Product page enrichment: destination PDPs for full catalog

AGENTS.md

@@ -1,5 +1,20 @@
 # Adazo agent notes
 
+## Product thesis
+
+**Women’s beauty & Riviera finish.** Discover on Adazo; buy on Amazon.
+
+## Product page enrichment
+
+Every catalog SKU should have long-form enrichment (review snapshot, item blog, FAQ, setup tips) so PDPs rank and convert as destinations—not thin affiliate hops.
+
+- **Rules:** [`docs/PRODUCT-ENRICHMENT-RULES.md`](docs/PRODUCT-ENRICHMENT-RULES.md)
+- **Skill:** `~/.grok/skills/product-page-enrichment` (`/product-page-enrichment`)
+- **Data:** `src/data/productEnrichments.ts` via `getProductEnrichment(slug)`
+- **UI:** `src/components/ProductEnrichment.tsx` on `Product.tsx`
+- **Research:** TinyFish via `npm run content:research` → synthesize original Adazo prose (warm Riviera; never clinical cure claims; never paste scraped reviews)
+- **SEO:** FAQPage JSON-LD when FAQs exist (`productSeo` in `seoData.ts`); `scripts/route-meta.ts` must call `productSeo` **with** enrichment
+
 ## Persona modeling / campaign images
 
 **Read and obey** [`docs/persona-portfolio-rules.md`](docs/persona-portfolio-rules.md) before generating or replacing any persona avatar, house campaign, or World Edit travel still.

