Metatronsdoob369/spectral-terrain ? reverse-engineered prompt

Build me a TypeScript project that helps agents work with code in a more structured way, with a clear system for planning, checking, and improving their actions as they go.

I want the app to include a core engine for agent behavior, a way to define contracts for what each agent or step is supposed to do, and some kind of telemetry so I can see what happened during a run. It should also support calibration or scoring so the system can learn from results and compare different approaches.

Please make the codebase organized, with separate areas for core logic, use cases, tests, docs, and a simple store for saved state. If it makes sense, add a few example workflows or sample domains so it’s easy to understand how everything fits together. Use the existing TypeScript setup, keep things clean and readable, and wire up enough tests so the main flows are covered. Look up current docs online if you need to.