Z3r0: Fully automated penetration testing

This is a submission for the GitHub Finish-Up-A-Thon Challenge

What I Built

I built Z3r0, a controlled multi-agent workbench for authorized security assessment, code auditing, internal review, and research.

Z3r0 coordinates a lead security agent with specialist agents for code audit, intelligence gathering, penetration validation, reverse engineering, and cryptography review. It combines a React workbench, FastAPI backend, PostgreSQL persistence, WebSocket streaming, OpenAI-compatible model providers, and Docker-backed sandbox tooling.

The goal is to make AI-assisted security work more structured, traceable, and bounded by explicit authorization.

Demo

yv1ing / Z3r0

A controlled multi-agent workbench for authorized security assessment, code auditing, internal review, and controlled research.

English · 中文

Architecture · Agent Team · Runtime Model · Deployment · Quickstart

---

⚠️ Legal Notice

This project may be used only within a lawful and explicitly authorized scope for security testing, assessment, and research. Any unauthorized, unlawful, or harmful use is strictly prohibited. The author assumes no responsibility for any consequences, losses, damages, legal liabilities, or unlawful acts caused by users.

This project is provided only for authorized security assessment, code auditing, internal review, and controlled research. It does not grant permission to test, access, scan, or affect any third-party system, network, service, account, or data. Users are solely responsible for obtaining and preserving authorization, defining scope, and complying with applicable laws, contracts, and authorization boundaries.

Z3r0 is a controlled multi-agent workbench for authorized security assessment, code auditing, internal review, and controlled research. It coordinates a lead security agent, domain specialists, and Docker-backed execution surfaces so planning, evidence…

View on GitHub

The Comeback Story

Z3r0 began as an ambitious security-agent prototype. The core idea was there, but it still needed the pieces that make a project feel complete: stable runtime behavior, agent delegation, persistent history, frontend streaming, sandbox control, deployment, and documentation.

During the finish-up work, I turned those pieces into a coherent system.

The biggest change was the agent runtime. Z3r0 now supports interrupt-driven execution, persistent background subagent jobs, streamed progress, and coordinator notifications when specialist agents finish their work.

I also added a clearer frontend event contract, so raw model events are normalized into stable events like thinking_delta, text_delta, tool_call, tool_result, and subagent_task.

Sandbox handling became more controlled as well. Command tools are only mounted when a running, authorized Docker sandbox is bound to the session, and sandbox state changes invalidate affected tools and tasks.

The project moved from a promising prototype to a self-hostable multi-agent security workbench with clear architecture, deployment steps, and safety boundaries.

My Experience with GitHub Copilot

GitHub Copilot helped me move faster while finishing the less glamorous but important parts of the project: FastAPI handlers, Pydantic schemas, React components, TypeScript types, Docker configuration, and documentation.

It was especially useful as a pair-programming assistant while refining runtime flows, event schemas, and integration code. For security-sensitive behavior, I treated Copilot as a drafting tool rather than an authority, reviewing the output carefully before using it.

Copilot helped keep momentum high while I focused on the architectural decisions that mattered most: controlled execution, traceable agent collaboration, and a workflow designed for authorized security review.