medical-structuring

Use this when cleaned medical data has patient_id/subject_id or multiple records per patient and needs patient-level validation, passed cohort filtering…

Use this when baseline cleaning is insufficient and Codex needs to generate task-specific cleaning code for medium/high-risk columns, validate shape/diff…

Use this whenever the user has scanned records, images, PDFs, OCR text, raw clinical documents, or asks to extract text before cleaning.

Use this when Codex should spawn subagents or run parallel reviews for data-cleaning criteria after a skill finishes, after validation reports are produced…

Use this whenever the user asks Codex to clean medical, clinical, EHR, MIMIC-like, tabular, OCR-derived, or diagnosis/label-building data.

Use this whenever a medical cleaning task needs diagnosis labels, ICD mapping, ML labels, cohort labels, mortality/readmission/diagnostic targets, or leakage…

Use this whenever a medical data task needs task-oriented column selection, feature filtering, schema profiling, PHI-aware field selection, or reducing a wide…

Use this at the start of any medical data cleaning task with a file, directory, mixed dataset, PDF/image/table collection, or unknown input.

Use this after each medical data cleaning run, validation failure, successful experiment, or when Codex wrote task-specific code that may be reusable.

Use this whenever a medical data cleaning task creates or changes cleaned.csv, label columns, validation criteria, hooks, or reports.

Use this after cleaning/labeling to produce leakage-safe ML-ready datasets from medical tables.