Connects Claude to Kling AI for video and image generation
Three tools with basic schemas and descriptions, but significant gaps in parameter documentation, output schema clarity, and error handling guidance. Tool names follow verb_noun convention (generate_*, check_*), which is good. However, descriptions lack actionable context for LLM selection, parameter descriptions are minimal, and output schemas are not formally documented. No input validation guidance, no error recovery hints, and no structured output specification. Baseline: avg tool description ~100 chars (below 194 baseline); most params lack detailed constraints or format guidance.
Check the status of a generation task.
Generate an image using Kling AI.
Generate a video using Kling AI.
Output schemas not documented. Tools return free-text strings (e.g., 'Video generation started! Task ID: {task_id}') instead of structured objects. LLMs cannot reliably extract task_id or parse status responses without explicit schema.
Parameter descriptions lack format/constraint guidance. 'duration' accepts 5 or 10 but description does not state this constraint. 'aspect_ratio' lists examples but no enum. 'quantity' says '1-4' in description but no min/max in schema. LLMs will guess invalid values.
Error handling does not guide recovery. Tools return raw error strings (e.g., 'API error: 401 - Unauthorized') without actionable next steps. No distinction between retryable (rate limit) vs user-fixable (invalid API key) vs fatal errors.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
Tool descriptions lack context for LLM selection. 'Generate a video using Kling AI' does not explain when to use generate_video vs generate_image, what prerequisites exist (API key setup), or what the return value enables (polling with check_task). Descriptions are 30-40 chars; baseline is 194.
No input validation or sanitization visible. 'prompt' parameter accepts any string with no length limit, injection guards, or format validation. LLMs could pass extremely long or malicious prompts without server-side checks.