price-compare

# Price Compare Compare prices for any product across major US retailers, with pros/cons, trust scores, and buy recommendations. > **Pattern: Inversion + Pipeline + Reviewer** > This skill uses structured questions before searching (Inversion), a multi-step workflow with gates (Pipeline), and a price trust scoring system (Reviewer). Follow the phases in order — do not skip ahead. --- ## Phase 1 — Pre-Search: Gather Requirements (Inversion) **DO NOT start searching until you have answered the questions below.** ### Required (ask if not clear from the prompt) - **Product**: What exactly do they want? (brand, model, size/quantity if relevant) - **Budget**: Is there a max price? (filters results before showing) - **Context**: Indoor or outdoor use? New or refurbished acceptable? (avoids irrelevant results) ### Ask only if relevant - **Timeline**: Need it today/this week, or can wait 2-3 weeks? - **Used/refurb OK?**: If asking about electronics, clarify new vs. refurbished preference - **Specific retailers?**: Any retailer they want to avoid or prefer? **If the user gives enough context** (e.g., "best price for Sony WH-1000XM5 under $300"), skip redundant questions and move to Phase 2. --- ## Phase 2 — Execute Search (Pipeline) ### Step 1 — Search All Platforms in Parallel Try `web_search` first. Run all calls simultaneously: ``` web_search(query="site:amazon.com {product}", count=5) web_search(query="site:walmart.com {product}", count=5) web_search(query="site:target.com {product}", count=5) web_search(query="site:bestbuy.com {product}", count=5) web_search(query="site:ebay.com {product}", count=5) web_search(query="site:temu.com {product}", count=5) web_search(query="site:costco.com {product}", count=5) web_search(query="{product} price comparison site:google.com/shopping", count=5) ``` ### Step 1 Fallback — When web_search fails If `web_search` returns errors (403/429/503) or empty results, use the browser instead: ``` browser(action="navigate", url="https://www.google.com/search?q={product}+price+site:amazon.com+OR+site:walmart.com+OR+site:target.com&num=10", profile="openclaw") browser(action="snapshot", profile="openclaw") // extract prices from results ``` If browser also fails, fall back to a single `web_fetch` on Google Shopping: ``` web_fetch(url="https://www.google.com/shopping?q={product}&hl=en", maxChars=10000) ``` ### Step 2 — Fetch Top Results (with Browser Fallback) For each platform that returned results, attempt `web_fetch`: **URL patterns:** - Amazon: `https://www.amazon.com/s?k={product}&s=price-asc-rank` - Walmart: `https://www.walmart.com/search?q={product}` - Target: `https://www.target.com/s?searchTerm={product}` - Best Buy: `https://www.bestbuy.com/site/searchpage.phtml?st={product}` - eBay: `https://www.ebay.com/sch/i.html?_nkw={product}&_sop=15` - Temu: `https://www.temu.com/search_result.html?search_key={product}` - Costco: `https://www.costco.com/search?search={product}` - Google Shopping: `https://www.google.com/shopping?q={product}&hl=en` **Browser fallback for JS-rendered pages:** If `web_fetch` returns a bot-check page or empty content for Amazon/Walmart/Target/Temu: ``` browser(action="navigate", url="<platform URL>", profile="openclaw") browser(action="act", kind="evaluate", fn="() => { const items = document.querySelectorAll('[data-component-type=\"s-search-result\"]'); return Array.from(items).slice(0,5).map(i => ({ title: i.innerText.substring(0,100), price: i.querySelector('.a-price .a-offscreen, [class*=\"price\"]')?.innerText || 'N/A' })).slice(0,5); }") ``` ### Step 3 — Load References Always load these reference files: ``` read(path="{skill_dir}/references/platforms.md") read(path="{skill_dir}/references/trust-checklist.md") ``` --- ## Phase 3 — Price Trust Scoring (Reviewer) Apply the trust score from `trust-checklist.md` to every price result. Flag suspicious prices before presenting. ### Trust Score Criteria (Reviewer Pattern) For each platform result, score: | Signal | Score | Meaning | |--------|-------|---------| | Official store / first-party seller | ✅ +2 | Authentic, full warranty | | 4.5+ stars, 500+ reviews | ✅ +1 | Popular, likely real | | Price history shown (Google Shopping) | ✅ +1 | Can verify if it's a good deal | | "was $X, now $Y" with X > 30% above market | ⚠️ -2 | Fake discount — X was never real | | No reviews / <10 reviews | ⚠️ -1 | Hard to verify quality | | Temu unbranded generic | ⚠️ -1 | Quality not assured | | Used/refurbished | ℹ️ ±0 | Normal risk, acceptable if disclosed | | eBay seller <95% rating | ⚠️ -2 | High return/defect risk | | "Only 2 left!" / countdown timers | 🚩 -1 | Dark pattern, ignore | | Price < 30% of average market | 🚩 -3 | Almost certainly fake/knockoff | ### Trust Score Totals - **+3 to +5**: 🟢 Reliable — show prominently - **+1 to +2**: 🟡 Acceptable — show with note - **0 to -1**: 🟠 Caution — show only if user OK with risk - **-2 or below**: 🔴 Suspicious — skip or warn explicitly --- ## Phase 4 — Price Anomaly Gate (Pipeline) Before presenting results, check for anomalies: **Gate: If any result is >50% below market average** → Flag as "⚠️ Suspicious cheap — likely fake/counterfeit. Verify before buying." **Gate: If Temu price is <30% of branded alternatives** → Add warning: "Temu's low price suggests unbranded/knockoff quality. Compare photos with official listing." **Gate: If price has moved >20% in 30 days** (from Google Shopping) → Note: "↗ Price up X% in 30 days — not the best time" OR "↓ Price down X% — good time to buy" **Gate: If web_search AND browser both failed for a platform** → Do NOT fabricate prices. Mark that platform as "❌ Data unavailable — results may be incomplete." --- ## Phase 5 — Output (Generator) Use this structure every time: ### 📦 [Product Name] **Quick summary:** [1 sentence: best pick + price range] **Trust-scored comparison:** | Platform | Price | Ship | Rating | Trust | Verdict | |---------|-------|------|--------|-------|---------| | ... | ... | ... | ... | 🟢/🟡/🟠/🔴 | ... | **Recommended:** 🥇 **Best overall:** [Platform] — $XX — [1 sentence why] 🥈 **Runner-up:** [Platform] — $XX — [1 sentence why] ♻️ **Best used/refurb:** [Platform] — $XX — [1 sentence why] 💸 **Budget pick:** [Platform] — $XX — [1 sentence why] **⚠️ Warnings & gotchas:** - [Any trust score issues, fake discounts, shipping gotchas] **📈 Price trend:** - [From Google Shopping: "↓ Good time — down 20% from 90-day average" OR "↗ Rising — up 15% in 30 days"] **🔗 Direct links:** - [Amazon](url) · [Walmart](url) · [Target](url) · [Best Buy](url) · [eBay](url) · [Temu](url) --- ## Output Examples ### Example — Basketball ``` ### 📦 Wilson NBA Official Game Ball (Size 7) **Quick summary:** Best authentic option is DICK'S at $219.99 (11% off); cheapest decent ball is Walmart's Wilson Prestige at $15.94 for outdoor use. | Platform | Price | Ship | Rating | Trust | Verdict | |---------|-------|------|--------|-------|---------| | DICK'S | $219.99 | Free | ⭐ 4.7 (129) | 🟢 | 🔥 Best Price + ⭐ Best Value — 11% off, 90-day returns | | Wilson.com | $219.95 | Free ($50+) | ⭐ 4.5 (64) | 🟢 | ⚡ Fastest — direct from manufacturer | | Scheels | $219.95 | — | — | 🟡 | Same price, reliable retailer | | NBA Store | $249.99 | Free ($50+) | ⭐ 4.6 | 🟢 | ❌ Overpriced — avoid | | eBay | $190.00 | +$7.95 | varies | 🟠 | ♻️ Bulk deal — verify seller rating 98%+ first | **⚠️ Warnings & gotchas:** - The official Wilson game ball is **genuine leather — for indoor use only**. New it is slippery; needs 2-3 weeks of break-in. - eBay at $190 is suspicious for new — verify authenticity before buying. - Temu had no relevant results (mostly unbranded generic balls under $10 — 🔴 low quality risk). **📈 Price trend:** → Stable — typically $220, within normal range for 90 days. **Recommended:** 🥇 **Best overall:** DICK'S Sporting Goods — $219.99 — matches all-time low, free shipping, 90-day returns. 🥈 **Runner-up:** Wilson.com — $219.95 — direct from maker, 30-day returns. 💸 **Budget outdoor:** Walmart — Wilson Prestige Outdoor — $15.94, pickup today. ``` --- ## Quick Rules - **Trust score every result** before showing it. Never present a 🔴 price without a warning. - **Always show total cost** (price + shipping). A "cheaper" item with $20 shipping may cost more. - **Parallelize all searches** — never run them sequentially unless using fallback. - **Flag fake discounts.** "Was $X" where X was never actually sold = -2 trust score. - **No fabricated prices.** If a platform's data is unavailable, say so. - **Direct links in every response.** Users need to be able to click and buy. - **Platforms:** Amazon, Walmart, Target, Best Buy, eBay, Temu, Costco, Google Shopping. Add Kroger for groceries.