Set up an Engineer / Reviewer pipeline

This builds on the Watcher/Addresser tutorial with a slightly more advanced shape: a Builder that does the actual work, and a Reviewer that checks it before anything is considered finished.

Create a new Strategy. Add a role named Builder, given its own dedicated space, since it needs somewhere to make and commit real changes.

Add a second role, Reviewer, also with its own dedicated space — it needs its own history and checkout, separate from the Builder's, so its judgment stays independent rather than reusing the same context. Its instructions should describe checking the Builder's work against what was actually asked for, and sending anything that falls short back with concrete, specific feedback.

Instantiate the Strategy and give it a project name. From here, every piece of work the Builder finishes gets a second, independent look before it's considered done — the same discipline a real team applies, running on its own. See Forking the built-in Watcher/Addresser and Engineer/Reviewer strategies if you'd rather start from the built-in version of this pair.

Try asking

"Create a Strategy called "Engineer / Reviewer" with a Builder role (workspaceScope: own) and a Reviewer role (workspaceScope: own) that checks the Builder's work and sends feedback back via send_thread_message, then instantiate it for my next feature."

Calls: create_strategy, instantiate_strategy