Strategy Saver

To save trading strategies to a JSONL catalog, the Strategy Saver persists approved strategies with extended metadata including optimized parameters, in-sample and out-of-sample metrics, grid statistics, and regime filters. It writes structured records to output/strategies.jsonl for reproducibility.

Logging lessons learned for discarded trading strategies is handled by updating the lessons_learned.md file. The skill records generalized observations from discarded strategies to inform and guide future explorations, ensuring knowledge capture from failed hypotheses.

Archiving trading strategies enforces an expanded output schema with fields like id_estrategia, fecha_creacion, hipotesis, parametros_optimizados, metricas_is, metricas_oos, grid_stats, filtro_regimen, and deployed. Evaluation fields remain null for external evaluators.

A trading strategy catalog handles regime filters and grid search statistics by enforcing an expanded JSONL output schema. The skill saves these extended metadata fields alongside optimized parameters and performance metrics to ensure full provenance and auditability.

Yes, you can leave evaluation fields null in a trading strategy JSONL catalog. The skill enforces an expanded output schema that intentionally leaves evaluation fields null, allowing external evaluators to independently assess the archived strategies.