This is a CLI tool wrapped as an MCP server with 5 tools. All tools have descriptions and input schemas are visible in source. However, schemas lack proper JSON Schema structure with explicit type declarations and constraints. Parameter descriptions are present but often generic (e.g., 'URL or domain'). No output schemas documented. Error handling descriptions mention exit codes but lack recovery guidance. Tools are read-only (good) but lack idempotent/destructive annotations. Average tool score: 42/100.

• Add explicit JSON Schema type, enum, minimum, maximum, and pattern fields to all parameters. E.g., 'min-score' should be {type: 'string', pattern: '^(\d|[1-9]\d|100)$', description: 'Exit 1 when score below threshold (0-100)'}.
• Document output schemas for each tool. E.g., 'Returns {score: number, layers: [{checks: [{id, status, message}]}], summary: string}'.
• Enhance error descriptions with recovery guidance. E.g., 'Exit 3: API error. Check ORA_API_KEY is set and valid; retry with --force to bypass cache.'
• Clarify parameter formats and constraints in descriptions. E.g., 'url: HTTPS URL (e.g., https://docs.example.com); required' and 'max-age: Cache window in seconds (e.g., 3600 for 1 hour)'.
• Rename 'journey' to 'run_agent_journey' or 'send_agent_to_site' to clarify action. Add description distinguishing it from 'deep-journey': 'Runs a curated intent; use deep-journey for free-text tasks with a partner API key.'