mcp-rss has clear, action-verb naming (fetch_, search_, get_, list_, add_, remove_) and all 6 tools have descriptions. Input schemas are present with type definitions. However, parameter descriptions are minimal (most under 50 chars), output schemas are not documented, and error handling lacks recovery guidance. The server follows basic tool structure but misses LLM-optimization patterns for descriptions and lacks composition guidance (e.g., which tool to call first, how results chain). No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles.

• Add outputSchema to each tool definition documenting the response structure. E.g., fetch_feed returns {title: string, description: string, link: string, entryCount: number, entries: [{id, title, link, pubDate, author, summary}]}.
• Expand parameter descriptions to 50-100 chars with format/constraint details. E.g., 'url: The RSS/Atom feed URL (must be http:// or https://, valid URI format)' and 'query: Search keyword(s) (1-200 chars, case-insensitive, matches title/summary/content/author)'.
• Add tool annotations: mark fetch_feed, search_entries, get_entry, list_subscriptions with readOnlyHint=true; mark add_subscription with idempotentHint=true (same URL+label = no-op); mark remove_subscription with destructiveHint=true.
• Enhance error messages with recovery guidance. E.g., 'Entry not found: {id}. Try calling search_entries() to find entries by keyword, or fetch_feed() to list all entries in the feed.'
• Add pagination to fetch_feed and search_entries. Accept optional limit (default 20, max 100) and offset parameters. Return total count and next_offset in response. Document in description: 'Returns up to 20 entries per call; use offset to paginate large feeds.'