Loading ToolBench
Significant gaps across naming, descriptions, and schemas. Tool names lack clear action verbs (buildVideo, buildWebtoon, confirmCharacter use camelCase inconsistently). Descriptions are present but generic or incomplete: 'Confirmcharacter tool description' is a placeholder; 'Updatescene tool description' is similarly unhelpful. Parameter descriptions are minimal (e.g., 'udpate scene' has a typo). Schemas use z.any() for speech_bubbles, defeating validation. No output schemas documented. Error handling is minimal, most tools lack recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All 7 tools are WRITE operations but descriptions do not state this clearly.
Create an .mp4 video with 2 seconds per scene image using ffmpeg concat mode.
Combine scenes vertically with alternating 30%-width speech bubbles and wrapped text.
Confirmcharacter tool description
Create a character image via ComfyUI WebSocket
Create a scene image using multiple character images and a prompt.
Update character image using existing one and new prompt
Updatescene tool description
Placeholder descriptions: 'Confirmcharacter tool description' and 'Updatescene tool description' provide zero context for LLM tool selection. These violate the pattern:tool-description requirement that every tool must have a non-empty, meaningful description (10 - 1024 chars).
Untyped parameter: buildWebtoon's speech_bubbles uses z.any().optional(), which defeats schema validation. LLMs cannot infer whether to pass an array, JSON string, or object. This violates pattern:constrained-input.
No output schemas documented. Tools return objects (e.g., video_url, frame_count, webtoon_id) but LLMs have no formal schema to plan downstream calls or extract fields. Violates pattern:tool requirement for documented return types.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
No error handling guidance. Tools throw generic errors (e.g., 'ffmpeg execution failed: ...') without recovery hints. LLMs cannot determine if the error is retryable, user-fixable, or fatal. Violates pattern:recovery-guide.
No tool annotations. All 7 tools are destructive (WRITE operations: create, update, build) but lack destructiveHint annotations. LLMs cannot distinguish safe reads from risky writes. Violates pattern:tool-annotation.