Two well-defined tools with complete input schemas, clear descriptions, and proper error handling. fetch_text and http_request both have detailed parameter documentation and structured output. Naming is verb-forward and action-clear. Main gaps: no output schema documentation in code, no tool annotations (readOnlyHint/destructiveHint), and parameter descriptions could be more prescriptive about constraints (e.g., URL format validation, timeout bounds). Error messages are actionable but could guide recovery more explicitly.

• Add tool annotations to tool definitions: FetchTextTool.readOnlyHint = true; HttpTool.destructiveHint = true. This signals to agents which tools are safe to retry and which modify state.
• Document output schemas formally. Create a Schemas.fetch_text_response and Schemas.http_request_response with zod definitions, then export them alongside inputSchema. Include field descriptions (e.g., 'truncated: boolean indicating if response was capped at max_chars').
• Enhance parameter descriptions with explicit constraints: 'timeout_ms: Request timeout in milliseconds (1 - 60000, default 10000)' and 'max_chars: Maximum characters to return (1 - 1000000, default 20000)'. Use the format 'must be X - Y' for numeric bounds.
• Improve error messages with recovery guidance. For timeouts, return 'Timeout after Xms. Retry with a longer timeout_ms or check if the server is slow.' For network errors, categorize: 'Network error (retryable): connection refused. Try again in a few seconds.'
• Add input validation for http_request headers and body. Reject header names with invalid characters; warn if body exceeds max_bytes before sending. Return clear validation errors like 'Invalid header name: contains newline character.'

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

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