| name | lemma-review |
| description | Review a notebook or analysis for data-science anti-patterns before it's trusted or shared. |
| homepage | https://github.com/tkpratardan/lemma |
| license | MIT |
Review an analysis
Deliver
Return findings ranked by severity, with concrete evidence, consequence, and a
specific correction. Distinguish correctness defects from optional
improvements.
Check
- Reconstruct the requested estimand or prediction target, observation unit,
population, and time boundary.
- Verify source identity, units, joins, missing-data handling, and preservation
of raw inputs.
- Check execution order and whether reported values are reproduced by current
cells.
- Audit splits, preprocessing, metrics, uncertainty, subgroup behavior,
leakage, and causal language as applicable.
- Confirm that tables and figures use honest denominators, labels, scales, and
uncertainty.
Do not rewrite the analysis merely for style, accept a clean-looking notebook
as proof, or report only aggregate metrics when a material group failure is
visible.
For detailed review patterns, read
references/deep-guide.md.
