Three tools with complete input schemas and detailed descriptions. All tools follow verb_noun naming (oxylabs_scrape, oxylabs_google_search, oxylabs_amazon_product). Descriptions are 200-300 chars, well above the 10-char minimum and within the 194-char baseline average. All required parameters are documented. However, output schemas are completely undocumented, the code comments mention response envelope structure but no formal schema is exposed to the LLM. Error handling is present (oxylabsError function) with HTTP status-specific guidance, but lacks structured error classification (retryable vs fatal). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all three being read-only operations. Parameters lack enum constraints where applicable (e.g., domain in oxylabs_amazon_product could be an enum). No pagination support documented, though MAX_CONTENT_CHARS cap suggests awareness of response size limits.

• Add formal output schema documentation for each tool. Example for oxylabs_scrape: { type: 'object', properties: { content: { type: 'string', description: 'Page HTML (capped at 20K chars)' }, status_code: { type: 'number' }, url: { type: 'string' } } }
• Add readOnlyHint: true to all three tool definitions to signal safety and enable agent optimizations.
• Convert 'domain' parameter in oxylabs_amazon_product to an enum: ['com', 'co.uk', 'de', 'co.jp', 'fr', 'it', 'es', 'nl', 'se', 'ca', 'au', 'in', 'br', 'mx', 'jp', 'cn'] with description 'Amazon marketplace domain suffix'.
• Document the 20K character cap for oxylabs_scrape in the tool description: 'Raw HTML is capped at 20,000 characters; longer pages are truncated.'
• Enhance error messages with explicit recovery guidance: 'HTTP 401: Invalid credentials. Verify _apiKey format is "username:password" from https://dashboard.oxylabs.io. Retry with corrected credentials.' (already partially done; formalize as structured error classification).