What problem does it solve? Persistent-homology findings on trajectory data need conventional baselines for interpretation and methodological comparison, but poorly designed baselines introduce leakage, sample mismatch, and unsupported claims. This Skill enforces a disciplined procedure for building clustering, classification, forecasting, anomaly-detection, and survival baselines that legitimately contextualize a TDA result. ## Core Features & Use Cases - Baseline Classification: Names the baseline class (descriptive, predictive, robustness, negative-control, interpretability) first, which determines the metrics and comparison language allowed. - Leakage and Comparability Audits: Mandates severe temporal-leakage checks and sample-provenance reconciliation against the PH pipeline before any comparison is made. - Provenance-Tracked Outputs: Requires a structured output record covering split, seed, metrics, limitations, and the PH result path the baseline contextualizes. - Use Case: A researcher comparing PH-derived market regimes against a standard clustering baseline uses this Skill to align cohort filters, verify no post-outcome features leak into predictors, and record seeds and splits so the comparison supports a paper claim. ## Quick Start Ask the assistant to build a conventional clustering baseline that contextualizes a specific persistent-homology trajectory result, including leakage checks and a provenance-tracked output record.
