14 tools with complete schemas and descriptions. Naming follows verb_noun pattern (check_*, run_*, crawl_*). Descriptions are detailed (100-300 chars) and explain WHAT/WHEN/HOW. All parameters typed with descriptions. Error handling is sophisticated (describeError maps 15+ error codes to actionable messages). However, output schemas are not formally documented in the tool definitions, responses are described in prose rather than JSON Schema. No pagination parameters on crawl_audit despite potentially returning many results. Some parameter descriptions could be more concise per LLM optimization guidelines (50-200 chars ideal).

• Add formal JSON Schema definitions for tool outputs. Document return types for seo_full_audit (Markdown structure, JSON object fields, HTML file path), crawl_audit (aggregated issues, counts), and run_lighthouse (Core Web Vitals metrics, scores).
• Add pagination to crawl_audit: include 'limit' (default 50, max 500) and 'offset' parameters; return 'total_pages_found' and 'next_offset' in response. Document: 'Large sites may require multiple calls.'
• Condense parameter descriptions to 50-200 chars. E.g., seo_full_audit.output: 'Report format: markdown (inline), json (structured), or html (shareable file).' instead of current 180+ char version.
• For check_keyword_placement, clarify matching behavior: 'Exact phrase match, case-insensitive. Density = (keyword_count / total_words) * 100. Stemming not applied.'
• Add error recovery hints to tool descriptions. E.g., check_https: 'If certificate validation fails, the site may be using a self-signed cert or incomplete chain, see error message for remediation steps.'

Inferred effective spec: 2025-06-18+.

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