Imported from ariesyous/imaxatscotia (AGENTS.md). Install upstream with npx skills add ariesyous/imaxatscotia. Copyright stays with the author.

AGENTS.md

Guidance for AI coding agents working in this repo.

What this project is

A static GitHub Pages site showing IMAX showtimes at four Cineplex theatres in the GTA (Scotiabank Theatre Toronto, Winston Churchill & VIP, Mississauga Square One, Vaughan). No backend, no build step. See README.md for the full architecture.

Layout

• index.html, style.css, app.js — the frontend. Plain HTML/CSS/vanilla JS on purpose; don't introduce a build step or framework for this. app.js builds a theatre tab bar from data/theatres.json (Scotiabank Theatre Toronto is the default shown on load) and re-renders the calendar from the selected theatre's data file when a tab is clicked.
• scripts/scrape.mjs — Node script (built-in fetch only, no dependencies) that hits Cineplex's undocumented theatrical API once per theatre listed in data/theatres.json and writes each theatre's own data/<slug>.json file.
• data/theatres.json — checked-in manifest of tracked theatres (Cineplex theatreId, slug, display name, output file). Add an entry here to track a new theatre; both the scraper and the frontend read from it.
• data/*.json (one per theatre, e.g. data/vaughan.json) — generated data, committed to the repo so the page has content without a build/deploy step. Treat these as artifacts, not something to hand-edit.
• .github/workflows/scrape.yml — runs the scraper daily and on workflow_dispatch, commits the data/*.json files when they change.

Constraints to respect

• The Cineplex API is unofficial and reverse-engineered (no public docs, key lives in Cineplex's own JS bundle). Don't assume its shape is stable — the scraper is written to skip bad/missing dates rather than overwrite good data with empty results, and to skip writing a theatre's file (leaving its last known-good data untouched) if every date fails for that theatre, without aborting the other theatres. It only exits non-zero if every theatre fails entirely. Preserve that fail-soft behavior in any changes.
• Showtimes aren't published on a contiguous rolling window — big releases get scattered advance-sale dates far in the future (see DAYS_AHEAD in scripts/scrape.mjs). Don't "optimize" the scraper to stop at the first empty date; it has to probe the whole window.
• No build step. Don't add bundlers, TypeScript compilation, or frameworks unless explicitly asked — the site is meant to stay simple enough to deploy by just pushing to master.
• GitHub Pages serves the master branch root directly. Any file at repo root is publicly served as-is.

Testing changes

• Run node scripts/scrape.mjs locally to regenerate data/showtimes.json and sanity-check the output.
• Serve the repo root with any static file server (e.g. npx serve or python -m http.server) and open it in a browser to check the UI.
• There is no automated test suite.