5. The confined execution path
The reason this documentation exists. Everything else here serves a model that answers questions; this part is what happens when the model stops answering and starts doing — running a command, writing a file, opening a socket on the machine you are reading this on.
One rule governs all of it: fail-closed — a confinement the harness cannot apply is an error, never a silent downgrade (the list).
Read the pages in this order. Each one assumes the one before it.
Tool execution harness is the spine: what a tool is, how one is selected, the control plane and the decision, boundary and execution layers a call passes, and what the harness knows about a call before it runs. Security model holds the authorization rules that decide whether it runs at all.
The three that follow are the execution layer’s mechanisms, in increasing order of subtlety: Sandbox for the filesystem and the process, Network backend for what it may reach, Resource control for what it may consume. Subtlety is the ordering because the failure modes get quieter: a blocked path is obvious, a blocked connection less so, and a process that merely runs slower than it should is the hardest to notice at all.
Final harness audit closes the section with the whole harness on one sheet: what is on by default, what is optional and what is experimental. Runtime internals is for contributors: who owns what inside the runtime.