toolkit

SEO drift baseline + compare · Playbook

Drift thresholds — tune this for your site

These thresholds encode the severity the dialga skill applies when a current snapshot is compared to a baseline. They ship as a starting point — adjust them per site or per migration.

Field Change Severity Verdict contribution
canonical any change (including added/removed) red FAIL
robots (meta robots / X-Robots-Tag) any change red FAIL
h1 text changed or H1 count changed red FAIL
lang (html lang) value changed or removed red FAIL
title text changed yellow REVIEW
description text changed yellow REVIEW
jsonLdTypes type set changed (added/removed type) yellow REVIEW
internalLinks count changed ±25% or more yellow REVIEW
wordCount changed ±50% or more yellow REVIEW
URL added / removed between runs present in one snapshot, absent in the other yellow REVIEW

How to tune: copy this table into your project’s docs and adjust the percentages per template. A news site with daily rewrites may tolerate title drift; a medical site may promote canonical drift to a hard blocker. The skill reads the same fields regardless — only the thresholds move.

Source: orbeetle (title/description/canonical/robots/lang/H1), porygon (JSON-LD types), klink (internal link count), plus word count from the body. The skill composes by shelling out to those CLIs when --urls is used, never by importing their internals.