This project provides a multi-layer harness for converting open-ended research goals into protocolized execution with durable artifacts and evaluable evidence surfaces. It structures research as units with declared inputs/outputs, acceptance criteria, and guardrails. The system combines a workflow protocol layer, execution ledger, evidence loop for run validation, and a capability surface with reusable AI-driven skills. It is designed to make AI-assisted research auditable, resumable, and improvable through visible contracts and structured intermediate artifacts.