The score now discriminates: every finding deducts by severity (GEO product-data rules weighted highest), hard caps unchanged, floor at 1 — real-world pages that scored 85–95 "ready" now land where they belong (a live supermarket page: 87 → 69 "needs_work").
New since 0.6.0
- Rendered vs crawler view modes in the extension: two tabs under the score swap the whole popup between what a browsing agent sees and what non-rendering AI crawlers receive — the gap measures your page's JavaScript dependence.
- Metric taxonomy across all surfaces: eight status tiles (readability, validity, completeness, consistency, trust, accessibility, transactability, freshness) with filtering and grouping — status, never a second score.
- Agentic-checkout trust rules researched against UCP (ucp.dev), OpenAI's ACP checkout schema, and Microsoft's MMC docs: policy links, seller identity, return-policy and shipping-details completeness.
- Opt-in
--onlinechecks: product-image size minimums (GS1-grade header sniffing, no imaging deps) and IndexNow key verification. - 13 more sourced rules including GS1 GTIN validation, ISO 4217, schema.org availability vocabulary, and prompt-injection detection (caps the score at 49).
Full details in CHANGELOG.md · rules with sources in docs/RULES.md
uvx --from catalogready-ai catalogready https://your-store.com/products/example