Skip to content

Repository files navigation

T-100AI — AI-Powered Offensive Security Terminal

Python Ollama License CI Offline Docker

T-100AI is a local AI-powered offensive security terminal. It combines a local LLM (Ollama) with a sandboxed execution engine, AI skills for pentesting, and a plugin system — all operating 100% offline with zero data leakage.

Architecture

flowchart LR
    U[User]
    T[T-100AI Terminal]
    L[Ollama Local LLM]
    S[Sandbox Executor]
    K[Security Tools]

    U -->|Commands / Prompts| T
    T -->|Context + Queries| L
    L -->|Generated Actions| T
    T -->|Execute Action| S
    S -->|Scope Validation| S
    S -->|Approved Call| K
    K -->|Raw Output| S
    S -->|Filtered Result| T
    T -->|Formatted Response| U
Loading

Quick Start

# 1. Clone
git clone https://github.com/Ruby570bocadito/T-100AI.git
cd T-100AI

# 2. Install
pip install -r requirements.txt

# 3. Ensure Ollama is running
ollama pull mistral:7b

# 4. Launch
python -m t100ai.cli.main

Requirements: Python 3.10+, Ollama installed and running, Linux/WSL environment.

Features

Feature Description
Local LLM Integration Full Ollama integration (mistral, llama3, qwen, gemma, deepseek). Zero data leaves your machine.
Sandboxed Execution Scope validation, rate limiting, auto-sudo approval, intelligent blocklist.
MCP Advanced Tool templates, chaining, auto-discovery, output parsers, interactive prompt builder.
AI Skills Recon, OSINT, Web, Post-Exploitation, Forensics, Active Directory, Reporting.
Multi-Agent Orchestrator Parallel sub-agents: Recon → Exploit → Analyst → Reporter. Conditional workflows.
Workflow Engine Conditional steps, loops, variables, interactive editor, real-time execution.
Guardrails Prompt injection detection, sensitive data filtering, tool approval gates, audit logs.
Plugin System Custom skill loader, community plugins, extensible tool registry.
Offline Mode Full capability without internet. Air-gapped by design. No telemetry.

Commands

Command Description
/help Show interactive help menu
/scan <target> Run AI-guided reconnaissance
/recon <domain> OSINT and subdomain enumeration
/post <session> Post-exploitation actions
/forensics <path> Forensic analysis on target
/workflow <name> Execute saved workflow
/skill <name> Load a security skill
/sandbox <cmd> Run command in sandbox
/config Edit configuration
/quit Exit

