The sample report, page by page.
The output is a document you can send. One score for the repository, every dimension scored underneath it, and each finding priced by what clearing it costs. Read the real one below, in print form and on a phone, before you send anyone a link.
One repository, two ways to read it.
Same payload behind both: the print document you attach to a proposal, and the web version for whoever reads everything on a phone. This is a real scan of Grove, a production codebase, health score 78.1. Only local file paths and the repository address were removed. The findings are as the engine reported them.
Page one is a bug in your code, already corrected.
The cover is not a stock illustration. It is your highest-severity mechanical finding shown before and after, so the first thing anyone reads is evidence. The seven pages after it narrow from one number to a named line.
- 01CoverYour highest-severity mechanical finding, before and after, corrected and verified at no cost.
- 02Executive summaryScore, scope, severity breakdown and the split by remediation tier, readable in ten seconds.
- 03Clustered scorecardForty dimensions rolled into seven families, each with its weakest and strongest member.
- 04Dimension detailEvery dimension scored out of ten, sorted worst first, on one page.
- 05Findings by costThe free ones lead, then bounded work, then decisions. Each carries evidence, a score delta and a disclosed cap.
- 06ComplianceNine frameworks, SOC 2 through the EU AI Act, with control citations backed by named findings.
- 07Control detailThe worst-scoring framework, article by article. Hit rows open into the findings that cite them, clean rows say so.
- 08MethodologyHow the numbers are made: syntax-tree analysis, three tiers, and a check that the changed file still parses.
print, web and phone are three renderers over one payload · nothing is written for the document that the scan did not produce