Prox Offensive · AI Multi-Agent Red Team Lab

Author: Felix Gutierrez (aka Don Trabajo)
Organization: Proxima Centauri Productions LLC · DBA Prox Offensive Information Security

---

Overview

This repository documents the development of a multi-agent, local-first AI lab designed for red-team research, training, and creative workflow automation.

Each AI agent serves a defined role in the offensive workflow — from structured reasoning to secure, offline synthesis — combining precision, privacy, and creativity.

---

Core Components

• ChatGPT Plus (GPT-4 Turbo) – orchestration and structured logic
• Claude Pro CLI – long-context reporting and documentation
• Gemini CLI – quick lookup and validation
• Local LLMs (gpt-oss 20B / DeepSeek) – on-device synthesis
• Duck.ai – browser reviewers for ensemble reasoning
• Atlas Browser – redacted research and drafting lane

---

/init Blueprint

For detailed setup instructions, see
➡️ AI_MultiAgent_RedTeam_Blueprint.md

---

Project Goals

• Strengthen red-team methodology through modular AI integration
• Maintain local sovereignty over sensitive data and findings
• Streamline documentation and creative publishing for security professionals

---

For details on the host CLI & AI environment this project runs in, see the Multi-Agent Lab docs.

For multi-agent prompt-injection / tool-abuse hardening (Arcanum taxonomy + OWASP LLM Top 10 adapted):

• security/README.md

See the Multi-Agent Lab docs:

https://github.com/DonTrabajo/proxoffensive-ai-multiagent-lab/blob/main/docs/host_cli_orchestration.md

License

MIT License © 2025 Felix Gutierrez
Part of the Prox Offensive / Proxima Centauri Productions initiative.