Basilisk v1.0.0 — AI Red Teaming Framework

[regaan]

What's New

🔬 Differential Mode

Compare LLM behavior across providers side-by-side. Detects behavioral divergences where one model refuses but another complies.

basilisk diff -t openai:gpt-4o -t anthropic:claude-3-5-sonnet-20241022

🛡️ Guardrail Posture Scan

Non-destructive recon-only security assessment. Produces an A+ to F security grade without running attacks. Safe for production.

basilisk posture -p openai -m gpt-4o

📋 Audit Logging

Forensic-grade, SHA-256 checksummed audit trails enabled by default. JSONL format with automatic API key redaction.

⚙️ CI/CD GitHub Action

Native action.yml for pipeline integration with SARIF baseline regression detection.

🆓 GitHub Models Support

Use GPT-4o, DeepSeek-R1, Llama, Mistral and more for FREE via GitHub Marketplace Models. No API key purchases required.

🖥️ Desktop App

Three new tabs: Differential comparison, Posture assessment with live grading, and Audit trail viewer.

---

Full Feature List

• 29 attack modules covering OWASP LLM Top 10
• Smart Prompt Evolution (SPE-NL) genetic algorithm
• 5 recon modules
• 5 report formats (HTML, JSON, SARIF, Markdown, PDF)
• Native C/Go performance extensions
• Universal provider support (OpenAI, Anthropic, Google, Azure, GitHub Models, Ollama)
• Electron desktop app (Windows, macOS, Linux)

---

Install

pip install basilisk-ai

---

Links

• 🌐 https://basilisk.rothackers.com
• 📖 Documentation
• ⚙️ GitHub Action

Built by Regaan — Rot Hackers

---

This discussion was created from the release Basilisk v1.0.0 — AI Red Teaming Framework.