Tagline: "An AI-powered penetration testing assistant that runs active audits, analyzes results, and auto-generates compliance-grade reports."
A local or containerized LLM orchestrator that runs a modular security audit workflow: Recon → Enumeration → Exploitation Simulation → Report.
It uses standard tools (Nmap, Nikto, Gobuster, Metasploit, etc.) under orchestration. The LLM reads outputs, interprets findings, and writes structured Markdown/HTML reports with risk ratings and remediation steps.
- Agent Brain: LLM (local via Ollama or API) with a function-calling interface for tool execution.
- Execution Sandbox: Isolated Docker container for running pentest utilities safely.
- Collector: Log aggregator for Nmap/XML, Nikto, and custom scripts.
- Analyzer: The LLM ingests parsed results, maps findings to CVE/CWE databases, and scores severity using CVSS logic.
- Report Generator: Auto-produces an HTML + PDF report with:
- Summary dashboard
- List of discovered hosts/services
- Detected vulnerabilities
- Exploitability & remediation
- Compliance checklist (CIS, OWASP, NIST 800-53 baseline)
- Python for orchestration
- LangChain or CrewAI for LLM workflow management
- Docker for sandboxing
- Nmap, Nikto, SSLyze, Lynis, Trivy, etc. as tool integrations
- Markdown → PDF pipeline using weasyprint or reportlab
- SQLite or JSON datastore for findings and context memory
Keep it light at first:
- CLI command like
sentinelaudit scan 192.168.1.0/24 --deep - Runs Nmap + Nikto
- Feeds logs into LLM
- Outputs report.md and report.pdf
- Done. That alone is portfolio gold.
TBD
TBD