| name | spotify |
| description | Control Spotify playback — play/pause, next/previous track, search the catalog, add tracks to playlists, and set volume. Mock backend (binding-only) for the thesis benchmark; no device required. |
Spotify Skill
Use this skill for music playback control and library actions. Most commands are pure
reflexer actions: one request → one playback command. Search is the read action.
Mock / binding-only. The spotify CLI here is a mock that echoes a JSON status line
without touching a real Spotify device or the Web API — we score command correctness,
not real playback (the developer has no device). To go live, back it with the Spotify Web
API (/me/player/...) and an OAuth token.
Commands (via bash)
CLI: python3 agents/skills/spotify/spotify.py <command> [args] (aliased spotify).
Playback control (the verb is the discriminator)
spotify play [--track "<query>"] — start/resume, optionally a specific track
spotify pause
spotify next
spotify previous
Search the catalog (read)
spotify search --query "<artist/track/album>"
Add a track to a playlist
spotify add --track "<track>" --playlist "<playlist>"
Set volume
spotify volume --level <0-100>
Flags
--track "<...>" track/artist to play or add--query "<...>" search terms (search)--playlist "<...>" target playlist (add)--level <0-100> volume percentage (volume)
Output format
One JSON line: {"ok": bool, "data": {...}, "error": null|{...}, "meta": {"mock": true}}.
Exit 0 on success, 2 on bad args.
Sibling families (why this skill is in the benchmark)
play / pause and next / previous are minimal-pair playback siblings — same topic,
opposite action. They extend the action-discrimination stress test (sibling collision)
beyond Home Assistant into a second domain.
