WikipediaAgent has 4 tools with basic descriptions and minimal parameter schemas. All tools have descriptions (10-200 chars range), but parameter documentation is sparse. Input schemas exist but lack depth: most parameters are simple strings with minimal constraint documentation. No output schemas are documented. Error handling is present but generic ('Error generating answer: {exc}'). Tool naming follows verb_noun convention (wikipedia_search, wikipedia_ask, wikipedia_get_article, wikipedia_index_article), which is good. However, the server is STDIO-only, which is a hard transport limitation. Descriptions are adequate but could be more specific about when to use each tool vs. alternatives. No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles.

• Add output schemas to all tools. Document the structure of returned results (e.g., wikipedia_search returns array of {title, text, chunk_index} objects). This enables LLMs to extract and chain data.
• Add parameter constraints: 'query' should specify max length (e.g., 100 chars), 'title' should clarify exact article title format (e.g., 'Python (programming language)' not 'python'). Use descriptions to state these rules.
• Add tool annotations: mark wikipedia_search, wikipedia_ask, wikipedia_get_article with readOnlyHint=true; mark wikipedia_index_article with destructiveHint=true. This signals to agents which tools are safe to retry.
• Improve error messages with recovery guidance. E.g., 'Article not found. Try wikipedia_search() to find the correct title, then retry with the exact title.' This guides the agent's next step.
• Add selection guidance in descriptions. E.g., 'Use wikipedia_search for exploratory queries; use wikipedia_ask for direct factual questions; use wikipedia_get_article when you know the exact title.'