botron

name	botron
description	Use when setting up, running, or managing the Botron autonomous red team agent — Docker-based multi-provider LLM pentesting framework. Covers install, demo, service management, testing, and provider configuration.
version	1.0.0
author	dazeb
license	Apache-2.0
metadata	{"hermes":{"tags":["security","red-team","pentesting","docker","llm","autonomous-agents","liteLLM"],"requires_toolsets":["terminal","web"],"related_skills":["docker-management","env-setup"]}}

Botron — Autonomous Red Team Agent

Botron is a Docker-based autonomous red team agent with 17 specialist agents orchestrated through LangGraph. It routes all LLM calls through a LiteLLM proxy supporting 10+ AI providers with automatic failover.

Fork of Decepticon by PurpleAILAB, enhanced with multi-provider support and stripped of proprietary Claude Code OAuth dependencies.

When to Use

• Setting up Botron for the first time
• Running the Metasploitable 2 demo
• Managing Docker Compose services (start, stop, logs, health)
• Running the Python test suite
• Configuring LLM providers
• Debugging service startup issues

Quick Start

# Clone and enter (or use the Hermes tap: hermes skills tap add dazeb/botron)
cd ~/projects && git clone https://github.com/dazeb/botron.git
cd botron

# Create .env from template
cp clients/launcher/internal/config/env.example .env

Edit .env — set at least one API key (ANTHROPIC_API_KEY, OPENAI_API_KEY, etc.) and add these defaults:

LITELLM_MASTER_KEY=sk-botron-master
LITELLM_SALT_KEY=sk-botron-salt
POSTGRES_PASSWORD=botron
NEO4J_PASSWORD=botron-graph
BOTRON_MODEL_PROFILE=eco

# Start all services
make dev
# or: docker compose up -d --build

# Open CLI in separate terminal
make cli

Service Management

All commands from ~/projects/botron/:

# Start services with hot-reload
make dev

# Start services (no watch mode)
docker compose up -d --build

# Check status
docker compose ps
make status

# Health checks
make health

# View logs
docker compose logs -f langgraph
docker compose logs -f litellm

# Stop (keep data)
docker compose down

# Stop + remove all volumes (full reset)
make clean

Run the Demo

cd ~/projects/botron
make demo

Launches Metasploitable 2, loads a pre-built engagement, and runs the full kill chain autonomously: port scan → vsftpd exploit → Sliver C2 implant → credential harvesting → internal recon.

Run Tests

cd ~/projects/botron

# Python tests (local venv)
python3 -m venv .venv && .venv/bin/pip install pytest httpx "langchain-core>=0.3.0" "langchain-openai>=0.3.0" pydantic pydantic-settings pyyaml
.venv/bin/python -m pytest tests/unit/llm/ tests/unit/core/ -v

# Go launcher tests
cd clients/launcher && go test ./...

For full test suite (including Docker-dependent tests), run inside the langgraph container after installing pytest there.

Architecture

botron/
├── botron/              # Python package — agents, LLM, tools, middleware
│   ├── llm/             # Model factory, router, model definitions
│   ├── agents/          # 17 specialist agents + prompts
│   ├── tools/           # Security tools (web, reversing, cloud, AD, research)
│   └── core/            # Config, logging, engagement loop
├── config/
│   └── litellm.yaml     # LiteLLM proxy model routing (10+ providers)
├── containers/          # Dockerfiles for each service
├── clients/
│   ├── launcher/        # Go CLI (onboard wizard, launcher)
│   └── web/             # Next.js dashboard
└── langgraph.json       # LangGraph graph definitions (17 agents)

Docker Compose Services

Service	Port	Purpose
botron-postgres	5432	LiteLLM + web dashboard DB
botron-neo4j	7474/7687	Attack chain knowledge graph
botron-litellm	4000	LLM API gateway (10+ providers)
botron-langgraph	2024	Agent API server
botron-sandbox	—	Hardened Kali red-team sandbox
botron-web	3000	Next.js dashboard

Supported LLM Providers

Provider	Models	Env Var
Anthropic	Opus 4.6, Sonnet 4.6, Haiku 4.5	ANTHROPIC_API_KEY
OpenAI	GPT-5.4, GPT-4.1	OPENAI_API_KEY
Google	Gemini 2.5 Flash	GEMINI_API_KEY
DeepSeek	DeepSeek-Chat, DeepSeek-Reasoner	DEEPSEEK_API_KEY
xAI	Grok-4, Grok-4-Mini	XAI_API_KEY
Groq	Llama 3.3 70B, Llama 3.1 8B, Mixtral	GROQ_API_KEY
Together AI	Llama 4 Maverick, DeepSeek-V3	TOGETHER_API_KEY
Fireworks	Llama 3.1 70B	FIREWORKS_API_KEY
MiniMax	M2.7	MINIMAX_API_KEY
Ollama	llama3.2 (local)	OLLAMA_API_BASE

Common Pitfalls

1. Missing .env file — Docker Compose requires .env at repo root. Copy from clients/launcher/internal/config/env.example and add at least one API key.

2. LangGraph fails to start — Check langgraph.json graph paths. All 17 agents must point to ./botron/agents/<name>.py:graph. If paths reference ./decepticon/ you're on an old version.

3. LiteLLM can't reach providers — Verify API keys are set in .env. Test directly:

   curl -X POST http://localhost:4000/chat/completions \
     -H "Authorization: Bearer sk-botron-master" \
     -H "Content-Type: application/json" \
     -d '{"model":"anthropic/claude-haiku-4-5","messages":[{"role":"user","content":"hi"}]}'
   

4. Old Decepticon containers conflict — If you previously ran upstream Decepticon, tear down first:

   docker compose --profile cli --profile victims down --volumes
   docker network rm decepticon-net 2>/dev/null
   

5. Python tests fail with import errors — The project depends on langchain, httpx, etc. Install full deps in a venv first (see test section above).

6. make dev uses old code — Rebuild images after code changes:

   docker compose build langgraph litellm
   docker compose up -d --wait langgraph
   

Verification Checklist

• .env exists with at least one API key and BOTRON_MODEL_PROFILE
• docker compose ps shows all services healthy
• curl -s http://localhost:2024/ok returns {"ok":true}
• curl -s http://localhost:4000/health returns 200
• make health passes all checks
• docker compose exec langgraph python -m pytest tests/unit/llm/ -q passes