Agent-Finance

Produces a decision-support verdict from the filled Excel model + full research cache. Reads out_* DCF outputs, computes a probability-weighted fair-value range (Bull/Base/Bear), derives a 5-level signal (ACCUMULATE / ADD / HOLD / REDUCE / AVOID), and emits a structured action plan across three time horizons. Use last in the pipeline, after excel-financial-model has been validated and opened in Excel (formulas must be calculated). Never labels output as investment advice.

Transforms MCP JSON caches into values.json + sources.json ready for fill_model.py. Applies Bull/Bear deltas. Never invents data — every missing datapoint is marked MISSING:<reason>.