Run the eval: every arm × every trial against the real claude CLI, with the
metric computed per run and aggregated per arm. Requires claude on PATH and
working model auth (e.g. ANTHROPIC_API_KEY). Thin wrapper over
runEvalWith with the real agent runner.
Run the eval: every arm × every trial against the real
claudeCLI, with the metric computed per run and aggregated per arm. Requiresclaudeon PATH and working model auth (e.g.ANTHROPIC_API_KEY). Thin wrapper overrunEvalWithwith the real agent runner.