The New AI Agent Primitive: Why Policy Needs Its Own Language (And Why YAML and Rego Fall Short)

AI agents now perform critical tasks like writing code and moving money, but existing policy tools like YAML configs and OPA/Rego were not designed for agent governance. The article introduces FPL (Faramesh Policy Language), a domain-specific language with first-class primitives for budgets, phases, delegation, and compile-time mandatory deny rules. FPL supports natural language policy compilation, backtesting against real decision records, and GitOps-native workflows.