A full eval suite for your drafting agent

In the Building Agents course you built a pre-contract enquiries agent in eve: it reads a synthetic contract pack (TA6, official copies, search results), drafts enquiries with evidence, and waits for a solicitor's approval before anything is "sent". This project gives that agent what every production agent needs before anyone changes its prompt again: a suite that says, with numbers and intervals, whether a change made it better or worse.

If you didn't build the Building Agents capstone, use your own agent or the title-register reviewer from Agent Foundations in Claude Code. Every milestone has a Claude Code equivalent. Use synthetic documents throughout: never paste real client data into any tool.