rseng-reproducibility

# Computational reproducibility Reproducibility is the ability of someone else - including the author in six months - to regenerate the results from the code and data. It is not one practice but a stack: pinned environments (rseng-reproducible-environments), scripted pipelines (rseng-workflows), versioned data (rseng-data-management) and controlled randomness, assembled so that ONE documented command rebuilds the results. This skill owns the assembly and its verification; the layers below have their own skills. The bar to aim for: a stranger with the repository and the README reproduces the paper's numbers without emailing anyone. ## The one-command bar - Everything scripted, nothing manual: any step a human performs by hand (a click, a copy, an "then edit line 12") is a reproduction failure waiting to happen. Encode the full path from raw data to final figures/tables in a workflow or top-level script (rseng-workflows). - One entry point, documented: `make reproduce`, `snakemake all` or ./run.sh - named in the README with expected runtime and resource needs. Long-running steps get cached intermediates so partial reruns are practical. - Outputs land in generated directories, mapped to the paper: which script makes Figure 3 must be answerable from the repo (a results/README or a figures manifest). - Configuration explicit: every parameter that shaped the published results lives in versioned config files, not command-line lore or notebook cell edits. ##...