Skip to content

Universal SEO skill for Claude Code. 26 sub-skills + 19 sub-agents covering technical SEO, E-E-A-T, schema, GEO/AEO, agent readiness (Lighthouse Agentic Browsing, WebMCP, llms.txt), backlinks, local SEO, e-commerce, international SEO, Google APIs, and PDF/Excel reporting. 9 optional extensions for live SEO data.

Merge pull request #393 from xunguangzlj-cloud/fix/crux-zero-baseline-trends

Agrici Danielde8095d19c2b
diff --cc CHANGELOG.md
index e6c4f34,f204169..4eca161
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@@ -9,55 -9,8 +9,57 @@@ and this project adheres to [Semantic V
### Fixed
+- **Merchant validation accepts a list of `priceSpecification` objects.**
+ Schema.org allows one price specification or several. A list crashed
+ `schema_ecommerce_validate.py` with `AttributeError` before any finding
+ was returned. Loyalty-tier detection now checks each object in the list.
+- **An og:image query no longer sets the page encoding.** `charset=` inside
+ another meta attribute was treated as `<meta charset>`. A later
+ `<meta charset="utf-8">` was ignored, so a canonical
+ `https://example.com/café` was decoded as `https://example.com/café`.
+ A real `charset` attribute, or an `http-equiv="content-type"` pragma,
+ is still the encoding.
+- **A shorter filler phrase no longer scores inside a longer one.**
+ `let's dive into` also matched `let's dive in`, so one sentence was two
+ hits. On a 320-token page that moved `filler_score` from 25 to 50,
+ `overall_quality` from 47 to 41, and raised the filler flag.
+- **Sitemap discovery keeps a robots.txt URL when a comment follows it.**
+ `Sitemap: https://example.com/products.xml # catalog` was dropped, so that
+ URL was never fetched and only the common paths were checked. A `#` comment
+ runs to the end of the line (Google's robots.txt syntax, and RFC 9309), and
+ a leading UTF-8 BOM is not part of the first field.
+- **IndexNow key check compares hosts, not URL substrings.**
+ `verify_key_published` treated `example.com` as present in
+ `https://notexample.com/...` and rejected `Example.COM` against
+ `https://example.com/...`. The key file host is compared as a hostname,
+ before the file is fetched.
+- **GBP verified status uses DataForSEO `is_claimed` (#334).** The checklist
+ said verification is not exposed and could be inferred from profile
+ completeness or a Maps listing. `is_claimed` is that field. The 50-point
+ scale is unchanged. An absent value is Unknown (manual check required).
+- **`/seo` keeps the command word (#342).** The hub skill said `/seo $1 $2`.
+ Claude Code indexes arguments from 0 (`$0` is the first word, `$1` the
+ second; https://code.claude.com/docs/en/skills), so `/seo audit <url>`
+ loaded as `/seo <url> $2` and dropped `audit`. The line now uses
+ `$ARGUMENTS`, which is the full input.
+- **News sitemap validation keeps the 50,000 URL limit (#335).** The skill
+ told the model to replace that check with a 1,000 cap whenever the `news:`
+ namespace was present. Google caps `<news:news>` tags at 1,000 and still
+ applies the general sitemap limits
+ (https://developers.google.com/search/docs/crawling-indexing/sitemaps/news-sitemap).
+- Recognize UTF-8 BOM XML sitemaps during format detection.
+- Preserve brackets around IPv6 hosts when normalizing drift baseline URLs.
+- Recognize the `none` robots directive as an indexing restriction in drift comparisons.
+- Avoid flagging a brand suffix when its text only occurs inside a longer word in the description.
+- Resolve relative image and link URLs using the first document base href.
+- Stop Indexing API batches after a quota response instead of sending more requests.
+- Report IndexNow submission errors in text mode without raising a KeyError.
+- GA4 `top-pages` reports now surface failed landing-page queries instead of
+ returning an empty successful report when the daily query succeeded.
+- `matomo_report.py top-pages` now reports a failed entry-page query as an
+ error and exits nonzero instead of returning an empty success (#395).
+ - CrUX History reports an increase from a zero CLS baseline as degrading, with
+ an undefined percentage (`null` in JSON), instead of stable with a 0% change.
## [2.5.0] - 2026-10-22