research-harness-audit

To audit raw data files without modifying them, perform a read-only inventory of files, variables, IDs, and missingness to generate a structured data_audit_report.md and traceable logs.

A raw data audit report includes file sizes, formats, variable inventory, missingness patterns, ID ranges, time checks, and a feasibility assessment to ensure data supports planned analysis.

You check data quality by scanning raw files to identify non-missing counts, detect missing patterns, and flag variables with high missingness without loading the entire dataset into memory.

Yes, you can target a specific raw data file using the --file argument, or apply a full-scan across all files under data/raw using the --all option to evaluate feasibility.

Audit logs are generated as logs/audit_YYYYMMDD.log files, capturing inventory details, time checks, and feasibility assessments to ensure full traceability of the raw data review.

Read-only data auditing evaluates whether study design and raw data support planned cleaning steps, but it cannot modify files, correct formatting inconsistencies, or resolve missing values directly.