The Agent Skills Directory

[COMMAND_EXECUTION]: Instructions for establishing persistence on a target system by appending a public key to the ~/.ssh/authorized_keys file.
[DATA_EXFILTRATION]: Commands for searching and reading sensitive files, including private SSH keys (e.g., id_rsa, id_dsa) and the .bash_history file, which may contain plaintext credentials or sensitive command history.
[COMMAND_EXECUTION]: Provides automated recipes for credential attacks, including brute-forcing and password spraying, using tools like Hydra and Medusa.
[REMOTE_CODE_EXECUTION]: Includes an embedded Python script that utilizes the paramiko library to automate SSH connections and execute arbitrary commands on remote hosts.
[DATA_EXFILTRATION]: Suggests using curl to probe for and download exposed private keys from web-accessible directories on a target server.
[COMMAND_EXECUTION]: Detailed instructions for network pivoting and bypassing security controls through local, remote, and dynamic (SOCKS proxy) SSH port forwarding.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from network scans and file content (e.g., banners, configuration audits) and processes it without explicit boundary markers or sanitization, while possessing capabilities for file modification and network operations. Evidence chain:
Ingestion points: Command output from nmap, ssh-audit, and curl (SKILL.md)
Boundary markers: Absent
Capability inventory: File system writes (authorized_keys), network operations (Nmap, Hydra), and remote command execution (paramiko) (SKILL.md)
Sanitization: Absent

SSH Penetration Testing