lloyal-ai/hdk ? reverse-engineered prompt

Build me a working starter project for this harness kit that I can run locally and poke at from the terminal.

I want a simple TypeScript app that shows how a harness owns a model, spawns a few agents from the same context, and keeps their work organized in one place. Make it feel like a real demo, not just a placeholder, so I can start it, ask for something like a plan or short research task, and see the agents share context and finish with a clear result.

Please wire up whatever is needed in this repo so the main pieces actually work together, with a clean example app, basic scripts, and a straightforward README for running it. If you need current docs for any of the model or runtime pieces, look them up online and keep the setup as simple as possible. I care most about the flow being easy to understand, the code being clean, and the example making the idea of a harness obvious right away.