The Agent Skills Directory

[PERSISTENCE]: The skill documents several methods for maintaining access to a target system, which are standard indicators of malicious persistence.
Evidence: Phase 6 includes commands to create cron job backdoors (echo "* * * * * /tmp/backdoor.sh" >> /etc/crontab) and establish SSH persistence by adding keys to ~/.ssh/authorized_keys.
[PRIVILEGE_ESCALATION]: The skill provides specific techniques to identify and exploit privilege escalation vulnerabilities.
Evidence: Phase 6 instructs on checking for SUID binaries (find / -perm -4000) and verifying sudo permissions (sudo -l). It also references tools like linpeas.sh and winpeas.exe for automated enumeration.
[COMMAND_EXECUTION]: The instructions contain a large volume of shell commands for active system interaction, scanning, and exploitation.
Evidence: Commands for nmap, gobuster, hydra, sqlmap, and the Metasploit Framework (msfconsole) are provided throughout Phases 2-5.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests data from external targets, creating a surface for indirect injection attacks.
Ingestion points: Reconnaissance and scanning outputs (Phase 2 & 3) such as DNS records, WHOIS data, and web service banners.
Boundary markers: Absent; there are no instructions to delimit or sanitize data gathered from targets before processing.
Capability inventory: The skill has access to shell execution, network tools, and exploitation frameworks.
Sanitization: None; the methodology does not include validation or escaping of external content before it is used in subsequent phases (e.g., exploitation).

Ethical Hacking Methodology