Produce a strict post-run analysis bundle from experiment logs, metrics, and artifacts. Use when comparing runs, generating tables or figures, checking whether differences look meaningful, or separating real signal from exploratory noise before writing conclusions.
Analyze first. Recommend second.
Turn raw experiment outputs into an auditable analysis bundle that distinguishes verified signal from exploratory observations and extends the canonical analysis artifact created during the background evidence-review stage.
experiment-visualization for per-run evidence:
../experiment-visualization to backfill them before declaring the figure unavailableresearch/analysis/analysis_report.mdresearch/analysis/stats_appendix.mdresearch/analysis/figure_catalog.mdTable N. ...research/analysis/history/YYYY-MM-DD_<slug>/ before replacing the live interpretation:
analysis_report.mdstats_appendix.mdfigure_catalog.mdanalysis_report.md, preserve the canonical shared schema so the background evidence-review sections stay readable if they already exist.results-report.analysis_report.md, stats_appendix.md, or figure_catalog.md must have a numbered caption immediately above it./external/... or /research/... when linking or embedding figures in Markdown artifacts.research/analysis/history/.The analysis bundle should answer:
assets/analysis-report-template.md.assets/stats-appendix-template.md.assets/figure-catalog-template.md.references/analysis-checklist.md.