Single tool 'generate' has a complete input schema with nested objects and enums, but critical gaps undermine quality. Tool description is generic (18 chars: 'Gherkin feature generation route'), below the 50-200 char LLM-optimized baseline. Parameter descriptions exist but are minimal (e.g., 'Feature title (required)' lacks format/constraint guidance). Output schema is undocumented, LLMs cannot plan downstream use. No error handling guidance, no recovery paths. Tool naming lacks action verb clarity ('generate' is vague; 'create_gherkin_feature' would be clearer). No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Risk marked READ_ONLY but not enforced in schema or description.

• Expand tool description to 80-150 chars: 'Generate a Gherkin feature file from a title, optional description, tags, and one or more scenarios. Each scenario contains Given/When/Then/And steps. Returns formatted Gherkin syntax ready for test automation.'
• Document the output schema explicitly: 'Returns {feature: string} containing formatted Gherkin syntax. Example: Feature: Login\n Scenario: Valid credentials\n Given user is on login page\n When user enters valid credentials\n Then user is logged in'
• Add format/constraint guidance to parameters: title (1-100 chars, no special chars), tags (array of 0-10 strings, each 1-50 chars, alphanumeric + hyphen), scenarios (array of 1-50 objects, each with name and 1-20 steps)
• Implement structured error responses with recovery guidance: 'Missing required field: title. Provide a non-empty string (1-100 chars).' and 'Scenarios array is empty. Provide at least one scenario with name and steps array.'
• Rename tool to 'create_gherkin_feature' and add tool annotation: readOnlyHint=false (since it generates output, not modifying state), idempotentHint=true (same input always produces same output)