{
  "markdown": "# Pricewin Skills Hub\n\n**Find the cheapest hotel deal in one command.** AI-agent skills that compare\n**live** prices across Booking.com, Agoda, Google Hotels & OpenTravel and hand\nback ranked best-value / cheapest / quality picks with direct booking links.\n\n> ⭐ **Most popular skill: [`pricewin-hotel-deal-finder`](https://www.skills.sh/price-win/pricewin-skills-hub/pricewin-hotel-deal-finder)** — standalone, no MCP server, no API keys.\n>\n> ```bash\n> npx skills add https://github.com/Price-Win/pricewin-skills-hub --skill pricewin-hotel-deal-finder\n> ```\n\n```\n🏨 Tokyo • Aug 12–15 • 3 nights • 2 guests\n━━━━━━━━━━━━━━━━━━━━\n🥇 BEST VALUE   Shinjuku Granbell Hotel   ✅ agoda $118  ·  booking $131  → save $13\n🥈 CHEAPEST     APA Hotel Shinjuku        ✅ google $94\n📊 18 hotels | agoda, booking, google, opentravel • prices in USD\n```\n\nWhy people install it:\n- 🥇🥈🥉 best-value / cheapest / quality picks, side by side\n- 4 sources, all normalized to **USD**, with clickable links to the cheapest OTA\n- Works for **any city worldwide** — even bot-hardened ones (Shanghai, Hangzhou, Bangkok) via a stealth Patchright daemon\n- No MCP server, no API keys — just `node` + `npx`\n\n---\n\nSkills for AI agents to search hotels, compare OTA prices, and surface the best\nbooking deals. Two families:\n\n- **MCP skills** — drive the Pricewin MCP server (`pricewin`).\n- **Standalone skill** — `pricewin-hotel-deal-finder` runs its own browser + API stack,\n  no MCP server required.\n\n**Repo:** `git@github.com:Price-Win/pricewin-skills-hub.git`\n\n## Skills\n\n| Skill | Type | Trigger | Sources / Tools |\n|-------|------|---------|-----------------|\n| `pricewin-hotel-search` | MCP | search/find hotels live for travel dates | `search_hotels_live` → `poll_search_results` (Agoda + Booking.com + Traveloka, plus `opentravelResults` OpenTravel listings) |\n| `pricewin-price-comparison` | MCP | compare prices, check room rates for a named hotel | `search_hotels_live` → `poll_search_results`, `get_ota_hotel_detail`, `get_hotel_detail`, `get_cancellation_policy` |\n| `pricewin-booking-assistant` | MCP | recommend room, book + pay, manage a reservation | `get_hotel_detail` → `create_booking`, `check_booking_status`, `recreate_payment_link`, `request_cancel_token` → `cancel_booking` |\n| `pricewin-hotel-deal-finder` | Standalone | hotel deals / price comparison for travel dates | Browser automation (Patchright) over Booking.com + Agoda + Google Hotels, plus the **OpenTravel** public API — no MCP server |\n\n## `pricewin-hotel-deal-finder` (standalone)\n\nSelf-contained skill: a long-running Patchright (stealth Chromium) daemon scrapes\nBooking.com, Agoda, and Google Hotels, while a single HTTPS call hits the\nOpenTravel public API. One command does everything:\n\n```bash\nnode bin/search.js \"<city>\" <checkIn YYYY-MM-DD> <checkOut YYYY-MM-DD> <adults> en-us\n```\n\n- **OpenTravel API base:** `https://api.opentravel.one`\n  (override with `OPENTRAVEL_API_BASE_URL`)\n- **Setup:** `skills/pricewin-hotel-deal-finder/install.sh` (Node deps + Chromium, ~200MB one-time)\n- **Cache:** `~/.cache/pricewin-hotel-deal-finder/selectors.json`\n\nSee `skills/pricewin-hotel-deal-finder/SKILL.md` for full usage and rules.\n\n**Roadmap:** hotels and **flights** ship today (flights via `skills/pricewin-flight-search/`, MCP-driven); **rental-car** price comparison is on the way — the same one-command, multi-source, USD-normalized deal-finding, extended across your whole trip. (Powered by [PriceWin](https://price.win) — *tìm giá tốt nhất* across 50+ travel platforms.)\n\n### Install (agents)\n\nThe skill follows the [agentskills.io](https://agentskills.io) `SKILL.md` standard and\npublishes to both ecosystems. `node_modules` and Chromium are **not** bundled — the\ninstall hook fetches them on first install (needs network + `node`/`npx` on PATH).