Agentic-tests

[javidsegura]

Agentic-test => an evaluator agent observes another agent's logs (thinking) and result (output) and provides feedback on performance and decides if to fail test, raise warning, or pass (possibly based on a passed rubric)

[javidsegura]

Maybe we could make agents to raise themselves an exception as well