19 tools with consistent verb_noun naming (get_, search_, update_, add_, remove_, list_). All tools have descriptions (avg ~80 chars, baseline 194 chars). Input schemas present with typed parameters and descriptions. However, output schemas are not documented in the source code, LLMs cannot see what fields to expect from responses. Parameter descriptions are minimal (often 1-2 sentences) and lack constraint details (ranges, enums, formats). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Error handling and recovery guidance absent from tool definitions.

• Document output schemas for all tools. For search tools, specify: returned_count (int), items (array of {full_name, primary_source, sources[], open_url}). For stat tools, specify field names and types.
• Add tool annotations: mark all READ_ONLY tools with readOnlyHint=true; mark WRITE tools with destructiveHint=true; mark idempotent tools (those with dry_run support) with idempotentHint=true.
• Enhance parameter descriptions with constraints: 'limit: Maximum results (1-50, default 20)'; 'query: Search string (1-500 chars, required)'; 'scope: One of: starred, knowledge, all'.
• Add error recovery guidance to tool descriptions: 'If search returns no results, try broader query or different scope. If repo not found, call global_search_repos first.'
• Document dry_run behavior explicitly: 'Set dry_run=true to preview changes without applying. Useful for validating changes before commit.'