docs/PRODUCT-ENRICHMENT-RULES.md

@@ -0,0 +1,138 @@
+# Product page enrichment rules (network)
+
+Reusable rule set for **Amazon-affiliate catalog PDPs** (Kyasi, Adazo, iBamboo, Mr Cuts, SurgiSyn, Chrome Cave, …). Goal: every product page is a **destination** for searchers—not a thin affiliate hop.
+
+**Agent skill (full deploy playbook for other repos):**  
+`~/.grok/skills/product-page-enrichment/SKILL.md` — slash `/product-page-enrichment`
+
+## Thesis
+
+Discover and decide **here**. Buy **on Amazon**. We own judgment, synthesis, and structure. Amazon owns inventory, price, shipping, and checkout.
+
+## Non-negotiables
+
+1. **Original prose only.** Never copy Amazon bullets, competitor reviews, or RTINGS/Wirecutter text. Research → synthesize → write in house voice.
+2. **Honest caveats.** Always include `skipIf` / caveats. Trust compounds conversion.
+3. **Footer affiliate disclosure only.** No Associate-tag UI, ASIN badges, or “we earn commission” banners mid-page. Keep Buy on Amazon CTAs.
+4. **Tag stays on links** (`tag=…` in href only). Never print the tracking ID on-page.
+5. **Ratings are approximate.** Star ratings and review counts from public consensus / listing snapshots; label “approx.” and tell shoppers to confirm on Amazon.
+6. **No medical / safety guarantees** on care/medical catalogs. Editorial, not clinical.
+7. **Sources optional, never paywalled dumps.** May list “research notes” (site names) without pasting full articles.
+8. **Enrichment is optional at render.** Missing enrichment → page still works with core catalog fields.
+
+## Content blocks (priority order)
+
+| Block | Required | Purpose | SEO / engagement |
+|-------|----------|---------|------------------|
+| **Review snapshot** | Yes | Verdict + love themes + caveats | Answer “is it good?” above the fold-ish |
+| **Best for / Skip if** | Yes | Fit filter in 2 lists | Reduces bounce, builds trust |
+| **Item blog** (long-form) | Yes | 3–5 original sections | Ranking depth, dwell time |
+| **FAQ** | Yes (≥4) | Buyer objections | FAQPage JSON-LD |
+| **Setup / first-week tips** | Prefer | Post-click utility | Engagement, return visits |
+| **Compared on this site** | Prefer | Links to sibling SKUs | Internal links, kit completion |
+| **Specs table** | Core catalog | Confirm details | Product schema |
+| **Buy on Amazon** | Core | Conversion | Multiple CTAs, sticky mobile |
+
+### Review snapshot schema
+
+```ts
+reviewSnapshot: {
+  verdict: string            // 1–2 sentences, house voice
+  love: string[]             // 3–5 buyer/expert themes (paraphrased)
+  caveats: string[]          // 2–4 honest caveats
+  bestFor: string[]          // 2–4 personas / use cases
+  skipIf: string[]           // 2–4 “not for you if…”
+  ratingNote?: string        // e.g. “~4.6 from tens of thousands of Amazon ratings”
+}
+```
+
+### Item blog schema
+
+```ts
+blog: {
+  title: string              // H2-worthy, search-aware, not clickbait spam
+  dek: string                // 1-line deck under title
+  sections: {
+    heading: string
+    body: string             // 80–180 words, paragraphs OK (use \n\n)
+  }[]                        // 3–5 sections
+}
+```
+
+**Blog section pattern (default):**
+
+1. Who this is really for  
+2. What it does better (or differently) than the obvious alternatives  
+3. Living with it (comfort, battery, app, quirks)  
+4. Trade-offs / what we checked before listing  
+5. Optional: care, setup, or “when to wait for a deal”
+
+### FAQ rules
+
+- ≥4, ≤8 questions.
+- Answers 40–90 words, plain language.
+- At least one FAQ about **price variance / Amazon listing**, one about **who should skip**, one about **fit vs a sibling** when catalog has a peer.
+- No keyword-stuffed Qs. Natural questions real shoppers ask.
+
+## Voice by brand (swap per site)
+
+| Site | Voice |
+|------|--------|
+| **Kyasi** | Don Draper × premium tech: short, confident, no fluff. Desire first. |
+| **Adazo** | Warm Riviera / beauty editorial. Sensory, never clinical hype. |
+| **iBamboo** | Calm natural living. Material truth over gadget hype. |
+| **Mr Cuts** | Groomed, fit, sharp. Male lifestyle without bro-speak. |
+| **SurgiSyn** | Calm caregiver. Clarity over cleverness. Not medical advice. |
+| **Chrome Cave** | Gamer den competence. Spec-smart, not toxic. |
+
+## Research pipeline (TinyFish)
+
+Free path preferred:
+
+1. **`search`** (3 queries per SKU):  
+   - `{name} review pros cons worth buying`  
+   - `{name} RTINGS Wirecutter CNET review` (or category peers)  
+   - `{name} amazon customer reviews common complaints praise`
+2. **`fetch_content`** top 4–8 editorial URLs (prefer RTINGS, Wirecutter, CNET, Verge, PCMag, SoundGuys, etc.).  
+   Amazon PDP often bot-blocks—optional, never required.
+3. **Synthesize** into schema. Cite mental sources; do not paste.
+4. **Store** as static data (`productEnrichments.ts` or JSON) keyed by `productId` / `slug`.
+5. **Refresh** when listing changes, major model revision, or quarterly content pass.
+
+Metered automation (`run_web_automation`) only when interaction is required (e.g. expand review facets). Prefer search + fetch.
+
+## Quality bar (ship checklist)
+
+- [ ] Every SKU with an ASIN has enrichment **or** explicit `todo` backlog issue  
+- [ ] Verdict does not repeat tagline verbatim  
+- [ ] Caveats exist and are specific (not “price may vary” only)  
+- [ ] Blog ≥3 sections (prefer 4–5); flagship SKUs ≥250 words body, accessories ≥120  
+- [ ] FAQ ≥4 with FAQPage JSON-LD  
+- [ ] Internal links to 1–2 similar products where catalog allows  
+- [ ] No Associate tag / ASIN chrome in UI  
+- [ ] Buy on Amazon still primary CTA  
+- [ ] Build + SEO verify pass after content ship  
+
+## What not to do
+
+- Thin pages: title + Amazon button + scraped bullets  
+- Fake “Verified Purchase” quotes  
+- Invented lab measurements  
+- Medical claims, cure language, absolute safety guarantees  
+- Stuffing brand + model into every sentence  
+- Duplicating the same blog across similar SKUs with only the name swapped  
+
+## Porting to another site
+
+1. Copy enrichment types + `getEnrichment(slug)` helper.  
+2. Copy Product page sections (or shared component package).  
+3. Swap voice table + category FAQ seeds.  
+4. Run TinyFish research batch for that catalog.  
+5. Author enrichment file; ship via PR gate.
+
+## Success metrics (post-ship)
+
+- Organic landings on `/product/*` (impressions + CTR)  
+- Dwell time / scroll depth on PDPs  
+- Amazon click rate **after** scroll past blog (intent quality)  
+- FAQ expand events (if instrumented)

package.json

@@ -14,7 +14,8 @@
     "fill:quota": "node scripts/bsr/fill-quota.mjs",
     "refresh:images": "node scripts/bsr/refresh-catalog-images.mjs",
     "refresh:weekly": "npm run import:bsr && npm run build && wrangler deploy",
-    "deploy": "npm run build && wrangler deploy"
+    "deploy": "npm run build && wrangler deploy",
+    "content:research": "node scripts/research-product-enrichment.mjs"
   },
   "dependencies": {
     "lucide-react": "^1.25.0",