The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides a comprehensive workflow for interacting with untrusted external network targets using various scanning and exploitation tools. This creates a surface for indirect prompt injection where malicious data returned by a target system (such as service banners, HTTP headers, or database content) could be interpreted as instructions by the agent.
Ingestion points: Command output from Nmap, Nikto, SQLMap, and Metasploit (SKILL.md).
Boundary markers: Absent. The skill does not provide instructions for the agent to encapsulate or sanitize output from these tools.
Capability inventory: Includes network scanning, service enumeration, credential brute-forcing, and vulnerability exploitation via system-level CLI tools.
Sanitization: Absent. The templates focus on command execution without output validation.
[COMMAND_EXECUTION]: The skill contains a large library of command templates for system enumeration, vulnerability scanning, and network exploitation. These include tools for credential cracking (Hydra, John the Ripper) and database manipulation (SQLMap).
[REMOTE_CODE_EXECUTION]: Includes instructions for utilizing msfvenom to generate various exploitation payloads (EXE, ELF, PHP, ASP). While these are intended for use on authorized targets during a security assessment, they represent the generation of executable code templates.

Pentest Commands