agentleFS
mantleCurvemantleCurve/claude-skills/skills/travel-journey-site/SKILL.md

travel-journey-site

Skill0 starsChanged 11 days agoMITView on GitHub

Turn a trip itinerary (from a poster image, a text plan, or a conversation) into a deployed, data-driven "living itinerary" website with a real map, distances from coordinates, train and flight timetables and fares, hour-aware Uber/cab estimates, traffic profiles, closed-day detection, climate normals, live weather and AQI, budget calculator and bookings checklist. Use when someone asks for an interactive travel plan, itinerary site, trip planner page, or to "replicate" the Agra–Delhi–Varanasi journey site for another trip.

  • Installs packages

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If the user can supply any illustration per city (their own photos work), cut them with layers.py or just save them as img/<cityKey>.webp …

    line 12
  2. Distinctive:

    If the input is an image, read it and transcribe every list before doing anything else.

    line 28
  3. Very specific:

    Search per item; a handful of queries per leg is normal.

    line 37
  4. Distinctive:

    Look up lat/lng for the home city, its airport, every city centre, airport, station, the place you'll sleep, and every sight.

    line 42
  5. Very specific:

    Run scripts/climate.py with each city's coordinates.

    line 45
  6. Very specific:

    If the user gave a poster, lift the artwork out of it with scripts/layers.py:

    line 57
  7. Distinctive:

    Keep blurbs original, short, and useful.

    line 96

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Travel journey site

Minimum input: a home city, two or more destinations, roughly how many days in each, and a rough month. Everything else (sights, fares, timetables, coordinates, climate, air) you research and fill in. With no poster there is no diorama art: the hero stage and chapter art collapse cleanly, and the page still carries the poster look through type, backdrop, embers and stamps. If the user can supply any illustration per city (their own photos work), cut them with layers.py or just save them as img/<cityKey>.webp with transparent surroundings.

You produce two files, index.html (a generic engine, copied from this skill unchanged) and data.js (everything trip-specific), test them headless, and deploy to GitHub Pages. The reference build is https://mantlecurve.github.io/agra-delhi-varanasi-journey/ ; its data file is template/data.example.js in this skill. Read that example before writing a new one.

1. Collect the itinerary

From the user's poster, text, or answers, establish:

  • home city and airport; each destination city with the days spent there
  • the sights per city (the poster's lists are the source of truth; do not invent extra stops)
  • the order of legs between cities and the intended mode (train, flight, cab)
  • departure date if known (otherwise pick a sensible default in the best season), travellers, ticket type (domestic vs foreign)

If the input is an image, read it and transcribe every list before doing anything else.

2. Research real facts (WebSearch / WebFetch), and record sources

For every inter-city leg: train number, departure and arrival times, running days, fare per class; flight duration and the current published fare range; cab fare and road distance. For every sight: entry fee for both ticket types, opening hours, closed day, typical visit length. For every city: airport and main station names, a typical hourly traffic profile (TomTom index, local reporting), UberGo-class rate card, notable airport-to-centre fare. Search per item; a handful of queries per leg is normal. Put each URL used into sources so the page's footer cites it. Fares change, so date the sources in the footer text.

3. Coordinates and climate

  • Look up lat/lng for the home city, its airport, every city centre, airport, station, the place you'll sleep, and every sight. The engine computes all distances from these (haversine × 1.3 inside cities, × 1.25 for rail, × 1 for flights). Get them right.
  • Run scripts/climate.py with each city's coordinates. It pulls three years of daily history from Open-Meteo's archive and prints the hi, lo and rain arrays you paste into climate. Set aqi (1 good … 5 very poor per month) from local reporting.
python3 ~/.claude/skills/travel-journey-site/scripts/climate.py "delhi 28.61 77.21" "agra 27.18 78.02"

3b. Visual layers (this is what keeps it from looking generic)

The engine expects img/backdrop.jpg, img/stone.jpg, img/skyline.jpg and one alpha WebP per city named img/<cityKey>.webp (the hero and each city chapter float these as parallax layers). If the user gave a poster, lift the artwork out of it with scripts/layers.py:

LO=50 HI=96 python3 ~/.claude/skills/travel-journey-site/scripts/layers.py poster.png site/img \
  "agra:35,275,435,730:45,288,170,405;45,545,205,712:42,292,431,650"

Each argument is name:cropbox:cut;cut...:keepbox in poster pixel coordinates. Cuts erase printed text so live HTML replaces it; the keep box drops everything outside the illustration (this is what removes the poster's card borders). Dark pixels become transparent, so the layer floats on any dark background. Always composite the results on a flat dark sheet and LOOK at them before using them. Make backdrop.jpg by blurring the whole poster heavily at low resolution, stone.jpg from a texture strip of its margins, skyline.jpg from its footer band (erase printed text by pasting plain patches). The template's img/ holds the reference set as a fallback when there is no poster.

