Three tools with complete input schemas and descriptions, but significant gaps in output documentation and error handling. Tool names lack action verbs (axure_fetch, axure_summary_prompt are noun-heavy). Parameter descriptions are present but generic. No documented output schemas, error recovery guidance, or idempotency guarantees. axure_health is trivial (empty input schema). The server targets a narrow domain (Axure prototypes) with reasonable parameter constraints (timeouts, limits), but lacks production-grade error handling and response structure documentation.

• Rename tools to start with action verbs: 'fetch_axure_content', 'build_axure_summary', 'check_axure_health'. This signals intent to LLMs and improves tool selection accuracy.
• Document output schemas explicitly. Define the structure of AxureFetchResult and the summary prompt format (fields, types, examples). Include this in tool descriptions or a separate schema reference.
• Add error handling with recovery guidance. E.g., 'If timeout occurs, increase timeoutMs or reduce maxImages. If OCR fails, disable enableOcrFallback.' Categorize errors as retryable vs. user-fixable.
• Enhance parameter descriptions with format constraints. E.g., 'ocrLanguage: Tesseract language codes separated by '+' (e.g., chi_sim+eng, eng, fra). See https://github.com/naptha/tesseract.js for valid codes.'
• Remove or repurpose axure_health. If kept, make it return useful diagnostics (Playwright version, Tesseract availability, memory usage) rather than a static 'ok: true' response.