     _______ _____ _____   ____
    |___  / |_   _|  __ \ / __ \
       / /  | | | |__) | |  | |
      / /   | | |  _  /| |  | |
     / /__ _| |_| | \ \| |__| |
    /_____|_____|_|  \_\\____/

AI-powered penetration testing agent. Point it at a target — it finds vulnerabilities.

Quick Start

# Install
curl -sSL https://raw.githubusercontent.com/Xyeino/ziro/master/scripts/install.sh | bash

# Configure
export ZIRO_LLM="openai/gpt-5.4"
export LLM_API_KEY="your-api-key"

# Scan
ziro --target https://your-app.com

Or install manually:

pip install git+https://github.com/Xyeino/ziro.git

Requires Docker running and Python 3.12+. First run pulls the sandbox image automatically.

What It Does

Ziro spins up sandboxed agents that autonomously:

Crawl and map attack surface
Identify injection points, auth flaws, misconfigs
Chain findings into multi-step exploits
Generate PoC scripts and evidence

Results land in ziro_runs/<run-name>/ — vulnerability reports, HTTP evidence, reproduction scripts.

Usage

# Local codebase
ziro --target ./app-directory

# GitHub repo
ziro --target https://github.com/org/repo

# Black-box web app
ziro --target https://your-app.com

# Authenticated testing
ziro --target https://your-app.com --instruction "Use credentials admin:pass123"

# Multiple targets
ziro -t https://github.com/org/app -t https://your-app.com

# Instructions from file
ziro --target api.example.com --instruction-file ./scope.md

# Headless (CI/CD, no TUI)
ziro -n --target https://your-app.com

Tools

| Tool | What it does | |------|-------------| | HTTP Proxy | Intercept, modify, replay requests | | Browser | Multi-tab automation for XSS, CSRF, auth flows | | Terminal | Interactive shells inside sandbox | | Python Runtime | Write and run custom exploits | | Metasploit | Search, inspect, execute MSF modules | | API Spec Parser | OpenAPI/Swagger analysis with risk scoring | | Attack Graph | Multi-step exploitation planning | | Evidence Collector | HTTP pairs, screenshots, command output | | Scope Guard | Enforces target boundaries | | Finding Dedup | No duplicate vulnerability reports |

Finds

Access control (IDOR, privesc, auth bypass), injections (SQL, NoSQL, command), server-side (SSRF, XXE, deserialization), client-side (XSS, prototype pollution), business logic (race conditions, workflow abuse), JWT/session flaws, infrastructure misconfigs.

Configuration

export ZIRO_LLM="openai/gpt-5.4"       # required
export LLM_API_KEY="your-api-key"        # required
export LLM_API_BASE="http://localhost:11434"  # for Ollama/local models
export ZIRO_REASONING_EFFORT="high"      # low/medium/high
export ZIRO_DASHBOARD=1                  # enable live web dashboard

Config auto-saves to ~/.ziro/cli-config.json.

Models that work well: openai/gpt-5.4, anthropic/claude-sonnet-4-6, vertex_ai/gemini-3-pro-preview

CI/CD

jobs:
  security:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v6
      - run: pip install git+https://github.com/Xyeino/ziro.git
      - run: ziro -n -t ./ --scan-mode quick
        env:
          ZIRO_LLM: ${{ secrets.ZIRO_LLM }}
          LLM_API_KEY: ${{ secrets.LLM_API_KEY }}

Warning: Only test targets you own or have explicit permission to test.

Ziro

Install / Use

README