well as staged meterpreter agents.
[DYNAMIC_EXECUTION]: Includes templates for runtime generation of shellcode using msfvenom and pwntools, alongside examples of in-memory execution using Python's exec function to avoid disk writes.
[DATA_EXFILTRATION]: Describes the setup of network listeners (e.g., netcat, pwncat-cs) and reverse connections to establish initial access and session persistence from target infrastructure.
[PRIVILEGE_ESCALATION]: Outlines methodologies for acquiring higher-level permissions on targets using tools like Metasploit's getsystem or impacket's secretsdump for credential harvesting.
[PERSISTENCE]: Includes instructions for maintaining long-term access to compromised systems by modifying shell profiles (~/.bashrc) or utilizing specialized persistence modules within exploitation frameworks.
[OBFUSCATION]: References the use of encoding techniques (e.g., x64/xor_dynamic) and payload transformation to evade detection by security controls such as WAFs and EDRs during engagements.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an ingestion surface by searching for and potentially executing code from external repositories (e.g., GitHub, exploit-db) which could contain malicious instructions.
Ingestion points: The skill searches for PoCs on public GitHub repositories like trickest/cve and nomi-sec/PoC-in-GitHub (SKILL.md, references/exploit-research.md).
Boundary markers: Includes explicit warnings to "manual code review required before any use" and "never run untested code against production."
Capability inventory: High-privilege execution capabilities including shell access, binary compilation, and protocol manipulation tools.
Sanitization: Primarily relies on operator-driven manual review and lab environment testing before proceeding to live targets.

vuln-exploit-technique