excel-hygienelisted

# Excel review — a second pair of eyes (taxonomy-driven) ## What to do 1. **Run the script** for precise, reproducible **mechanical** checks (openpyxl, no LibreOffice). It needs `openpyxl` — if the import fails, `pip install openpyxl` first: ``` python3 scripts/check_excel.py <path/to/file.xlsx> ``` 2. **Then review against the taxonomy below.** The script is a **floor, not a ceiling.** Read each sheet and reason through every class, using your general business/finance/spreadsheet knowledge. Goal (per Panko 2008): be **expansive — suggest issues to check, don't just confirm a fixed list.** ## The error taxonomy — your review framework (Panko–Halverson, revised 2008) Work top-down. Each class has a different **detection rate** — the ones lowest for the human eye (omission/lapses and qualitative) are where you must look hardest; that's where real value is added. ### 1. Violations vs Errors - **Violation** — breaks a policy or compliance rule (even if the math is right): a banned/hardcoded rate, a missing required control/disclaimer/sign-off, a regulated calc done off-standard. *You usually can't know client policy — flag "confirm against your standards."* - **Error** — inadvertent. Everything below. ### 2. Qualitative vs Quantitative - **Qualitative** — the number is *not* wrong today, but the design is risky (latent). Often the most dangerous because it survives review. Examples: - **Hardcoding / "jamming"** — a constant buried in a formula (`=B2*