results-analysis

To analyze ML experiment results, provide your raw experiment artifacts for validation. The skill inventories metrics, training curves, and baselines to enforce a reproducible analysis workflow, outputting a strict bundle with statistics and figures.

Validating experimental pipelines requires inventorying metric tables, seeds, and baselines from your artifacts. This skill enforces evidence-first rigor by validating evaluation metadata and producing a strict, reproducible analysis bundle with explicit caveats when data is missing.

Compute confidence intervals and effect sizes by feeding evaluation metrics into this analysis workflow. It performs descriptive and inferential statistics, applying appropriate tests with corrections, and documents the results in a generated stats-appendix.md file.

Yes, you can create a figure catalog for publication-ready plots. The workflow generates figures from your ML results and outputs a figure-catalog.md file that enforces caption requirements and tracks all visual artifacts.

Analyzing ML experiment results with missing data triggers explicit caveats and blockers. The skill enforces a strict validation process during artifact inventory, ensuring traceability and preventing incomplete analyses from proceeding silently.