Module cli 

Eval harness CLI surface (EH5).

Wires the runner + persistence + regression into the eval subcommand: run executes a suite through a tool-less orchestrator, scores it, persists results, compares to a baseline, prints a report, and returns a non-zero exit code on regression (a CI gate). list/baseline inspect stored data.

Functions

run_eval

Execute an eval subcommand. Returns a process exit code (0 ok, 1 regression, 2 usage/IO error). Never panics.