The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill makes extensive use of the sudo command to perform actions that require root privileges on the host system. This includes launching network interceptors, modifying network interfaces, and performing low-level packet injection.
scripts/bettercap_mitm.sh: Executes bettercap with root privileges for ARP/DNS spoofing and HSTS bypass.
scripts/mitm6_relay_launcher.sh: Runs mitm6 and ntlmrelayx as root to perform IPv6 takeover and NTLM relay attacks.
scripts/vlan_hop.py: Requires root for raw packet creation and sending via the Scapy library.
scripts/wifi_attack.sh: Uses sudo to manage wireless interfaces (airmon-ng) and launch rogue access points (hostapd).
[COMMAND_EXECUTION]: The skill provides numerous scripts that act as wrappers for high-risk offensive security binaries. The agent can use these to execute complex attack chains involving network poisoning, traffic manipulation, and remote exploitation.
The skill orchestrates the execution of tools like responder, bettercap, ntlmrelayx, mitm6, nmap, and the aircrack-ng suite.
Several shell scripts automate multi-step attacks, such as mitm6_relay_launcher.sh which manages a full IPv6 takeover and relay session.
[DATA_EXFILTRATION]: The skill is designed to intercept and harvest sensitive authentication data from the local network. While it does not hardcode a remote exfiltration server, the capability to capture and relay credentials represents a significant risk of data exposure.
scripts/responder_loot_parser.py: Designed specifically to extract NetNTLM hashes from Responder logs for offline cracking.
scripts/wifi_attack.sh: Captures WPA2 handshakes and WPA-Enterprise (MSCHAPv2) credentials for theft.
The documentation describes NTLM relaying to SMB, LDAP, and MSSQL to perform unauthorized actions as the compromised user.
[INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted external data, creating a large attack surface for indirect prompt injection.
Ingestion points: The agent processes external data sources including network maps, service lists, and network foothold positions defined in SKILL.md.
Boundary markers: The instructions lack delimiters or explicit warnings to ignore instructions that might be embedded in network scan results or intercepted traffic.
Capability inventory: The skill possesses high-privilege capabilities including raw socket access (vlan_hop.py), file system writes for loot collection, and the ability to execute arbitrary network poisoning tools.
Sanitization: There is no evidence of validation or sanitization of external data before it is interpolated into agent actions or analyzed for further steps.
[DYNAMIC_EXECUTION]: Several scripts generate and execute transient code or configuration files at runtime, which is a common technique for hiding or automating dangerous activities.
scripts/bettercap_mitm.sh: Dynamically generates a Bettercap caplet configuration file in a temporary location and immediately executes the tool using that configuration.
scripts/wifi_attack.sh: Creates hostapd configuration files on the fly to stand up rogue wireless access points for credential capture.

network-attack