Story Forge has 2 tools with basic descriptions and minimal parameter documentation. Both tools accept a single string parameter ('query') with adequate descriptions, but lack structured output schemas, error handling guidance, and parameter constraints. Naming follows verb_noun convention (get_*, generate_*), which is good. However, descriptions are functional but generic (194-char baseline suggests these are below average). No input validation, enum constraints, or recovery guidance. The tools are simple wrappers around underlying functions, which is acceptable, but the MCP interface lacks production-grade polish.

• Add explicit output schema to both tools. Document that get_latest_info_mcp returns a string summary with source attribution, and generate_video_script_mcp returns a ~200-word script string. Consider returning structured JSON with fields like {summary: string, sources: [{title, url}]} for better chaining.
• Add error handling with recovery guidance. Wrap Tavily and OpenAI calls in try-except blocks that return structured errors: {error: string, retryable: bool, suggestion: string}. E.g., 'Tavily API timeout. Retryable. Try again in 5 seconds.'
• Constrain the 'query' parameter. Add description: 'User query (1-500 characters). Examples: "latest AI trends", "stock market update". Avoid special characters; alphanumeric + spaces preferred.'
• Document pagination and limits. Update get_latest_info_mcp description to state: 'Returns up to 5 summarized results from web search. No pagination available; results are pre-filtered by relevance.'
• Add tool annotations (readOnlyHint) to both tools in FastMCP registration to signal they are safe, non-destructive operations.