We release patches for security vulnerabilities in the following versions:

We take the security of strands-mlx seriously. If you discover a security vulnerability, please follow these steps:

DO NOT open a public GitHub issue for security vulnerabilities.

Instead, please report security issues by emailing:

📧 [email protected]

Please include the following information in your report:

• Description - Clear description of the vulnerability
• Impact - Potential impact and severity
• Steps to Reproduce - Detailed steps to reproduce the issue
• Proof of Concept - Code or commands demonstrating the vulnerability
• Environment - Python version, OS, hardware (Apple Silicon, etc.)
• Suggested Fix - If you have ideas on how to fix it

• Acknowledgment - We'll acknowledge receipt within 48 hours
• Investigation - We'll investigate and validate the report
• Updates - We'll keep you informed of our progress
• Resolution - We'll work on a fix and coordinate disclosure
• Credit - We'll credit you in the security advisory (if you wish)

⚠️ Training data may contain sensitive information:

# ❌ DON'T: Store sensitive data in training files
session = MLXSessionManager(session_id="passwords", storage_dir="./data")
agent("My password is secret123")  # Saved to JSONL!

# ✅ DO: Sanitize or exclude sensitive conversations
session = MLXSessionManager(
    session_id="safe_training",
    storage_dir="./secure_data"
)
# Be mindful of what data gets trained on

🔐 Trained adapters can leak training data:

• LoRA adapters contain model weights trained on your data
• Don't share adapters trained on private/sensitive information
• Review training data before sharing adapters on HuggingFace

📁 Session managers write to disk:

# Be careful with file paths
session = MLXSessionManager(
    session_id="user_input",  # ⚠️ Sanitize user-provided IDs
    storage_dir="/safe/directory"  # ✅ Use absolute paths
)

🤖 Only load models from trusted sources:

# ✅ Official mlx-community models
model = MLXModel(model_id="mlx-community/Qwen3-1.7B-4bit")

# ⚠️ Verify adapter sources before loading
model = MLXModel(
    model_id="mlx-community/Qwen3-1.7B-4bit",
    adapter_path="untrusted/adapter"  # Could contain malicious code
)

• Issue: Conversation data saved to JSONL files
• Impact: Sensitive information may persist on disk
• Mitigation:
  • Use secure directories with proper permissions
  • Implement data retention policies
  • Sanitize sensitive data before training

• Issue: Adapters can memorize training data
• Impact: Private information could be extracted from adapters
• Mitigation:
  • Review training data before sharing adapters
  • Use differential privacy techniques (future feature)
  • Limit adapter distribution

• Issue: Agents can execute tools with system access
• Impact: Malicious prompts could abuse tool capabilities
• Mitigation:
  • Carefully select which tools to enable
  • Validate tool inputs
  • Use principle of least privilege

• Issue: Image/audio/video files processed by models
• Impact: Malicious media could exploit vulnerabilities
• Mitigation:
  • Validate media file formats
  • Use trusted media sources
  • Implement file size limits

Security updates will be released as patch versions (e.g., 0.2.5 → 0.2.6) and announced via:

• GitHub Security Advisories
• PyPI release notes
• Repository CHANGELOG.md

strands-mlx relies on several dependencies. We monitor these for vulnerabilities:

• mlx - Apple's ML framework
• mlx-lm - Language model support
• mlx-vlm - Vision/audio/video support (optional)
• strands-agents - Core agent framework

Run security checks:

# Check for known vulnerabilities
pip install safety
safety check

# Or with uv
uv pip install safety
safety check

We believe in responsible disclosure. If you report a vulnerability:

1. We'll work with you to understand and validate the issue
2. We'll develop and test a fix
3. We'll coordinate public disclosure timing
4. We'll credit you in the security advisory (unless you prefer anonymity)

Thank you for helping keep strands-mlx and its users safe! 🛡️

• Security Issues: [email protected]
• General Issues: GitHub Issues
• Discussions: GitHub Discussions