Three tools with complete schemas and descriptions, but significant gaps in parameter constraints, output documentation, and error guidance. edit_video has 12 parameters with minimal validation hints; check_status and afterai_info lack output schema documentation. No enums for constrained fields (colorCorrectionStyle, subtitleStyle, spokenLanguage, subtitleLanguage). Error messages are helpful but lack recovery guidance for some failure modes. Naming is clear (verb_noun pattern), but parameter descriptions could be more prescriptive about formats and constraints.

• Add enum constraints to edit_video: colorCorrectionStyle: enum(['1', '2', '3', ...]), subtitleStyle: enum(['1', '3', ...]), spokenLanguage: enum(['en', 'bg', 'es', ...]), subtitleLanguage: enum(['same', 'en', 'es', 'fr', ...]). Document valid values in tool description or via API reference.
• Document output schemas for all tools. For check_status, declare: {headline: string, percent: number, sub?: string, drive_link?: string}. For afterai_info, declare: {api: object} or {reference: string}. For edit_video, declare: {jobId: string, taskId?: string, trackUrl: string, remaining: number}.
• Add format/constraint hints to edit_video parameters: spokenLanguage='ISO 639-1 code (2 chars)', watermarkText='max 50 chars', notes='max 500 chars'. Use parameter descriptions to guide LLM input validation.
• Enhance error messages with recovery steps. For 402: 'Out of credits. Subscribe at https://getafterai.eu/#pricing, then retry with the same videoUrl.' For check_status not found: 'Job not found. Verify the job ID from edit_video response, or check your email for completion notifications.'
• Add idempotency guidance to edit_video description: 'Submitting the same videoUrl and parameters multiple times will create separate jobs. Use check_status to verify completion before resubmitting.' This prevents accidental duplicate edits.