dezhurny: eval
config: workspace/my_agent/examples/full
case report-on-clean-cards: scenario success | expected success | actual success | pass
case ask-on-ambiguity: scenario needs_input | expected needs_input | actual needs_input | pass
case stop-on-missing-cards: scenario missing_file | expected error | actual error | pass
result: 3/3
evidence: synthetic_script — проверка harness и настроек my_agent, не качество модели