\n\n```bash\n# OpenClaw / ClawHub\nclawhub skill publish ./skills/pricewin-hotel-deal-finder \\\n  --slug pricewin-hotel-deal-finder --name \"PriceWin Hotel Deal Finder\" --version 0.7.0 --tags travel,hotel\nopenclaw skills install pricewin-hotel-deal-finder            # users / agents\n\n# Hermes — install straight from this repo (github: source pulls the whole dir)\nhermes skills install github:Price-Win/pricewin-skills-hub/skills/pricewin-hotel-deal-finder\n\n# Manual / dev\ngit clone git@github.com:Price-Win/pricewin-skills-hub.git\ncd pricewin-skills-hub/skills/pricewin-hotel-deal-finder && ./install.sh\n```\n\n## Security & data handling\n\nEvery skill ships a `SECURITY.md` next to its `SKILL.md` — what it executes, what\nleaves the machine, and who operates the backend:\n\n| Skill | Runs code? | Data it sends | Disclosure |\n|---|---|---|---|\n| `pricewin-hotel-deal-finder` | Yes — local Patchright daemon | search query only (city, dates, guests) | [SECURITY.md](skills/pricewin-hotel-deal-finder/SECURITY.md) |\n| `pricewin-hotel-search` | No | search query | [SECURITY.md](skills/pricewin-hotel-search/SECURITY.md) |\n| `pricewin-price-comparison` | No | search query + `queryText` | [SECURITY.md](skills/pricewin-price-comparison/SECURITY.md) |\n| `pricewin-flight-search` | `install.sh` only (MCP registration) | route + dates | [SECURITY.md](skills/pricewin-flight-search/SECURITY.md) |\n| `pricewin-booking-assistant` | No | **guest PII** (name, phone, email) — real bookings | [SECURITY.md](skills/pricewin-booking-assistant/SECURITY.md) |\n\nThe deal-finder's local daemon is loopback-only, token-authenticated (`0600`\nstate file), rejects non-loopback `Host` headers, and keeps the Chromium sandbox\non. No skill ever handles card numbers: payment happens on the provider's page.\nPrivacy policy: <https://price.win/en/privacy-policy>.\n\n## FAQ\n\n**Is Booking.com or Agoda cheaper?** Neither wins universally — it depends on the hotel, dates, and region. Independent tests (CNBC's review of ~200,000 hotel searches) found **Agoda had the lowest rate about 34% of the time** globally, and it tends to win in **Asia-Pacific** while **Booking.com pulls ahead more often in Europe**. That is exactly why `pricewin-hotel-deal-finder` queries **both** (plus Google Hotels and OpenTravel) in a single run, shows the per-night gap, labels the cheapest source per property, and links you straight to it — so you never have to guess.\n\n**How many sources does it compare, and how fast?** Up to **4 sources** (Booking.com, Agoda, Google Hotels, OpenTravel), all normalized to **USD**. A cached city returns in **~30–60 seconds**; a brand-new city pays a one-time Agoda discovery cost of **2–4 minutes**, then joins the cache.\n\n**Do I need API keys or an MCP server?** No. `pricewin-hotel-deal-finder` is standalone — `node` + `npx` is all it needs. The only external dependency is Chromium (~200 MB), fetched once by the install hook.\n\n**Does it work for bot-hardened cities?** Yes. A stealth Patchright (stealth Chromium) daemon handles anti-bot detection, so cities like Shanghai, Hangzhou, and Bangkok resolve where vanilla scrapers get blocked.\n\n**How are prices normalized?** Booking.com returns USD natively; Agoda, Google Hotels, and OpenTravel geo-lock to VND by IP and are converted via a live FX rate. Every price in the output is USD.\n\n**What does one command return?** Ranked **best-value 🥇 · cheapest 🥈 · quality 🥉** picks, each hotel name a clickable link to its cheapest OTA, plus a \"More good deals\" list grouped by source and a USD footer of exactly which sources returned data.\n\n> _Last verified: 2026-07-08 · reviewed quarterly._\n\n## MCP skills — prerequisites\n\n- Pricewin MCP server running (`pricewin-mcp`, stdio transport, auto-launched by Claude)\n- Backend API accessible\n\n```json\n{\n  \"mcpServers\": {\n    \"pricewin\": {\n      \"command\": \"node\",\n      \"args\": [\"/path/to/pricewin-mcp/dist/index.js\"]\n    }\n  }\n}\n```\n",
  "bytes": 7732,
  "sha": "583f2c441a09e5d7c05f5d4fd4a5f82a35158a85e574d18d19bbf97a7548c64a",
  "repo_slug": "price-win/pricewin-skills-hub",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/skl_price_win_pricewin_skills_hub_pricewin_h_857af4db/readme"
}