The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes privileged system commands to alter network configurations, including sudo sysctl -w net.ipv4.ip_forward=1 to enable packet forwarding and sudo iptables to redirect TCP traffic (ports 80, 443) to local interceptors.
[COMMAND_EXECUTION]: The Python script scripts/agent.py utilizes dynamic module loading via __import__("time") to manage timing during the ARP spoofing loop.
[DATA_EXFILTRATION]: The workflow describes the interception of cleartext network traffic and the logging of potentially sensitive POST data and session cookies to local files (captured_forms.log and captured_cookies.log).
[CREDENTIALS_UNSAFE]: The skill implements automated credential extraction, storing intercepted authentication tokens and form data in plain-text logs on the local filesystem without encryption.
[PROMPT_INJECTION]: The skill creates an ingestion point for untrusted network data by sniffing traffic using Scapy and mitmproxy. This establishes an indirect injection surface where malicious payloads embedded in network traffic could influence the agent's behavior during analysis.

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