Braveli convenes a coding-agent council

Field notes

Braveli structures a coding run as a routed discussion among architect, implementer, and reviewer roles rather than a fixed sequence of agents. Its evidence gate compares reported work with the Git worktree and project tests, and separately looks for files that were written but never added to Git. The page illustrates the workflow with a reviewer returning an uncovered test branch to the implementer, while routing history and thread context persist across a restart.