Two tools with minimal definitions. Both have basic descriptions (60-80 chars, below the 194-char baseline) and simple input schemas with single string parameters. No output schemas documented. No error handling guidance. No parameter validation constraints (enums, ranges, patterns). Descriptions lack WHEN/WHY context and dependency hints. No tool annotations (readOnlyHint present in metadata but not in schema). Naming is acceptable (verb_noun pattern: tavily_search, read_page) but descriptions are generic and under-optimized for LLM selection.

• Expand descriptions to 150-250 chars. Include WHAT (searches web/reads page), WHEN (when search snippets insufficient/need full context), and WHAT IT RETURNS (list of results with URLs/snippets or full page text).
• Document output schemas. tavily_search should return {results: [{title, url, snippet, source}], total_count}. read_page should return {content, url, status_code}.
• Add parameter constraints: query min 1 char, max 500; url must be valid HTTP(S) URL with format validation.
• Add error guidance: 'If tavily_search returns no results, try a simpler query or different keywords. If read_page times out (>30s), the URL may be unreachable or very large.'
• Expose tool annotations in schema via readOnlyHint: true so LLMs see these are safe, idempotent read operations.