跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
其他技能

inflight-movie-picker

Recommend what to watch on a flight from the airline's in-flight catalogue, ranked by IMDb rating and the user's taste. Use when the user asks what to watch on their flight or plane, what movies are on Emirates / ice, United, KLM, British Airways, Singapore Airlines, Lufthansa, JAL, Cathay Pacific, Qantas, ANA or American Airlines this month, for plane movie picks, or whether a specific film is available on a given airline. Also triggers on "inflight movie picker", "seatback cinema", "what's on the plane" and "in-flight entertainment" questions.

len5ky · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add len5ky/seatback-cinema --skill inflight-movie-picker

技能说明


name: inflight-movie-picker description: Recommend what to watch on a flight from the airline's in-flight catalogue, ranked by IMDb rating and the user's taste. Use when the user asks what to watch on their flight or plane, what movies are on Emirates / ice, United, KLM, British Airways, Singapore Airlines, Lufthansa, JAL, Cathay Pacific, Qantas, ANA or American Airlines this month, for plane movie picks, or whether a specific film is available on a given airline. Also triggers on "inflight movie picker", "seatback cinema", "what's on the plane" and "in-flight entertainment" questions.

Inflight movie picker

Local monthly catalogues for Emirates, United, KLM, British Airways (ba), Singapore Airlines (singapore), Lufthansa, Japan Airlines (jal), Cathay Pacific (cathay), Qantas, ANA and American (american), matched to IMDb and ranked by a blend of IMDb, Letterboxd, RT and Metacritic plus the user's Letterboxd taste and taste.json. Run everything from the project root /home/lensky/Dev/plane_movie with bun (~/.bun/bin on PATH). bun run cli with no arguments lists the registered airline ids.

Answering "what should I watch on my flight?"

  1. Find out the airline and, if it matters, the flight month (catalogues change monthly).
  2. Make sure data is fresh, then rank:
    bun run cli stats --airline united                  # snapshot for this month? match rate?
    bun run cli fetch united [--month YYYY-MM]          # only if the month is missing (cached per month)
    bun run cli match --airline united                  # after any fetch or imdb sync
    bun run cli picks --airline united --top 15 [--genre Thriller] [--min-rating 7] [--language English] [--min-year N] [--max-year N] [--year N]
    
    picks defaults to Emirates when --airline is omitted.
  3. Present 5-10 picks with one line each: why it fits (genre, rating), runtime (fits the flight?), IMDb link.
  4. For "is X on the plane?": bun run cli search "<title>" searches every fetched airline (or --airline X).

bun run cli imdb sync reloads IMDb (downloads ~1.2 GB if files are older than 7 days, ~3 min load). Only run it when stats shows 0 IMDb titles or the user asks for fresher ratings.

Taste and ratings

The simplest taste input is the user's IMDb export: imdb.com > Your Ratings (and Watchlist) > three-dot menu > Export, CSVs dropped into data/imdb-ratings/, then bun run cli taste import (joined on tconst, no title matching). A Letterboxd export works too. If picks prints "taste: taste.json only", offer this:

  1. Letterboxd > Settings > Data > Export your data; drop the zip into data/letterboxd/.
  2. bun run cli taste import (re-run after each new export; it replaces the previous import and lists unmatched films). Watched and rated films are then excluded, watchlist films are boosted, and genre/director affinities learned from the ratings show up as "you rate X +0.7 (3 films)" reasons.

Extra rating sources (optional, slow the first time): bun run cli ids sync, then bun run cli ratings sync [--source letterboxd|rt|mc|tmdb] [--airline X]. Scraped sites go at 1 request/s with a 30-day cache, so a first full run is about an hour per source; scope it with --airline or --limit N. TMDB needs TMDB_API_KEY in .env (free key from themoviedb.org > Settings > API) and is skipped otherwise. Picks then blend IMDb, Letterboxd, RT audience and Metacritic; each pick shows its LB, RT and MC scores on its first line.

taste.json stays as manual overrides: liked_genres, disliked_genres (IMDb names such as Sci-Fi, Thriller, Horror), min_rating, min_votes, seen (titles or tt ids, excluded from picks), favourites (titles; their directors get a small boost). When the user says they have seen a film, add it to seen.

Caveats to tell the user when relevant

  • All sources are undocumented airline endpoints: personal, low-volume use. Never fetch the same airline-month more than once a day (raw responses are cached under data/raw/; --refresh forces a refetch). Never loop retries, especially against BA (Akamai).
  • BA, Singapore and Qantas only publish the current cycle, so next month cannot be fetched for them until it starts. The others accept a future --month. Lufthansa, JAL and ANA make one detail call per film (~130-410 requests, cached), so do not refetch them casually.
  • Qantas is the large international seatback catalogue; domestic and smaller aircraft differ.
  • Matches are graded high/medium/low/none from title, year, director and runtime. Picks use only high and medium. Emirates has no year field, so its regional-language films and concert specials match least often.
  • A catalogue can differ by route, aircraft or country; only the airline's public web catalogue is fetched.