The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents and automates the execution of shell commands on remote targets via XXE-to-SSRF exploitation chains, specifically providing reverse shell payloads using the PHP expect wrapper or Redis cron job injection.
[REMOTE_CODE_EXECUTION]: Provides ready-to-use payloads and a Python script (scripts/xxe_probe.py) designed to achieve code execution on remote systems by leveraging XML parser misconfigurations.
[DATA_EXFILTRATION]: Contains instructions and code for Out-of-Band (OOB) exfiltration of file contents using HTTP, DNS, and FTP protocols, including a built-in FTP server to capture exfiltrated data.
[CREDENTIALS_UNSAFE]: Includes a library of payloads targeting sensitive system files and credentials, such as SSH private keys (id_rsa), environment variables containing secrets (/proc/self/environ), and cloud provider metadata (AWS/GCP/Azure IAM credentials).
[PROMPT_INJECTION]: The skill identifies a significant attack surface for indirect prompt injection as it ingests untrusted XML data and target URLs through the xxe_probe.py script. These ingestion points lack explicit boundary markers or sanitization, while the capability inventory includes network communication and file writing, creating a path for potential exploitation by malicious external data.

xxe-phantom