# Panel Hop

> A fan convention schedule planner. Paste the programme (day | start-end | room | title) and tag each session must,
> want, maybe or fixed (a ticket or appointment you hold, like an autograph or photo-op slot). The browser lays every
> day out as a TV-guide grid - rooms as rows, half hours across - and flags clashes, hall changes with no time to walk
> or queue, days with no lunch or dinner gap and days longer than you want to stand. A metered planner picks one path
> through the weekend with arrival times, drops with a backup (usually a later repeat), meal breaks and a
> fits / trim / overbooked verdict. https://panel-hop.skillsafe.ai/

## What is free (runs in the browser)

- Reading the programme: pipe-, tab- or comma-separated lines (a spreadsheet CSV export with separate start and end
  columns works), day or date headers, 24-hour or am/pm times, `start-end` or `start +minutes`, tags (must / want /
  maybe / fixed, `!!` / `!` / `?` shorthand), and sessions that run more than once (same title = a repeat).
- The guide: one grid per day, sessions coloured by tag, a "Starred" strip that shows every overlap of your starred
  sessions in red, and a faint wash over the lunch (11:30-14:30) and dinner (17:30-20:30) windows.
- Flags: clash (fix when both sessions are must or fixed and no repeat can settle it), tight hop (less time than the
  walk plus the must-see queue), no meal, long day, hard hours, maybe overlaps, repeats; a page reading of fits / trim /
  overbooked.
- Day cards: hours of sessions, time on your feet against your limit, lunch and dinner gaps.
- Re-plan loop: paste the final programme over a saved weekend and the page names what is new, gone or moved, and
  puts your old tags back on sessions that came in untagged (matched by day and title).
- Exports: a share-sheet send of the path (with the .ics where the device allows), Markdown, a path CSV, an .ics calendar file (set the first day's date), and the day's guide as SVG.

## The metered plan (task `plan`)

Input (scalars; objects JSON-encoded as strings): `task`, `today`, `programme`, `facts`, `session_list`,
`page_reads`, `page_verdict`, `notes`. Reply: one JSON object with `title`, `verdict` (fits | trim | overbooked),
`summary`, `path` [{session, arrive, leave, says, why}], `drops` [{session, says, why, instead}], `breaks`
[{day, from, to, kind: meal | rest, why}], `tips`.

The page re-checks every reply: ids must be real and used once, every starred session must be placed or dropped,
quotes must appear in the programme as written, and clashes, walks, meal gaps and day length are re-run on the minutes
the path actually occupies (arrive to leave). Lost must-sees and fixed slots, moves to a repeat, late joins and a
verdict that does not match the path are all named.

## Rules of thumb it uses

- Fixed slots first, then must-sees, wants, maybes. A repeat later in the weekend is the cheapest way to settle a
  clash.
- Arrive at a must-see your queue time early; leave the walk between rooms; never miss more than half of a session.
- Find a real gap for lunch on any day that spans 13:00, and for dinner on any day that spans 19:00.
- Times and rooms are the programme as pasted; walking, queueing and meal minutes are the attendee's own settings, not
  measurements of any venue.

## Sources

- Built on the SkillSafe skill @onewave-ai/itinerary-optimizer: https://skillsafe.ai/skill/@onewave-ai/itinerary-optimizer
- API tutorial: https://panel-hop.skillsafe.ai/api.html
- Why now: anime convention searches up 49% year on year (https://www.risingtrends.co/trends/activities-hobbies-trends-2026);
  2026 conventions reported record turnstile counts and 1,400+ hours of programming in one weekend
  (https://www.comicshoplocator.com/2026/07/12/anime-expo-2026-record-422000-attendance-mainstream/).
- Sample conventions, guests and sessions are made up.