Two tools with identical structure but critical gaps. Both have basic descriptions (60-70 chars, acceptable) and complete input schemas with proper types and enums. However, descriptions lack WHEN/WHY guidance and actionable context. No output schema documented. Parameters lack format constraints and validation hints. No error handling guidance. Tool names are verb-noun but generic ('generate_from_*'), does not clearly distinguish library vs studio templates to LLMs. API key exposed as parameter (critical security issue). No pagination, no result limits documented. Schemas present but descriptions are minimal for LLM decision-making.

• Move apiKey to server-side environment variable (ORSHOT_API_KEY). Remove from tool parameters. Use DEFAULT_API_KEY from config (already loaded in index.ts), do not expose to LLM.
• Document output schema in tool descriptions: 'Returns {success: boolean, data?: string (base64), url?: string, task_id?: string, status?: string}'. Include in code comments for clarity.
• Expand descriptions to 80-150 chars with WHEN/WHY: 'Generate an image from a pre-built library template. Use this for quick generation with minimal customization. Provide templateId and optional modifications (key-value pairs). Returns base64 data or download URL.'
• Add parameter constraints in descriptions: 'modifications: JSON object with template-specific keys (e.g., {"text": "Hello", "color": "#FF0000"}). See template docs for valid keys.' Validate and return clear errors if keys are invalid.
• Implement error recovery guidance: catch API errors and return 'Template not found. Available templates: [list]. Did you mean one of these?' or 'API rate limited (429). Retry in 60 seconds.' This guides LLM self-correction.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry