Product page enrichment: destination PDPs for full catalog

[SteveSimonson]

Fixes #73

Summary

• Port Kyasi PDP enrichment pattern to Adazo (review snapshot, item blog, FAQ, setup tips)
• Enrich all 64 curated catalog SKUs in Riviera / beauty editorial voice
• Wire productSeo(p, enrichment) + routeMeta so raw HTML includes FAQPage JSON-LD
• Mid-page Buy on Amazon CTA after enrichment; optional render if missing
• Docs: PRODUCT-ENRICHMENT-RULES.md, AGENTS section, npm run content:research

Risk

Low runtime risk (additive UI). Content is original synthesis, not scraped listings. Bundle size grows with enrichment data.

Test plan

• npm run build green
• routeMeta: 64/64 product routes include FAQPage
• Spot-check live PDP after deploy: sections, FAQ expand, Amazon CTA, footer disclosure

Live samples (post-deploy)

• https://adazo.com/product/dyson-airwrap-multi-styler
• https://adazo.com/product/cerave-hydrating-facial-cleanser

[SteveSimonson]

Adversarial review — PR #74 (destination PDPs)

Reviewed wiring, affiliate compliance, security, SEO/routeMeta, build, and spot-checked enrichment quality (CeraVe, Dyson Airwrap, AFFY/jewelry SKUs).

Build / types / SEO wiring — PASS

• npm run build green (tsc -b + Vite; sitemap + routeMeta regenerated).
• npm run lint (oxlint): 0 errors; pre-existing \$ escape warnings in products.ts (not introduced here).
• productSeo(p, enrichment?) correctly optional; FAQPage JSON-LD only when FAQs exist (src/lib/seoData.ts + faqPageJsonLd).
• scripts/route-meta.ts does call productSeo(p, getProductEnrichment(p.slug)) — no drift risk for raw-HTML injection.
• Worker routeMeta: 64/64 product routes include FAQPage.
• Coverage: 64 enrichments for 64 catalog SKUs; getProductEnrichment optional at render; mid-page Buy CTA only when enrichment present (page still works without).

Affiliate compliance — PASS

• Associate tag lives only in affiliateUrl / amazonProductUrl query (tag= in href).
• Not printed in ProductEnrichment.tsx UI, ASIN chrome, or mid-page commission banners.
• Buy CTAs retain rel="noopener noreferrer sponsored" (primary, secondary, after-enrichment).
• Footer disclosure path unchanged (BRAND.affiliateDisclosure in Layout).

Security — PASS

• No dangerouslySetInnerHTML on enrichment fields.
• Verdict, blog bodies, FAQs, tips rendered as React text (Paragraphs splits on \n\n only). Safe against HTML injection from enrichment data.

---

Enrichment originality / quality — FAIL (blocking)

Not Amazon paste (good), but the corpus is template-machined with catalog field injection. That violates this PR’s own rules in docs/PRODUCT-ENRICHMENT-RULES.md:

Duplicating the same blog across similar SKUs with only the name swapped

1. Systemic FAQ grammar bug (64/64)

Skip FAQ answers are missing separators:

You need prescription-strength actives only Also reconsider if you refuse any fragrance...

Same join bug on Dyson, Olaplex, jewelry, etc. Ship-visible broken English on every PDP FAQ.

2. HTML entities left literal (")

Jewelry/gold copy embeds raw Amazon entities, e.g. AFFY / Jewelry Affairs:

27" L, 22"

React text will show " to shoppers — not a curly quote. Looks scraped/broken.

3. Spot checks (not house editorial)

SKU	Issue
CeraVe (cerave-hydrating-facial-cleanser)	Shared skincare shell; love chip "Highlight shoppers notice: …" is raw catalog glue; skip FAQ grammar break; generic bestFor/skipIf recycled with La Roche / Neutrogena
Dyson Airwrap (dyson-airwrap-multi-styler)	Flagship still gets generic tools template: “collecting bottles for the algorithm”, “humid bathrooms / skin or scalp”, no Coanda/weight/hair-type/attachment-kit specificity. Verdict is stock one-liner (“look considered, feel finished…”)
AFFY / Houston / CHOW jewelry	Full truncated Amazon titles repeated every section; beauty-counter metaphors on fine jewelry; love themes like "Big-ticket luxury jewelry" / "CHOW SANG SANG piece"; dek/title truncation with …

Template phrase density (illustrative counts in productEnrichments.ts):

• Also reconsider — 64
• Highlight shoppers notice: — 64
• beauty like wardrobe — 33
• In a category full of twins — 31
• crowded shelf of maybes — 25
• You are not collecting bottles for the algorithm — 31

4. Double-period / glue artifacts

Bodies like …A statement tool for the vanity.. (Dyson) show string-concatenation without cleanup.

---

Missing vs stated Kyasi / rules pattern — nits (non-blocking)

Expected	Status
Review snapshot + bestFor/skipIf	Present
Item blog 3–5 sections	Present (structure OK; substance thin)
FAQ ≥4 + FAQPage JSON-LD	Present (64/64)
Setup tips	Present
Research notes (names only)	Present + UI disclosure
Mid-page Buy on Amazon	Present
Compared on this site (Prefer)	Not in enrichment type/UI — PDP still has similarProducts / also-like grids, so partial coverage
Fit-vs-sibling FAQ (rules)	Rare / mostly generic “check Amazon” FAQs

Bundle note: main JS ~818 kB min (~170 kB gzip) after enrichment data — called out in Vite; acceptable for v1 but watch growth.

---

What to fix before merge

Must-fix (content gate):

1. Repair FAQ skip answers globally — punctuation/conjunction between skipIf items (not bare Also reconsider).
2. Decode / strip HTML entities in names used in titles/FAQs (" → "); prefer short display names, not full Amazon title strings.
3. De-templatize flagships + category representatives (at minimum Dyson, one luxury, one skincare, one jewelry): product-specific caveats, living-with notes, and verdicts that do not reuse the same shell with {name} swapped. Strip "Highlight shoppers notice:" glue.
4. Category-aware shells — jewelry/watches must not say “bottles,” “barrier,” or “beauty like wardrobe.”

Nice-to-have (can follow-up):

5. Optional comparedOnSite: { slug, reason }[] + UI block per rules Prefer.
6. Code-split or lazy-load enrichment module if bundle pressure matters.
7. Lint-time assertion: no Also reconsider / &#\w+; / repeated identical blog body across >N SKUs.

Architecture notes (no change required)

Wiring matches the skill playbook: types → data → ProductEnrichmentSections → optional render → productSeo(..., enrichment) in app and routeMeta. Compliance and XSS posture are solid. This is a content quality reject, not a framework reject.

VERDICT: REQUEST_CHANGES

[SteveSimonson]

Re-review — PR #74 (post content rewrite)

Checked against the prior REQUEST_CHANGES must-fixes, spot-sampled CeraVe, Dyson Airwrap, Dazzlingrock (jewelry), Gucci / LV (handbags), and re-ran npm run build.

Prior blockers — status

Prior must-fix	Status
FAQ skip join bug (only Also / missing separators)	FIXED — zero hits for only Also / Also only
Literal HTML entities (", etc.)	FIXED — zero hits
"Highlight shoppers notice:" catalog glue	FIXED — zero hits
Product features in love + blog (flagships)	IMPROVED — CeraVe ceramides / non-foaming / fragrance-free land in love + “What separates…”; Dyson multi-styler + attachments similarly present
Category-aware shells (no bottles/barrier on jewelry)	IMPROVED — jewelry/handbags use fine-goods / seller-diligence framing, not skincare “bottles” metaphors
Unique who-bodies	PARTIAL — full first bodies are unique via product name, but normalized templates are 33/64 (skincare flagships are product-specific; jewelry / gold / watches / handbags still share category shells)

Build

• npm run build green (sitemap 91 URLs / 64 products, tsc -b, Vite client build OK).

Spot checks

CeraVe (cerave-hydrating-facial-cleanser)

• Love is product-real: non-foaming hydrating formula, ceramides, fragrance-free, material story.
• Who job is SKU-specific (dermatologist-developed hydrating cleanser…).
• Still template frame (“not for people collecting random Amazon tabs…”) but substance is no longer empty glue.

Dyson Airwrap (dyson-airwrap-multi-styler)

• Love/blog cite multi-styler system + curl/wave/smooth attachments.
• Caveats about premium price / weekly use / learning curve are reasonable.
• Still light on Coanda / heat / hair-type / kit variance depth — acceptable as nit, not the old generic “tools” emptiness.

Jewelry (dazzlingrock-necklace-lk8m)

• Prior beauty-counter / bottles language gone.
• Love still thin (Big-ticket luxury jewelry, Dazzlingrock piece).
• Who job shared with other $1k+ fine jewelry SKUs; long truncated Amazon titles still echo through FAQ/blog.

Handbags (gucci-tote-kfwh, louis-vuitton-pochette-4blf)

• Pre-loved condition / authenticity diligence appears in who + living-with.
• Love includes brand + pre-loved/resale — better than pure generic, still category-cloned across LV/Gucci/Prada.

---

Remaining nits (non-blocking for content architecture; fix preferred before merge)

High-priority glue (still ship-visible on ~every PDP who-section):

1. Missing space: vanity andwardrobe — 64/64 entries. Looks like string concat without a separator (and + wardrobe).
2. Trailing double periods on job clauses: e.g. …dry skin.., …on Amazon.., …for the vanity.. — same class of glue artifact called out last review.

Medium (category template residue):

3. Awkward skip FAQ phrasing: handbags system / jewelry system / tools system / skincare system you love and will not rotate — systems language is odd for a necklace or tote.
4. Universal FAQ closer still category-blind: space, skin, scent preference, or sport on handbags/jewelry/watches.
5. Luxury jewelry/gold/watches/handbags who-bodies remain fill-in-the-blank shells (name + brand swapped). Fine for long-tail SKUs; not “house editorial” for hero luxury.
6. Truncated Amazon titles (…) still used as prose subjects in FAQ questions and section openers — prefer short display names.

Nice-to-have (unchanged from prior):

7. Optional comparedOnSite / fit-vs-sibling FAQ still sparse.
8. Bundle size (enrichment data in main chunk) — watch later; not a merge blocker.

---

Assessment

The rewrite clears the blocking content gate from the last review: broken FAQ grammar, HTML entities, Highlight glue, and the worst cross-category shell misuse are gone. Flagship skincare/tools now carry real product features in love/blog. Wiring/compliance/security findings from the first pass still stand (no re-regression spotted).

Left work is global polish + deeper uniqueness on luxury clusters, not another architecture rewrite. The two glue typos (andwardrobe, ..) should be a mechanical fix across the generator/data before merge if easy.

VERDICT: APPROVE_WITH_NITS