25 tools with consistent naming (browser_* prefix) and basic schemas. All tools have descriptions (10-50 chars, below ideal 50-200 char range). Input schemas present but minimal, most parameters lack detailed constraints, enums, or format specifications. No output schemas documented. Error handling not visible in provided code. Parameter descriptions are terse (e.g., 'CSS selector or Playwright locator' repeated 8+ times). No tool annotations (readOnlyHint/destructiveHint). Risk classifications present but not integrated into descriptions. Composition is sound (single responsibility per tool), but descriptions don't guide LLM selection or explain when to use each tool vs. alternatives.

• Expand parameter descriptions to 50-100 chars with format/constraint guidance. E.g., 'CSS selector (e.g., #button-id, .class-name) or Playwright locator (e.g., page.locator("text=Click me")). Use CSS for simple queries, Playwright locators for complex matching.'
• Add enum constraints to parameters with fixed values: direction=[up|down|left|right], action=[accept|dismiss], browser=[chromium|firefox|webkit|chrome|msedge].
• Document output schemas for all tools. E.g., browser_screenshot returns {type: 'image', data: base64, width: number, height: number}; browser_get_text returns {text: string, selector: string}.
• Integrate tool annotations into descriptions. Prefix WRITE tools with '[Destructive]' and READ_ONLY tools with '[Read-only]'. Add idempotent hints for tools safe to retry (e.g., browser_navigate, browser_screenshot).
• Expand tool descriptions to 80-150 chars with WHEN/WHY guidance. E.g., 'Take a screenshot of the current page to verify visual state or capture evidence. Use before/after screenshots to confirm UI changes.' This helps LLMs select the right tool among similar options.