The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes sensitive commands with elevated privileges to modify host networking.
Evidence: sudo sysctl -w net.ipv4.ip_forward=1, sudo iptables -t nat -A PREROUTING -i eth0 -p tcp --dport 80 -j REDIRECT --to-port 8080.
[PRIVILEGE_ESCALATION]: Extensive use of sudo grants the agent control over the network stack and system configuration files.
Evidence: sudo bettercap, sudo mitmproxy, and modification of /etc/ettercap/etter.dns.
[DYNAMIC_EXECUTION]: The skill writes a Python script to the file system and executes it during runtime to automate data extraction.
Evidence: Writing extract_creds.py using cat and running it with sudo mitmproxy -s extract_creds.py.
[DATA_EXFILTRATION]: The skill is designed to intercept and record sensitive communications, including authentication cookies and POST data, into local log files.
Evidence: Scripts writing to captured_forms.log and captured_cookies.log.
[INDIRECT_PROMPT_INJECTION]: The skill creates a vulnerability surface by processing untrusted network data without sanitization.
Ingestion points: Network packets intercepted via mitmproxy and scapy.sniff in scripts/agent.py.
Boundary markers: None present in instructions or scripts.
Capability inventory: sysctl, iptables, bettercap, mitmproxy, scapy.send, scapy.sniff (all requiring high privileges).
Sanitization: No sanitization or validation of intercepted content before logging.

conducting-man-in-the-middle-attack-simulation