This section documents the operating-tooling layer for the AI agents that build this repo — hooks, skills, memory system, MCP servers, and prompts that shape how workers operate while implementing and developing the codebase.

It is deliberately separate, in context, from the rest of the docs:

• Other sections (Contracts, Internal / governance, and the product documentation) describe the repo and its product — what we are building and the state of that work.
• This section describes how the workers work — the tooling layered on top of the agents who develop the repo, not the product they produce.

Treat this as the living home for discussions and activities about agent operating tools. Add and update pages here as the hook stack, skill library, memory model, MCP setup, and prompt protocols evolve.

Pages

• Hooks — the hooks wired into the agent harness: what fires, what each one injects or enforces, and where the code lives.
• Skills — the skill library: where skills live, the curated catalog, the searchable index, and how the hook surfaces them.
• Memory — the shared memory store: the save contract, the project-tag taxonomy, recall, and where it's heading.
• MCP — the MCP server setup: the mcp-router aggregator, the capability families behind it, and Windows pitfalls.