Shannon
An AI-powered penetration testing agent that automates vulnerability assessment with multi-phase analysis and durable workflow orchestration.
Overview
Shannon is an AI-powered penetration testing framework for defensive security analysis. It automates vulnerability assessment by combining external reconnaissance tools (nmap, subfinder, whatweb) with AI-powered code analysis, orchestrated through Temporal durable workflows for crash recovery and auditability.
Key Features
- 5-Phase Pipeline: Reconnaissance, vulnerability analysis, exploitation validation, impact assessment, and report generation—each phase builds on findings from the previous.
- Multi-Agent Architecture: Specialized agents for recon, vulnerability analysis, exploitation, and reporting work in parallel with coordinated handoffs.
- Durable Workflows: Temporal orchestration ensures every analysis run can recover from crashes, resume from checkpoints, and produce complete audit trails.
- OWASP Coverage: Tests for injection, XSS, SSRF, broken authentication, and other OWASP Top 10 vulnerabilities.
- Tool Integration: Wraps industry-standard tools (nmap, subfinder, whatweb, Playwright) alongside Claude-powered code analysis.
- Audit Logging: Every action, finding, and decision is logged for compliance and review.
- Docker-Based: Fully containerized with Docker Compose for reproducible environments.
Technical Architecture
Temporal serves as the workflow engine, coordinating multi-phase analysis activities. Each phase is a Temporal activity that can run reconnaissance tools, invoke AI analysis, or validate findings. The Claude Agent SDK powers the AI reasoning layer for vulnerability classification and exploitation strategy.
Core components:
- Workflow Engine: Temporal server managing durable, recoverable analysis workflows.
- Recon Activity: nmap, subfinder, whatweb for external reconnaissance.
- AI Analyzer: Claude Agent SDK for vulnerability classification and code review.
- Report Generator: Structured findings with severity, evidence, and remediation.
Technology Stack
- Language: TypeScript
- Orchestration: Temporal (durable workflows)
- AI: Claude Agent SDK (Anthropic)
- Tools: nmap, subfinder, whatweb, Playwright MCP
- Infrastructure: Docker Compose, Express.js
- Config: YAML with JSON Schema validation
Current Status
Active with 5-phase pipeline operational, parallel agent execution, crash recovery, and structured reporting. Currently expanding vulnerability detection patterns and improving exploitation validation accuracy.
Have questions about Shannon?
Try asking the AI assistant! Here are some ideas:
Related Projects
ClawGuard
A security analysis platform for AI agent skills that scans for malware, prompt injection, and supply chain risks using multi-layer analysis.
PromptKart
A prompt engineering and evaluation suite with tracks, heats, and leaderboards for systematic LLM testing.
Gloss
A local-first, privacy-preserving alternative to Google's NotebookLM with RAG-powered chat using local LLM inference via Ollama.