Fix the raw-HTML JSON-LD extraction command — it is line-based and manufactures fake hydration divergences - repo: wakqasahmed/ai-visibility-skills - category: gap - severity: high - evidence: the raw pass and the hydrated pass use different extraction engines. - Raw pass (line-based `grep`, truncates at the first `<` and cannot span lines): `skills/ai-visibility/schema-markup-audit/references/checks.md:10` and `skills/ai-visibility/ai-visibility-audit/references/checks.md:40` — `curl -s "$URL" | grep -oiE ']+application/ld\+json[^>]*>[^<]*' | sed 's/^]*>//' | python3 -m json.tool` - Hydrated pass (multi-line Python regex, `re.S`): `skills/ai-visibility/schema-markup-audit/references/checks.md:36` and `skills/ai-visibility/ai-visibility-audit/references/checks.md:67` — `re.findall(r']+application/ld\+json[^>]*>(.*?)', html, re.S | re.I)` - Reproduced on a fully server-rendered page with pretty-printed JSON-LD (what Yoast, Rank Math, Shopify and most SSR frameworks emit): ``` === pack's raw-pass extraction command (checks.md:10) === Expecting value: line 2 column 1 (char 1) exit=1 === pack's @type listing command (checks.md:65) === --- done, types printed above (empty means none found) === hydrated-pass regex (checks.md:36) on the same file === 1 block(s) found by the hydrated-pass regex ``` - Second, separate defect in the same file: `skills/ai-visibility/schema-markup-audit/references/checks.md:65` uses `grep -oE '

wakqasahmed analysis page

Waqas Ahmed