# Liner Clues > A host's kit for a guess-the-artist music round. List tonight's artists (one per line, optional notes after a |); a metered lane writes five name-free sound clues per artist, hardest first, and the page plays them off the back of a record sleeve with a scoreboard (5 points for a guess on the first clue, down to 1 on the last) and a podium. https://liner-clues.skillsafe.ai/ ## What it does - Free, in the browser (cluekit.js, sleeveview.js): the list is cleaned and capped at 8 rounds, duplicates dropped, the sound traits in each note counted, one-word names without notes flagged. The game reveals one clue at a time, awards points to the player or team who guessed, marks rounds nobody got, keeps a scoreboard with a dot per round, and draws a podium when every playable round is over. Generated cover art per round (it never hints at the answer) and a record that slides out with the answer on its label. - Paid lane `deck`: one round per artist with a status - `ready` (an act the model knows: five clues across rhythm, harmony, instruments, production, genre fusion, energy, then the decade), `notes_only` (an act it does not know, but your notes describe two or more sound traits: two to five clues from your notes alone), `need_notes` (unknown or a name several acts share, with too little in the notes: no clues, a request for what to add). Each clue is marked as from your notes or from general knowledge, plus a one-line description of the sound, short phrases usable as an AI-music prompt, and the clue the host should double-check. - The page re-checks every reply: a word of the artist's name in a clue, capitalised words that may be names, numbers that are not decades, internal field names in prose, the clue count and sources each status allows, and rounds missing or out of order. - Exports: a Markdown host sheet, the deck as JSON, the scores as CSV, the sound of a round as an AI-music prompt. Saved rounds (the list, the deck and the scores) live in the person's SkillSafe account (collection `decks`). ## Contract - Run input: `task` (`deck`), `level` (`casual` | `mixed` | `nerd`), `crowd`, `artists` (JSON array of {n, name, notes, note_traits}), `page_flags` (JSON). See https://liner-clues.skillsafe.ai/api.html - Reply: one JSON object {task, title, summary, rounds: [{n, name, status, dna, clues: [{rung, dimension, text, source}], prompt_elements, host_check, ask}]}. ## Caveats - Clues marked "check" come from the model's general knowledge and can be wrong about a sound; read them before the party. Artist names are only answers to guess; the app is not affiliated with any artist. ## Sources - @jwynia/musical-dna (describe an artist's sound across six dimensions without naming them) - the clue deck.