Design rules that the template already follows, keep them: warm near-black ground, cream DM Serif Display titles, Bitter slab body at 18px+ and weight 500–800 (figures in Bitter 800), Kalam 700 script for stamps and taglines in saffron, hairlines not boxes, dioramas unboxed with drop shadows, embers, grain. Every metric gets an icon chip (chip('thermo', …) etc. in the engine). Text must read from across a room: nothing under 14px, body copy 18px, weights bold. Never use Inter, Roboto, Helvetica or other sterile sans faces, never glass cards or gradient blobs.

4. Write data.js

Copy template/data.example.js, keep the shape, replace the content. City keys must match the WebP file names. The chapter taglines live in the engine's CITYTAG map near renderDays; set one per city key. Key rules:

  • legs[].options[0] is the default; give trains a classes array with matching fare pairs.
  • days[].items use kind ∈ sight | ride | walk | train | flight | meal | free. Rides name from/to as a sight id or one of home, homeAirport, stay, hub, rail, rail2. A ride that happens in a city other than the day's city MUST set cityRef. fixed: [lo, hi] overrides the estimate for prebooked cabs. metro: true uses flat metro fares.
  • closed strings are matched against weekday names, so write "Friday", "Monday", "Never".
  • festivals[].m is 1-based month. packing lists are chosen by the month's heat, cold, rain, smog.
  • defaultStart is an ISO date. Pick one whose weekdays clear every closed rule and train running day.
  • Air: each city needs air.monthly (12 CPCB AQI averages, from CPCB bulletins or press coverage of monthly averages), air.fog (0–3 per month) and air.note. Top-level airHourly is the 24-hour multiplier (overnight inversion high, early afternoon low). The engine rates every outdoor hour, flags fog on early-morning sights, builds the Air section and pulls a live five-day forecast. For a smog-season trip, research the latest monthly averages; they move year to year.
  • Keep blurbs original, short, and useful. No filler.

Then copy template/index.html next to it. The hero title, the three <img src="img/…webp"> lines in the hero stage, the three .tag taglines, the CITYTAG map, the nav, the section copy and the <title> are the trip-specific text in the engine: replace them for the new cities (add or remove .dio blocks to match the city count). Leave the script logic alone.

5. Test headless before deploying

cd <site dir> && python3 -m http.server 8790 &
node ~/.claude/skills/travel-journey-site/scripts/check.js http://127.0.0.1:8790/

check.js needs playwright installed somewhere on NODE_PATH (the scratchpad from earlier sessions has it; npm i playwright otherwise) and a Chromium under ~/.cache/ms-playwright. It fails loudly on page errors and prints how many days and legs rendered, the map stats, closed-day clashes, and the budget total. Zero page errors and the right day count are required. Then take a couple of screenshots and look at them: hero, map, one day, weather.

Then run scripts/scroll-check.js <url>: it scrolls the page like a person WITHOUT overriding the reveal styles and reports anything still at opacity 0, the chapter order, and per-day opacity. The first check overrides reveal CSS to screenshot, so it cannot catch a block that never reveals.

Common failures: a ride whose from/to doesn't resolve (missing cityRef), a sight id typo, a leg fromPt/toPt key the city doesn't have, duplicate element ids, a scroll-reveal applied to a container taller than the viewport (it never reaches the threshold).

6. Deploy

bash ~/.claude/skills/travel-journey-site/scripts/deploy.sh <site dir> <github-owner>/<repo-name> "one-line description"

Creates the public repo, pushes main, enables Pages from the root, waits for the build, and curls the live URL plus data.js. Never push into an existing repo the user didn't name.

7. Report

Give the live URL first. Then, in plain prose, what the page does (map, legs, day-by-day, weather, traffic, budget, bookings) and which numbers were verified against which sources. Say clearly that ride fares are estimates from rate cards and traffic profiles, not live quotes.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.