Project Structure

T-100AI/
├── pyproject.toml                    # Build config + dependencies
├── requirements.txt                  # All dependencies
├── run.sh / run.bat                  # Quick launch scripts
├── Dockerfile                        # Multi-stage Kali + Python
├── docker-compose.yml                # T-100AI + Ollama stack
├── .github/workflows/ci.yml          # CI pipeline (lint + test + coverage)
├── src/t100ai/
│   ├── cli/
│   │   ├── main.py                   # Typer CLI entry point + REPL
│   │   └── session_commands.py       # Session command handlers
│   ├── core/
│   │   ├── engine.py                 # Main execution engine (2000+ lines)
│   │   ├── config.py                 # Configuration (Pydantic)
│   │   ├── session.py                # Session management
│   │   ├── sandbox.py                # Sandbox executor
│   │   ├── guardrails.py             # Prompt injection + data filtering
│   │   ├── permissions.py            # Permission levels
│   │   ├── llm_handler.py            # LLM interaction handler
│   │   ├── command_executor.py       # Command execution
│   │   ├── command_router.py         # Command routing
│   │   ├── templates.py              # Prompt templates
│   │   ├── models.py                 # Data models
│   │   ├── audit.py                  # Audit logging
│   │   ├── engagement.py             # Engagement tracking
│   │   ├── report_generator.py       # Report generation
│   │   ├── session_manager.py        # Session persistence
│   │   ├── storage.py                # Storage backend
│   │   ├── tool_service.py           # Tool service
│   │   ├── wordlist_loader.py        # Wordlist loading
│   │   ├── mitre.py                  # MITRE ATT&CK integration
│   │   ├── mitre_navigator.py        # MITRE Navigator
│   │   ├── i18n.py                   # Internationalization
│   │   └── log_rotation.py           # Log rotation
│   ├── llm/
│   │   ├── client.py                 # Ollama client
│   │   ├── handler.py                # LLM handler
│   │   ├── connection_manager.py     # Connection management
│   │   ├── prompt_builder.py         # Dynamic prompt builder
│   │   └── service.py                # LLM service
│   ├── skills/
│   │   ├── base.py                   # Base skill class
│   │   ├── manager.py                # Skill manager
│   │   ├── recon.py                  # Reconnaissance skill
│   │   ├── osint.py                  # OSINT skill
│   │   ├── web.py                    # Web security skill
│   │   ├── postex.py                 # Post-exploitation skill
│   │   ├── forense.py                # Forensics skill
│   │   ├── ad.py                     # Active Directory skill
│   │   ├── report.py                 # Reporting skill
│   │   └── advanced_framework.py     # Advanced skill framework
│   ├── mcp/
│   │   ├── registry.py               # Tool registry
│   │   ├── advanced_registry.py      # Advanced registry
│   │   ├── executor.py               # Tool executor
│   │   └── tool.py                   # Tool definitions
│   ├── plugins/
│   │   ├── base.py                   # Plugin base class
│   │   ├── plugin_manager.py         # Plugin manager
│   │   ├── marketplace.py            # Plugin marketplace
│   │   └── examples/                 # Example plugins
│   ├── workflows/
│   │   ├── definitions.py            # Built-in workflows
│   │   └── executor.py               # Workflow executor
│   ├── agents/
│   │   └── orchestrator.py           # Multi-agent orchestrator
│   ├── analysis/
│   │   ├── attack_graph.py           # Attack graph analysis
│   │   ├── chain_of_custody.py       # Evidence chain of custody
│   │   ├── cvss_scorer.py            # CVSS scoring
│   │   ├── finding_cluster.py        # Finding clustering
│   │   ├── ioc_manager.py            # IOC management
│   │   ├── kill_chain.py             # Kill chain analysis
│   │   ├── purple_team.py            # Purple team exercises
│   │   └── risk_prioritizer.py       # Risk prioritization
│   ├── api/
│   │   └── server.py                 # API server
│   ├── compliance/
│   │   └── frameworks.py             # Compliance frameworks
│   ├── utils/
│   │   ├── audit.py                  # Audit utilities
│   │   ├── errors.py                 # Error handling
│   │   ├── history.py                # Command history
│   │   ├── logging.py                # Structured logging
│   │   ├── perf_profiler.py          # Performance profiling
│   │   └── sensitive.py              # Sensitive data handling
│   └── wordlists/
│       └── dictionaries.py           # 700+ integrated wordlists
├── wordlists/
│   └── dictionaries.py               # Standalone wordlist module
└── tests/                            # Test suite (20+ test files)

Docker

# Full stack (T-100AI + Ollama)
docker compose up -d
docker compose exec t100ai python -m t100ai.cli.main

# Or standalone
docker build -t t100ai .
docker run -it --rm \
  -e T100AI_OLLAMA_HOST=http://host.docker.internal:11434 \
  t100ai

Configuration

Configuration via environment variables:

Variable Default Description
T100AI_OLLAMA_HOST http://localhost:11434 Ollama API endpoint
T100AI_DATA_DIR . Data directory
OLLAMA_MODEL Model name (via .env)

Or via pyproject.toml + config file system.

Development

pip install -e ".[dev,ollama,export,workflows]"

# Run tests with coverage
pytest tests/ -v --cov=src/t100ai

# Lint
ruff check src/ tests/

# Type check
mypy src/t100ai --ignore-missing-imports

Security & Ethics

T-100AI is designed for authorized security professionals only. Always:

  • Obtain explicit written permission before testing any system
  • Use only in isolated environments for production testing
  • Follow responsible disclosure practices
  • Never use against systems you don't own or have written authorization for

Disclaimer: The authors assume no liability for misuse. You are responsible for complying with all applicable laws.

License

MIT © Ruby570bocadito

About

T-100AI — AI-Powered Offensive Security Terminal. Local LLM + sandboxed execution + pentesting skills. 100% air-gapped.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages