This is an experimental research project for evolving AI agent architectures. If you discover a security issue, please open an issue with the label security.

Please do not disclose vulnerabilities publicly until they have been addressed.

• The agent_forge/ module — ensure generated agent specs cannot produce harmful behavior
• The Docker sandbox — validate that agent execution is properly isolated
• The dashboard — ensure it doesn't expose sensitive evolution data

• This project generates AI agent blueprints; the output is experimental and not validated for production use
• Agent blueprint evolution is inherently experimental — no guarantees about safety or correctness of generated agents
• Docker sandbox isolation depends on the host Docker configuration