MCP server for controlling audio playback
This server has three tools with significant quality gaps. All three have basic descriptions (30-40 chars), which is below the 50-200 char baseline for LLM-optimized descriptions. The input schemas are present but minimal: list_audio_files and stop_playback have empty input objects ({}), while play_audio has only a single 'filename' parameter with no constraints, validation rules, or format guidance. Tool descriptions lack actionable context about when to use them, what they return structurally, or how to handle errors. Parameter descriptions are missing entirely (play_audio's 'filename' has only 'The audio file to play', no guidance on format, validation, or error recovery). No output schemas are documented, forcing LLMs to infer response structure. Error handling is present in the code but not surfaced in the tool definitions, LLMs cannot predict failure modes or recovery paths from the schema alone. The tools are well-named (verb_noun pattern) and focused (single responsibility), but the definitions lack the depth required for confident LLM selection and invocation.
List all available audio files in the audio directory
Play an audio file
Stop playback
Tool descriptions are too brief (25 - 30 characters) and lack actionable context. They do not explain when to call each tool, what it returns structurally, or what can go wrong.
play_audio parameter 'filename' has no format constraints, validation rules, or error recovery guidance. The description 'The audio file to play' does not tell LLMs whether to pass a full path, just the name, or a pattern; what extensions are valid; or how to handle missing files.
No output schemas are documented for any tool. LLMs cannot predict the response structure (what fields are returned, types, or how to extract values for downstream tools).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 50 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 29 | - | v1 |
Error handling logic exists in the code (try/except blocks, ctx.error calls) but is not exposed in the tool schemas. LLMs cannot plan for failure modes or know when to retry vs ask the user for help.
list_audio_files and stop_playback have empty input schemas ({}), which is correct, but this pattern is not explicitly documented in the tool descriptions, LLMs should be told these tools take no parameters.