COCOFELON — Offensive Security Engineer

The standard penetration testing workflow suffers from a critical OpSec vulnerability: sending proprietary binaries and sensitive logs to public AI models. This post details the engineering of a Controller/Agent split architecture. All source code is versioned in the Jericho repository to ensure your local infrastructure remains reproducible.

I. Infrastructure Mapping

Controller Node (Parrot OS): Executes payloads and network discovery.
Inference Server (Dell Inspiron 16+): Dedicated Ollama instance.
Bridge: Inter-process communication via local network socket (11434).

II. Deployment: Inference Server

On the Windows Inference Server, we must modify the binding to accept network-level traffic and configure the firewall to expose the API. Repository Reference: jericho/ai-inference/scripts/windows_ollama_bind.ps1

# Jericho Suite: Inference Server Network Binding
# Binds to 0.0.0.0 for external controller access
[Environment]::SetEnvironmentVariable("OLLAMA_HOST", "0.0.0.0", "Machine")
New-NetFirewallRule -DisplayName "Jericho Ollama API" -Direction Inbound -LocalPort 11434 -Protocol TCP -Action Allow

III. Deployment: CLI Bridge

The provides the interface to pipe directly into the inference engine. By reading from the buffer, we avoid storing sensitive data on disk. Repository Reference: jericho/ai-inference/scripts/parrot_agent.py

#!/usr/bin/env python3
import requests, sys, json
OLLAMA_HOST = "http://192.168.1.50:11434/api/generate"
def stream_response(prompt, model):
    payload = {"model": model, "prompt": prompt, "stream": True}
    with requests.post(OLLAMA_HOST, json=payload, stream=True) as response:
        for line in response.iter_lines():
            if line:
                sys.stdout.write(json.loads(line).get("response", ""))
                sys.stdout.flush()
if __name__ == "__main__":
    # Piped data context injection
    piped_data = sys.stdin.read() if not sys.stdin.isatty() else ""
    stream_response(f"Analyze the following data: {piped_data}", "deepseek-coder")

IV. Operational Example

Use this for live log analysis during engagement:

cat access.log | ./parrot_agent.py -m deepseek-coder