llm-wiki has 10 tools with consistent naming (wiki_* prefix) and complete input schemas. All tools have descriptions (avg 95 chars, within baseline 34-392). However, descriptions are terse and lack LLM-optimized guidance on WHEN to use each tool or what to do with results. Output schemas are documented in code but not exposed in tool definitions. Error handling is present but minimal, most tools return JSON errors without recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite 9 read-only and 1 write tool. Parameter descriptions are adequate but lack format constraints (e.g., 'page_type' enum values not enforced in schema). wiki_new is the only write tool but lacks confirmation/dry-run pattern.

• Add readOnlyHint: true to all 9 read-only tools and destructiveHint: true to wiki_new. This enables LLMs to reason about safety and plan accordingly.
• Expand tool descriptions to 100-150 chars with explicit WHEN/WHY guidance. E.g., wiki_search: 'Search the wiki by keyword using BM25 or semantic search. Use this first to find relevant pages, then call wiki_expand to read full content.'
• Add enum constraints to schema for page_type and severity parameters. Change from type:string to type:string with enum: ['entity', 'concept', 'source', 'answer', 'all'].
• Implement a wiki_new_preview tool or add a dry_run parameter to wiki_new so agents can see what will be created before committing. Include a confirmation step for force=true.
• Enhance error responses with recovery guidance. E.g., 'Page not found: xyz. Try wiki_search("xyz") to find similar pages, or wiki_list() to see all available pages.'