Travel Guidebook is a JavaScript-based AI agent skill hosted on GitHub at geekjourneyx/travel-guidebook, with 55 stars. It takes a single sentence describing a trip—such as "北京到上海，5.1 到 5.5，五日深度游" (Beijing to Shanghai, May 1 to 5, five-day in-depth tour)—and outputs a complete, typeset PDF travel guidebook around 30 pages and 7MB in size, plus a self-contained HTML version. The project handles everything from research to final export using parallel AI agents and optional integration with Amap maps for real-time data like POIs and routes.

Unlike basic template fillers or data dumps, it produces guides with a distinct aesthetic: parchment-style backgrounds, serif title fonts, inline SVG decorations like compass icons and day badges, and structured color coding for info blocks. Each guide feels like a printed booklet, complete with daily journal spaces outlined for notes or ticket stubs.

Core features

The skill runs a multi-stage process powered by agents:

  • Parallel research agents gather itinerary details, weather, attractions, food, and lodging in tandem.
  • Amap MCP integration (optional) pulls live POI images, addresses, ratings, walking routes, and transit options from Gaode Maps.
  • Editorial writing crafts narrative descriptions with literary flair, like drop-cap initials for route intros.
  • Typography-driven layout uses custom CSS for elements such as double-line borders, functional color systems, and printable PDF export via Playwright/Chromium.
  • Comprehensive sections cover overview, prep checklists, daily breakdowns, themed essays, and appendices.

These combine into a 6-stage pipeline: environment checks, input parsing, parallel research, content synthesis, HTML rendering, and PDF generation.

Getting it running

Setup requires Node.js version 18 or higher. First, install Playwright for PDF export:

npm install playwright
npx playwright install chromium

Add the skill via Copilot CLI:

npx skills add https://github.com/geekjourneyx/travel-guidebook

For enhanced map data, configure the Amap MCP server in ~/.copilot/mcp-config.json:

{
  "mcpServers": {
    "amap-maps": {
      "command": "npx",
      "args": ["-y", "@amap/amap-maps-mcp-server"],
      "env": {
        "AMAP_MAPS_API_KEY": "your_amap_api_key"
      }
    }
  }
}

Obtain a free Web Services API key from the Gaode Open Platform.

The project files include SKILL.md for agent definitions, references/ with CSS norms (layout-css.md), chapter templates (chapter-templates.md), and report templates (report-template.md), plus scripts/html2pdf.mjs for export.

Usage in practice

Once installed, trigger it in Copilot CLI or Claude Code by describing a trip. The skill auto-detects keywords like "路书" (travel notes), "旅行指南" (travel guide), or "行程规划" (itinerary planning). Examples:

北京到上海，5.1 到 5.5，五日深度文化游

成都到重庆，自驾三日游，美食为主

杭州周末两日游，带老人和小孩

Processing yields files like beijing_shanghai_guidebook.html (82KB) and beijing_shanghai_guidebook.pdf (7.1MB) in the current project directory.

What the output includes

Guides follow a fixed structure for print-ready use:

Section Key elements
Cover Compass SVG, route title with tags, double-line frame
Journey overview Days, mileage, season, budget; core highlights
Pre-trip prep Documents/gear/weather; budget tables (economy/mid/high-end); app recs
Daily chapters Day badge SVG; schedule cards (transit/weather); narrative routes; POI cards (images/address/rating); food (photos/avg price/signature dishes); lodging; tips; dotted journal space
Themed essays 2-3 pieces on history, culture, local views
Appendix Emergencies/phrases; expense tables; playlist
Back cover Journey reflections

This format suits printing or digital reading, with self-contained HTML for previews.

Design details

The aesthetic draws from a defined system: parchment base, serif headings, inline SVGs, and five principles (visible in project images). Colors distinguish info types—e.g., transport in blue, food in warm tones. CSS in references/layout-css.md handles drop caps, badges, and responsive cards.

Who this fits

Frequent travelers or planners who want personalized, visually appealing guides beyond generic apps like TripAdvisor or Google Maps itineraries. It's ideal for multi-day trips with specifics like self-driving (Chengdu-Chongqing) or family outings (Hangzhou with elders/kids). Users comfortable with AI tools like Copilot CLI benefit most, as it automates research into a polished artifact. Output quality depends on agent prompts and Amap data for accuracy.

Those needing offline basics without AI might skip it; it's tied to agent ecosystems and requires setup.

The MIT-licensed source lives at github.com/geekjourneyx/travel-guidebook, including full templates for tweaks.

View on GitHub →