Loading ToolBench
Pure, modular, plugin-based video editing and creation toolkit. No orchestrator, no LLM, no autonomous pipeline — an external agent drives one plugin at a time.
Single tool 'voice.voicebox_server' has a comprehensive input schema with 12 well-typed parameters (action, backendDir, python, host, port, dataDir, modelsDir, waitMs, pollMs, probeTimeoutMs, baseUrl, out). However, the tool description is extremely generic ('Lifecycle control for the vendored Voicebox TTS backend...') and lacks actionable guidance on WHEN to use it, WHAT it returns, or HOW to interpret results. No output schema is documented. The action parameter uses an enum (start, stop, restart, status) which is good, but most other parameters lack sufficient context in their descriptions (e.g., 'Bind host' does not explain why an agent would override it, or what the default behavior is). The tool name 'voice.voicebox_server' is awkwardly namespaced and does not follow verb_noun convention (should be 'start_voicebox', 'stop_voicebox', etc., or a single 'control_voicebox' with action as enum). No error handling guidance is visible. No indication of what the tool returns or how the agent should interpret success/failure.
Lifecycle control for the vendored Voicebox TTS backend (vendor/voicebox/speech). Actions: start, stop, restart, status.
Tool name does not follow verb_noun convention. 'voice.voicebox_server' is a noun phrase; should be 'start_voicebox', 'stop_voicebox', 'restart_voicebox', 'get_voicebox_status' or a single 'control_voicebox' tool with action enum.
Tool description is generic and lacks actionable context. Does not explain WHEN to call it, WHAT it returns, or HOW to interpret results. Should state: 'Manages the Voicebox TTS backend lifecycle. Use to start the service before generating speech, check health status, or stop it when done. Returns current service state (running/stopped) and health details.'
No output schema documented. LLM cannot plan downstream actions or extract required fields. Must document what the tool returns for each action (e.g., status action returns {state: 'running'|'stopped', port: number, health: boolean}).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
Parameter descriptions are minimal and lack context. E.g., 'Bind host' does not explain why an agent would override it or what happens if omitted. Should be: 'Network interface to bind the Voicebox service to. Default 127.0.0.1 (localhost only). Set to 0.0.0.0 to allow remote connections.'
No error handling guidance. Tool description does not explain what errors are possible (e.g., port already in use, Python interpreter not found, backend directory missing) or how the agent should recover.