Identity
SheetBrain Detector Set v0.4 is the current detector bundle referenced by SheetBrain audit reports under the Spreadsheet Trust Standard.
| Standard version | v0.1 |
|---|---|
| Detector set name | SheetBrain Detector Set v0.4 |
| Detector set version | v0.4 |
| Release date | 2026-05-05 |
| Status | Published and immutable |
Content hash:
67bfcab7558fb1ddd4c61e33b12b83a92fc4071c0a707742ae0546bc071ad760
Once published, this detector set version is not silently edited. Detector additions, removals, severity changes, and verdict-policy changes require a new detector set version.
Included detectors
| Issue type | Severity | What it checks |
|---|---|---|
broken_reference |
error | Flags active formula failures such as #REF! outputs and lookup formulas that request a column or row outside the referenced range. |
visible_formula_error |
error | Flags formula cells returning visible error values: #N/A, #VALUE!, #DIV/0!, #NAME?, #NUM!, #NULL!, or #ERROR!. Literal text values are not flagged. |
mixed_text_number |
warning | Flags columns that mix numeric and text-stored values, a common source of silent SUMIFS, COUNTIFS, and AVERAGEIFS mistakes. |
duplicate_lookup_key |
warning | Flags lookup-key columns with duplicate values when used by VLOOKUP or XLOOKUP, because those formulas silently return the first matching row. |
inconsistent_formula_pattern |
warning | Flags one-off formula deviations inside same-column contiguous formula runs when a conservative repeated pattern is clear. |
missing_formula_pattern |
warning | Flags hardcoded or blank cells that sit inside a repeated formula column when the two formulas immediately above and two immediately below all share the same normalized pattern. |
Verdict thresholds
| Verdict | Threshold |
|---|---|
| Critical | errorCount > 0 |
| Has open issues | errorCount === 0 && warningCount > 0 |
| Trusted | errorCount === 0 && warningCount === 0 |
Informational findings do not affect the verdict in v0.4. The formula-pattern detectors emit warnings only, so they can move a report to Has open issues but not Critical.
Scope limits
Detector Set v0.4 evaluates the bounded data sent to the SheetBrain backend for the audit. Public reports describe this as data cells evaluated (headers excluded).
- Maximum evaluated rows per sheet: 1,000 data rows.
- Maximum evaluated columns per sheet: 50 columns.
- Headers are used for context but are not counted as evaluated data cells.
- The scan evaluates the workbook snapshot sent at audit-generation time, not future changes.
Limitations
- Does not verify source data accuracy, completeness, authorization, or business assumptions.
- Does not certify that a spreadsheet is suitable for a specific legal, tax, audit, investment, or procurement decision.
- Does not validate external data sources, API integrations, named-range references outside the evaluated scope, or third-party add-on output.
- Does not detect schema drift or formula drift after a column move or rename. SheetBrain Repair handles those separately with a schema snapshot and AI-assisted rewrites.
- Formula-pattern checks are same-column only in v0.4 and intentionally skip ARRAYFORMULA, INDIRECT, OFFSET, formulas without A1 references, unsupported reference styles, and contexts without a clear local pattern.
- The missing-formula detector requires exactly two matching formulas above and two matching formulas below. It skips sheet starts/ends, full blank rows, likely subtotal/footer rows, and mismatched neighbor patterns.
v0.4 changelog
- Added
missing_formula_patternas a warning detector for conservative same-column formula holes. - Kept
inconsistent_formula_patternfrom v0.3 as a warning detector for formula outliers. - Kept verdict thresholds, scope limits, and Standard version unchanged from v0.3.
Prior detector sets: SheetBrain Detector Set v0.3, SheetBrain Detector Set v0.2, and SheetBrain Detector Set v0.1.