The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/agent.py script utilizes subprocess.run to manage network interfaces, such as adding VLAN subinterfaces and loading the 8021q kernel module. These calls use argument lists rather than shell strings, which follows best practices to prevent command injection.
[OBFUSCATION]: A hex-encoded byte sequence is present in scripts/agent.py. This sequence corresponds to the binary structure of a Dynamic Trunking Protocol (DTP) frame used for switch spoofing. This is a functional requirement for packet crafting and is not considered malicious obfuscation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted network data by sniffing CDP, LLDP, and DTP frames via Scapy's sniff() function in scripts/agent.py.
Ingestion points: Network interface eth0 via sniff() calls.
Boundary markers: Absent (not applicable as data is not treated as natural language instructions).
Capability inventory: Raw packet sending (sendp), interface management (ip link), and kernel module loading (modprobe).
Sanitization: The script logs metadata (MAC addresses, timestamps) but does not execute logic based on the content of the captured frames, significantly limiting the attack surface.
[PRIVILEGE_ESCALATION]: The skill requires sudo privileges for capturing raw network traffic and modifying system network configurations. This is an expected and necessary requirement for the intended use case of network security testing.

performing-vlan-hopping-attack