Spreadsheet Formula Integrity Linter

Run deterministic formula checks before a workbook generated or edited by an AI agent is accepted.

Scenario and prerequisites

• - Use after workbook contract normalization.
• Provide formulas as redacted strings in the normalized JSON.
• This is a static linter, not an Excel calculation engine.

Input and output

Input is the normalizer output. Output is PASS or HOLD_FOR_HUMAN_REVIEW, with error markers, formula pattern counts and findings.

Installation and execution

Run python scripts/lint_formula_integrity.py --input ../spreadsheet-workbook-contract-normalizer/examples/output.json --output examples/output.json. Review every finding and keep the original report; do not delete a failing formula to obtain a pass.

Recovery and boundaries

The linter flags #REF!, #DIV/0!, #VALUE!, #NAME?, #N/A and formula pattern outliers. It does not calculate formulas, resolve external links or prove business correctness.

Acceptance

The sample must produce a report with PASS and zero error markers. Add #REF! to a formula to produce HOLDFORHUMAN_REVIEW.