The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md provide detailed examples of shell and PowerShell commands for performing offensive security operations. These include reconnaissance tools such as amass and nmap, as well as high-risk credential harvesting and lateral movement tools like SafetyKatz, Rubeus, secretsdump.py, and psexec.py.
[EXTERNAL_DOWNLOADS]: The skill documentation references several external offensive security frameworks including Cobalt Strike, Sliver, Havoc C2, and GoPhish. It also requires the installation of the impacket and attackcti Python libraries, which are utilized for advanced penetration testing and red teaming activities.
[DATA_EXFILTRATION]: Instructions within the skill explicitly describe methods for simulating data exfiltration. Examples provided include using dnscat2 for exfiltrating data over DNS and general techniques for moving data through Command and Control (C2) channels.
[PROMPT_INJECTION]: The skill processes external data which creates an attack surface for indirect prompt injection. 1. Ingestion points: The scope and objectives arguments in scripts/agent.py. 2. Boundary markers: No delimiters or warnings are used to separate user input from the rest of the generated plan. 3. Capability inventory: The agent is instructed on how to use invasive tools including secretsdump.py, psexec.py, and dnscat2 across all engagement phases. 4. Sanitization: The agent.py script performs no validation or escaping of the input data before including it in the output.

conducting-full-scope-red-team-engagement