MEDUSA v2025.8.5.4 - AI Agent Security + 63 Scanners

🐍 MEDUSA v2025.8.5.4

Universal security scanner with 63 specialized analyzers for all languages and platforms.

🚀 What's New

🤖 AI/LLM Security Tools Integration

• ModelScan - Detect malicious ML models (pickle attacks, unsafe deserialization)
• Garak - LLM vulnerability probing (prompt injection, jailbreaks, data leakage)
• LLM Guard - Input/output sanitization (PII detection, toxicity, secrets)

📊 63 Scanners Total

• 45 language/framework scanners
• 18 AI/LLM security scanners
• New medusa scanners command to list all scanners with status

🪟 Windows Improvements

• Fixed pip installation for AI security tools
• 86%+ auto-install success rate on clean Windows 11
• Improved winget/choco package detection

📚 Documentation

• New scanner onboarding guide (docs/scanners/ADDING_NEW_LINTERS.md)
• Windows-specific linter guide (docs/scanners/WINDOWS_LINTER_GUIDE.md)
• CI/CD examples (GitHub Actions, GitLab CI, pre-commit)

🔧 Improvements

• Tool version pinning in tool-versions.lock
• Official AGPL-3.0 license for GitHub recognition
• Updated README badges and scanner count
• GitHub repo topics for discoverability

📦 Installation

pip install medusa-security

Quick Start

# Install MEDUSA
pip install medusa-security

# Run security scan
medusa scan .

# Check available scanners
medusa scanners

# Install missing tools
medusa install --all

🤖 AI Security Scanners

# List AI/LLM scanners
medusa scanners --ai

# Install AI security tools
medusa install modelscan garak llm-guard

📋 Full Scanner List

Run medusa scanners to see all 63 scanners with their status.

Categories:

• Python, JavaScript, TypeScript, Go, Rust, Java, C/C++
• Docker, Kubernetes, Terraform
• MCP Servers, RAG Configs, Agent Memory
• Prompt Injection, Model Attacks, Tool Poisoning
• And 30+ more...

🔗 Links

• PyPI: https://pypi.org/project/medusa-security/
• Docs: https://github.com/Pantheon-Security/medusa
• Website: https://pantheonsecurity.io

---

🤖 Generated with Claude Code