A Model Context Protocol server for controlling Spotify playback and searching tracks
This server has significant quality gaps across naming, descriptions, and schema documentation. While 6 tools are defined with basic input schemas, the descriptions are extremely brief (mostly under 30 chars), parameter descriptions are missing entirely, output schemas are not documented, and error handling lacks recovery guidance. The tools themselves follow reasonable verb_noun naming (play_track, pause_track, etc.), but lack the depth needed for LLM selection and safe composition. Tool composition is reasonable, each tool has a single responsibility, but the lack of explicit output documentation and error guidance prevents reliable chaining.
Get information about the currently playing track
Skip to the next track on Spotify
Pause playback on Spotify
Play a track on Spotify by track URI
Skip to the previous track on Spotify
Search for tracks on Spotify
Four tools (pause_track, next_track, previous_track, get_current_track) have empty input schemas (no parameters documented), but this is not an immediate blocking issue since they accept no parameters. However, two of these lack any parameter description text for clarity.
All tool descriptions are critically brief (under 50 characters). Per the rubric baseline (avg 194 chars, p10=34, p90=392), descriptions should be 50 - 200 chars to guide LLM selection. Examples: 'Pause playback on Spotify' (27 chars) tells the agent WHAT but not WHEN to call it or what it returns. Should expand to explain state change, return value, and prerequisites.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 35 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 38 | - | v1 |
No output/return schemas documented for any tool. LLMs cannot know what fields to expect from get_current_track or search_tracks, making it impossible to plan downstream calls or extract the right data. For example, search_tracks should document: returns array of track objects with [track_id, track_name, artist, album, duration_ms, uri] and pagination info if applicable.
Parameter 'q' in search_tracks lacks a description. Per the rubric, every parameter must have a description explaining what it controls, 'q' alone does not clarify whether it searches track names, artists, albums, or all. Should read: 'Search query string (searches track names, artists, and albums)'.
Parameter 'track_uri' in play_track has a description but lacks format/example guidance. LLMs often pass invalid URIs. Should expand: 'The Spotify URI of the track to play (format: spotify:track:ID, e.g. spotify:track:11dFghVXANMlKmJXsNCQvf). Can be obtained from search_tracks results.'
No error handling guidance in tool descriptions. If play_track fails because no device is active, or search_tracks returns empty results, LLMs have no recovery path. Descriptions should include: 'If a device is not active, an error will be returned, ensure a device is active before calling. Call get_current_track to check device status.' and 'If no tracks are found, the agent should try a simpler or broader search query.'
No pagination support visible for search_tracks. If a search returns hundreds of results, the response could blow the context window. Per the rubric baseline, paginated results should accept limit/offset and return a total_count or next_cursor. search_tracks should document: 'Returns first 20 results by default. Use limit and offset parameters for pagination.'
Tool composition: search_tracks returns track data (presumably including track_uri), which should be directly passable to play_track. This is good. However, without documented output schemas, the agent cannot confirm this chain will work. get_current_track should return currently_playing track object with track_uri to enable downstream calls.
No state management or idempotency guidance. If play_track is called twice with the same track_uri, does it restart the track or do nothing? If pause_track is called when already paused, is it idempotent? Descriptions should clarify: 'Idempotent: calling twice with the same input produces the same effect as calling once.'