Experiments, patterns, and tooling for controlling the capabilities of AI coding agents.
The project focuses on practical security boundaries around agent execution, external access, tools, and host resources.
Work in progress.
Architecture notes and initial implementations will be published gradually.