Loading ToolBench
An MCP server that transcribes audio or video from URLs using OpenAI's Whisper model. Downloads media, converts to WAV, and performs transcription.
Single tool with minimal schema and weak descriptions. TranscriptTool has a basic input schema (file_path: string) but lacks output schema documentation, parameter constraints, and error handling guidance. Description is 287 chars (within range) but generic, does not explain when to use it vs alternatives, prerequisites (ffmpeg, torch), or expected output format. No enum constraints, no format specifications, no recovery guidance for common failures (file not found, conversion errors, device unavailability). Tool name lacks action verb prefix (should be 'transcribe_audio' or 'transcribe_file'). Error handling returns plain strings instead of structured error objects with actionable recovery steps.
A smolagent tool for transcribing audio and video files into text. This tool utilises Whisper for transcription and ffmpeg for media conversion, enabling agents to process multimedia inputs into text. The tool supports robust file handling, including format conversion to WAV and dynamic device selection for optimal performance.
Tool name lacks action verb. 'TranscriptTool' should be 'transcribe_audio' or 'transcribe_file' to signal intent to LLMs.
No output schema documented. Tool returns string but LLMs cannot infer structure (plain text? JSON? error format?). Must document return type and example structure.
Parameter 'file_path' lacks format constraints. No description of accepted formats (mp3, wav, mp4, etc.), path resolution behavior, or size limits. LLMs will guess.
Error handling returns plain strings ('Error: File not found...') instead of structured errors with recovery guidance. No distinction between retryable vs fatal errors.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | <=2025-11-25 | v2 |
Description does not state prerequisites (ffmpeg, torch, CUDA optional) or expected output format. LLMs cannot determine if tool is suitable without explicit guidance.