← Agent Transactable

Public Scoring Methodology

We only score what we can evidence. UNKNOWN is never silently converted into FAIL.

Status values

Dimension score

Weighted verified status values divided by verified test weight × 100.

Overall weights

Confidence

Evidence coverage and sampled-page depth determine LOW, MEDIUM or HIGH confidence. Lower confidence widens the displayed readiness range. This is a methodological range, not a statistical 95% confidence interval.

Crawl posture

V3.1 uses a maximum six-page HTML crawl budget, same-origin preference, robots.txt awareness, timeouts and response-size limits. It does not perform authenticated or transactional actions.

Versioning

Scanner 3.1.0 · KB-2026.07.22 · SM-1.0