A dictionary of several thousand objects accumulates code nobody has parsed, colours nobody chose and labels nobody translated. This family of tools measures it, and records every measurement as a run.
Parse, then run
- Every body parsed Each stored code body is parsed in its declared language and never executed. On the reference dictionary: 3,981 bodies, 29 failing, 150 fragments unresolved, 633 declared in the wrong language.
- An object run on a budget An object executes through six probe stages on a virtual thread with a time budget, always rolled back. The dictionary run loops it, a thousand objects by default.
- Every lint is a run Recorded with a Markdown report, environments enforced, five failure classes, and the run document rendered as a page.
Census
- Presentation Grid widths, hidden columns and group bands; a colour inventory checked against the curated palette with a consolidation rollup.
- One contract for code columns All 126 code columns have one oracle for who evaluates each and where; 2,730 hide expressions were measured, twelve SQL and the rest expression language.
- Translation and security A translation census per area and language, and a ten-rule security lint.
The static review of September measured statements. This measures everything else, and it is the same pattern: never execute, always record.