CodeSense has 2 tools with basic definitions but significant quality gaps. sayHello is a trivial greeting with minimal utility. getDocs has a reasonable description (73 chars) but both tools lack comprehensive parameter descriptions, output schema documentation, and error handling guidance. Input schemas are present but sparse. No tool annotations (readOnlyHint, destructiveHint). Descriptions fall short of the 50-200 char LLM-optimized baseline. Parameter descriptions are minimal or absent. No documented output structure. Error paths (e.g., 'No sources found' returning empty array) lack recovery guidance.

• Add comprehensive parameter descriptions to getDocs: 'libraryName (string, required): The name of the library or framework (e.g., "react", "django", "numpy"). Used to search for official documentation.' and 'langName (string, required): The programming language for code generation (e.g., "javascript", "python", "typescript"). Constrains search results to language-specific docs.'
• Document the output schema explicitly in the tool description: 'Returns either: (1) markdown string if docs URL found, (2) string content if GitHub repo found, (3) array of SearchResult objects if no suitable URL found. Each SearchResult has {title, link, snippet}.'
• Add error recovery guidance: When no sources found, return error message 'No documentation found for {libraryName} in {langName}. Try: (1) verify library name spelling, (2) use a more specific query (e.g., "react hooks" instead of "react"), (3) ask the user for a direct documentation URL.'
• Remove sayHello or repurpose it. If kept, add a meaningful description explaining when an agent should call it (e.g., 'Verify server connectivity and readiness before attempting documentation lookups.').
• Add tool annotations to schema: set readOnlyHint=true for both tools to signal they are safe to call without side effects.