Server has 5 tools with complete input schemas using Zod validation and clear descriptions. Naming follows verb_noun convention (get_*, search_*, install_*). However, descriptions are brief (10-50 chars for most tools), lacking LLM-optimized context about WHEN to use each tool and what happens next. Output schemas are not documented, callers must infer structure from code. Error handling is minimal; no recovery guidance or actionable error messages. Tool composition is reasonable but search_ready_directory lacks pagination documentation despite accepting a limit parameter.

• Document output schemas for all tools. Add a comment block or JSDoc showing the structure returned by each tool, e.g., 'Returns {content: [{type: "text", text: string}]}'.
• Expand tool descriptions to 50 - 150 characters. Include WHEN to use the tool and what the LLM should do next. E.g., 'get_agent_install_rule: Returns a system prompt rule that coding agents should follow when auto-installing ContinueWith. Use this to configure agent behavior before calling install_continuewith_widget.'
• Add actionable error messages to install_continuewith_widget. Catch common failures (npx not found, directory not found, framework not recognized) and return 'Framework not recognized. Supported: html, nextjs, astro. Did you mean: nextjs?' instead of raw stderr.
• Implement pagination for search_ready_directory. Add a 'cursor' or 'offset' parameter to the response and accept it as input. Document in the tool description: 'Returns up to limit results. Use cursor from response to fetch next page.'
• Sanitize arguments passed to spawn(). Use an allowlist for framework, mode, layout values. Validate siteKey against a regex (alphanumeric + underscore). Reject any input containing shell metacharacters (|, &, ;, $, etc.).