Task 5

Enter a ticker. We run the fundamental agent (Task 3, from the latest 10-K) and the technical agent (Task 4, from as-of indicator readings) over one common window, then an LLM arbiter decides how to combine them — picking one policy from a fixed menu (AND / OR / weighted / gated / defer). The arbiter sees each agent's reasoning but not its realized returns, so the combine policy can't be fit to the test window. The combined position is then backtested, lookahead-free, against buy-and-hold and the S&P 500.