Single tool 'crawl_website' has a complete input schema with 6 parameters, all typed and described. Tool description is adequate (145 chars, within 10-1024 range). However, output schema is undocumented, the tool returns a plain string with no structured format specified. Parameter descriptions are present but generic; they lack actionable constraints (e.g., max_depth range, delay bounds, mode enum values). No error handling guidance, the catch-all exception handler returns a bare error message without recovery hints. Tool name 'crawl_website' is clear and verb-based, but the description does not explain WHEN to use this tool vs alternatives, or what the output structure contains. Missing: pagination support (no limit on crawled_pages), idempotency guarantees, and permission/scope declarations. The tool performs a destructive write operation (creates files) but has no dry-run or confirmation pattern.

• Convert output to structured JSON: {success: boolean, page_count: integer, output_dir: string, index_path: string (if mode='crawl'), urls_path: string (if mode='explore'), error: string (if failed)}. Document this schema in the tool description.
• Add enum constraint for 'mode' parameter: mode: enum ['crawl', 'explore']. Add min/max bounds: max_depth (1-10), max_pages (1-1000), delay (0.1-60.0). Update parameter descriptions to include these constraints.
• Enhance error handling: categorize errors as retryable (network timeout, rate limit) vs user-fixable (invalid URL, permission denied). Return {success: false, error_code: string, error_message: string, recovery_hint: string}.
• Add dry_run parameter (boolean, default false) to preview crawl without writing files. Return estimated page count and output directory path for confirmation.
• Document result limits in tool description: 'Returns up to max_pages crawled pages. If limit reached, use output_dir to access partial results. For large sites, increase max_pages or run multiple crawls with different URL subsets.'