The Orchestrator

This chapter is not about a Cursor plugin or an OpenAI tool. It is not here to assign these concepts to easy-to-install software components, because they are not software components. There is no product called a permission envelope, no plugin named directives, no setting called learning sources. The components are the taxonomy this book will keep using, and the next two chapters extend it in two directions: Chapter 4 shows the formats — the shared files and schemas that begin to standardize the components — and Chapter 5 surveys the products, sorting each one by which components it supplies and which it leaves to somebody else. Read every product named there through that taxonomy: none of them is the definition, none is the only way to build one, and all of them are simply what happens to be available right now. The point of naming the components is that an orchestrator has to be able to think in them, ask about them, and notice when one is missing.

The next chapter shows what that looks like in practice: how an orchestrator approaches a problem a developer would otherwise own, what gets delegated, what gets measured, and what the work looks like once the system can act on its own. [^ch03_oasp]: NVIDIA, “NVIDIA Launches Open Agent Safety Platform to Secure Agents From Testing to Deployment,” September 28, 2026, https://nvidianews.nvidia.com/news/open-agent-safety-platform, and NVIDIA Technical Blog, “NVIDIA Open Agent Safety Platform: A Reference for Continuous In-Silicon Agent Monitoring,” https://developer.nvidia.com/blog/nvidia-open-agent-safety-platform-a-reference-for-continuous-in-silicon-agent-monitoring/ — OpenShell verifiable policy over files, networks, tools, processes, and credentials; Justin Boitano: “Agents can drift when instructions are ambiguous. An agent cannot be expected to fully police its own behavior… Infrastructure needs to enforce explicitly.” Announced September 28, 2026. The reading of the launch as an admission about permission envelopes is this book’s interpretation; NVIDIA’s framing is safety governance. Verified September 29, 2026.

HQ 8 — Authored. I wrote the argument, structure, examples, and prose. AI tools were used for research assistance and light editorial polish; the voice, ideas, and substance are mine.