Labcoat biomedical ml hygiene
Skill qchiujunhao/labcoat/skills/labcoat-biomedical-ml-hygiene
Skills, hooks, and CLI checks for safer biomedical AI coding agents.
npx -y skills add qchiujunhao/labcoat --skill labcoat-biomedical-ml-hygieneAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Biomedical ML hygiene for leakage, patient-level splits, subject/sample identifiers, site/batch leakage, label leakage, calibration, external validation, and preprocessing fit-before-split risks. Use when building or reviewing ML code for biomedical data.
SKILL.md
1.3 KB, as published. Nobody here has run it
Biomedical ML Hygiene
Prevent optimistic biomedical ML results.
Decision Ladder
- Identify prediction unit: patient, subject, sample, slide, cell, visit, site, or batch.
- Split by independent unit before preprocessing fit.
- Use group-aware cross-validation when repeated samples exist.
- Check label leakage, site/batch proxies, and duplicate-derived features.
- Require calibration and external validation language for deployment claims.
Allowed
- Flag leakage risks and suggest group-aware splitting.
- Ask for cohort structure and prediction unit.
- Use minimal tests that fail when patient IDs cross splits.
Disallowed
- Do not report clinical performance as validated without external evidence.
- Do not fit scalers/normalizers on the full dataset before splitting.
- Do not ignore patient/site/batch grouping.
Output
Use:
- Prediction unit:
- Leakage risks:
- Split strategy:
- Validation limits:
Read references/leakage-patterns.md for patterns.