Three tools with clear verb-noun naming (search_sessions, list_sessions, restore_session) and reasonable descriptions (120-180 chars). All tools have input schemas with types and descriptions. However, output schemas are not documented, responses are inferred from code inspection. Parameter descriptions are present but lack detail on constraints, formats, and error recovery. No error handling guidance for LLMs. Missing idempotency hints for restore_session (a write operation). Descriptions do not explain when to use each tool vs. alternatives or what happens on failure.

• Document output schemas for all three tools. Example: search_sessions returns {results: [{id: string, project: string, matches: number, snippets: string[], title?: string}], total: number}. Include field types and meanings.
• Add min/max constraints to numeric parameters: max_results (1-100, default 10), context_snippets (1-10, default 3). State these in parameter descriptions.
• Enhance restore_session description: 'Restores a session by creating/updating its registration file in Claude Desktop. Idempotent, safe to call multiple times. Returns success or error with recovery steps (e.g., if session not found, call list_sessions to verify ID).'
• Add error recovery guidance to all tools. Example: 'If search returns no results, try list_sessions to verify sessions exist. If restore_session fails with 'file write error', check that ~/.claude directory is writable.'
• Clarify tool selection in descriptions. search_sessions: 'Use when you know a keyword to find. Slower but precise.' list_sessions: 'Use to browse recent sessions or get an overview. Faster, no search needed.'